The latest articles on DEV Community by ogc16 (@ogc16). https://dev.to/ogc16 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F885223%2F4a51669c-3728-4f5f-bdba-03e0b0da756f.png https://dev.to/ogc16 en ogc16 Thu, 27 Aug 2026 12:37:17 +0000 https://dev.to/ogc16/uca-oaua-oaoe-aeu-ee-oaaoi-oaoaoe-ao-4l4d https://dev.to/ogc16/uca-oaua-oaoe-aeu-ee-oaaoi-oaoaoe-ao-4l4d <p>╪▒╪│╪º┘ä╪⌐ ╪¿╪¡╪½┘è╪⌐</p> <h2> ╪º┘ä┘à┘ä╪«╪╡ </h2> <p>┘ä┘à ┘è╪╣╪» ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ê╪╣╪»┘ï╪º ╪¿╪╣┘è╪»┘ï╪º. ╪Ñ┘å┘ç ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è ╪º┘ä╪ó┘å. ╪¬┘ü╪¡╪╡ ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ┘â┘è┘ü ╪¬╪╣┘à┘ä ╪ú╪»┘ê╪º╪¬ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪╣┘ä┘ë ╪Ñ╪╣╪º╪»╪⌐ ╪¬╪┤┘â┘è┘ä ╪º┘ä╪¬╪╣┘ä┘è┘à ΓÇö ┘à┘å ╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬╪»╪▒┘è╪│ ╪º┘ä╪┤╪«╪╡┘è ╪º┘ä╪¬┘è ╪¬╪¬┘â┘è┘ü ┘à╪╣ ┘â┘ä ┘à╪¬╪╣┘ä┘à╪î ╪Ñ┘ä┘ë ╪º┘ä╪¬╪╡╪¡┘è╪¡ ╪º┘ä╪░┘â┘è ╪º┘ä╪░┘è ┘è┘é╪»┘æ┘à ╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘ü┘è ╪½┘ê╪º┘å┘ì╪î ┘ê╪╡┘ê┘ä┘ï╪º ╪Ñ┘ä┘ë ╪º┘ä╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪ú╪«┘ä╪º┘é┘è╪⌐ ╪º┘ä╪╡╪º┘à╪¬╪⌐ ╪º┘ä╪¬┘è ╪¬╪▒╪º┘ü┘é ┘â┘ä ╪º╪¿╪¬┘â╪º╪▒.</p> <p>╪º┘ä╪▒╪ú┘è ┘ç┘å╪º ┘ä┘è╪│ ╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪│┘è╪¡┘ä ┘à╪¡┘ä ╪º┘ä┘à╪╣┘ä┘à┘è┘å. ┘ä┘å ┘è╪¡╪»╪½ ╪░┘ä┘â. ╪¿╪»┘ä┘ï╪º ┘à┘å ╪░┘ä┘â╪î ╪¬┘é╪»┘æ┘à ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ╪º╪»╪╣╪º╪í┘ï ╪ú┘â╪½╪▒ ╪»┘é╪⌐: ╪Ñ┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è╪î ╪Ñ╪░╪º ╪º╪│╪¬┘Å╪«╪»┘à ╪¿╪¡┘â┘à╪⌐╪î ┘è┘à┘â┘å┘ç ╪ú┘å ┘è╪¡╪▒┘æ╪▒ ╪º┘ä┘à╪╣┘ä┘à┘è┘å ┘ä┘è┘é┘ê┘à┘ê╪º ╪¿┘à╪º ┘ä╪º ╪¬╪│╪¬╪╖┘è╪╣ ╪º┘ä╪ó┘ä╪º╪¬ ┘ü╪╣┘ä┘ç ΓÇö ╪º┘ä╪Ñ╪▒╪┤╪º╪» ┘ê╪º┘ä╪Ñ┘ä┘ç╪º┘à ┘ê╪º┘ä╪¬┘ê╪º╪╡┘ä. ╪║┘è╪▒ ╪ú┘å ┘ç╪░╪º ╪º┘ä┘ê╪╣╪» ┘è╪╣╪¬┘à╪» ┘â┘ä┘è┘ï╪º ╪╣┘ä┘ë ╪º┘ä╪╖╪▒┘è┘é╪⌐ ╪º┘ä╪¬┘è ┘å╪«╪¬╪º╪▒ ╪¿┘ç╪º ╪¿┘å╪º╪í ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘ê┘å╪┤╪▒┘ç╪º.</p> <p>╪¬╪▒╪│┘à ╪º┘ä┘ü╪╡┘ê┘ä ╪º┘ä╪¬╪º┘ä┘è╪⌐ ┘à╪│╪º╪▒┘ï╪º: ┘à┘å ╪º┘ä╪¼╪░┘ê╪▒ ╪º┘ä╪¬╪º╪▒┘è╪«┘è╪⌐ ┘ä┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐╪î ┘à╪▒┘ê╪▒┘ï╪º ╪¿╪º┘ä┘à╪┤┘ç╪» ╪º┘ä╪¡╪º┘ä┘è ┘ä╪ú╪»┘ê╪º╪¬ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è╪î ┘ê╪╡┘ê┘ä┘ï╪º ╪Ñ┘ä┘ë ╪º┘ä╪ó╪½╪º╪▒ ╪º┘ä┘é╪º╪¿┘ä╪⌐ ┘ä┘ä┘é┘è╪º╪│ ╪╣┘ä┘ë ╪º┘ä╪¬╪╣┘ä┘à╪î ┘ê╪ú╪«┘è╪▒┘ï╪º ╪Ñ┘ä┘ë ╪º┘ä╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪ú╪«┘ä╪º┘é┘è╪⌐ ┘ê╪º┘ä╪│┘è╪º╪│┘è╪⌐ ╪º┘ä╪¬┘è ┘ä╪º ┘è┘à┘â┘å ┘ä╪ú┘è ╪╖╪º┘ä╪¿ ╪ú┘ê ┘à╪╣┘ä┘æ┘à ╪ú┘ê ╪╡╪º┘å╪╣ ┘é╪▒╪º╪▒ ╪ú┘å ┘è╪¬╪¼╪º┘ç┘ä┘ç╪º.</p> <h2> ┘ü┘ç╪▒╪│ ╪º┘ä┘à╪¡╪¬┘ê┘è╪º╪¬ </h2> <ol> <li>╪º┘ä┘à┘é╪»┘à╪⌐</li> <li>┘å╪¿╪░╪⌐ ╪¬╪º╪▒┘è╪«┘è╪⌐ ┘à┘ê╪¼╪▓╪⌐ ╪╣┘å ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐</li> <li>┘â┘è┘ü ┘è╪╣┘à┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è</li> <li>╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪┤╪«╪╡┘è ┘ê╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐</li> <li>╪º┘ä╪¬┘é┘è┘è┘à ┘ê╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪º┘ä╪ó┘ä┘è╪º┘å</li> <li>╪»┘ê╪▒ ╪º┘ä┘à╪╣┘ä┘à ┘ü┘è ╪╣╪╡╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è</li> <li>┘é┘è╪º╪│ ┘à╪º ┘è┘å╪¼╪¡: ╪º┘ä╪ú╪»┘ä╪⌐ ┘ê╪º┘ä┘å╪¬╪º╪ª╪¼</li> <li>╪º┘ä╪ú╪«┘ä╪º┘é ┘ê╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ┘ê╪º┘ä╪¬╪¡┘è┘æ╪▓</li> <li>╪º┘ä╪│┘è╪º╪│╪⌐ ┘ê╪º┘ä╪¬┘å┘ü┘è╪░</li> <li>┘à╪│╪¬┘é╪¿┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à</li> <li>╪º┘ä╪«╪º╪¬┘à╪⌐</li> <li>╪º┘ä┘à╪▒╪º╪¼╪╣</li> </ol> <h2> 1. ╪º┘ä┘à┘é╪»┘à╪⌐ </h2> <p>┘å╪╣┘è╪┤ ┘ü┘è ╪╣╪╡╪▒ ╪ú╪»┘ê╪º╪¬ ╪º╪│╪¬╪½┘å╪º╪ª┘è╪⌐.</p> <p>┘å┘ü╪│ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬┘è ╪¬┘é┘ê╪» ╪º┘ä╪│┘è╪º╪▒╪º╪¬ ╪░╪º╪¬┘è╪⌐ ╪º┘ä┘é┘è╪º╪»╪⌐ ┘ê╪º┘ä╪¬╪┤╪«┘è╪╡ ╪º┘ä╪╖╪¿┘è╪î ╪¬┘à┘å╪¡ ╪º┘ä┘è┘ê┘à ╪╖╪º┘ä╪¿┘ï╪º ┘à╪╣┘ä┘æ┘à┘ï╪º ┘ä╪º ┘è╪¬╪╣╪¿ ╪ú╪¿╪»┘ï╪º╪î ┘ê┘ä╪º ┘è╪¡┘â┘à ╪╣┘ä┘è┘ç╪î ┘ê┘è╪¬╪░┘â╪▒ ┘â┘ä ╪Ñ╪¼╪º╪¿╪⌐ ╪ú╪╣╪╖╪º┘ç╪º ╪╣┘ä┘ë ╪º┘ä╪Ñ╪╖┘ä╪º┘é. ╪Ñ┘å┘ç ╪º╪¡╪¬┘à╪º┘ä ┘à╪░┘ç┘ä. ┘ê┘ç┘ê ┘à┘ê╪¼┘ê╪» ╪¿╪º┘ä┘ü╪╣┘ä.</p> <p>╪¬╪»┘ê╪▒ ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ╪¡┘ê┘ä ┘à╪º ┘è╪¡╪»╪½ ╪╣┘å╪»┘à╪º ┘è┘ä╪¬┘é┘è ┘ç╪░╪º ╪º┘ä┘ê╪º┘é╪╣ ╪¿╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è.</p> <p>┘ä╪╖╪º┘ä┘à╪º ┘â╪º┘å ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¿╪╖┘è╪ª┘ï╪º ┘ü┘è ╪º┘ä╪¬╪║┘è╪▒. ╪ú┘ü╪│╪¡╪¬ ╪º┘ä╪│╪¿┘ê╪▒╪º╪¬ ╪º┘ä╪│┘ê╪»╪º╪í ╪º┘ä┘à╪¼╪º┘ä ┘ä┘ä╪│╪¿┘ê╪▒╪º╪¬ ╪º┘ä╪¿┘è╪╢╪º╪í. ┘ê╪º┘ä╪│╪¿┘ê╪▒╪º╪¬ ╪º┘ä╪¿┘è╪╢╪º╪í ┘ä┘ä╪╣╪º╪▒╪╢╪º╪¬. ┘ê╪º┘ä╪╣╪º╪▒╪╢╪º╪¬ ┘ä┘ä╪ú╪¼┘ç╪▓╪⌐ ╪º┘ä┘ä┘ê╪¡┘è╪⌐. ┘ä┘â┘å ╪¬╪¡╪¬ ┘â┘ä ╪╖╪¿┘é╪⌐ ┘à┘å ╪º┘ä╪╣╪¬╪º╪» ╪º┘ä╪¼╪»┘è╪»╪î ╪╕┘ä ╪º┘ä┘ç┘è┘â┘ä ╪º┘ä╪ú╪│╪º╪│┘è ┘à╪ú┘ä┘ê┘ü┘ï╪º ╪¿╪╣┘å╪º╪»: ┘à╪╣┘ä┘æ┘à ┘ê╪º╪¡╪»╪î ┘ê╪╖┘ä╪º╪¿ ┘â╪½┘è╪▒┘ê┘å╪î ┘ê┘à┘å┘ç╪¼ ╪½╪º╪¿╪¬╪î ┘ê╪│╪º╪╣╪⌐ ╪¬┘å╪│╪º╪¿ ╪╣┘ä┘ë ┘å╪¡┘ê ┘à╪¬╪╖╪º╪¿┘é ┘ä┘ä╪¼┘à┘è╪╣. ┘è┘ç╪»╪» ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪¿┘â╪│╪▒ ┘ç╪░╪º ╪º┘ä┘ç┘è┘â┘ä. ╪Ñ┘å┘ç ┘è┘é╪»┘æ┘à ╪Ñ┘à┘â╪º┘å┘è╪⌐ ╪¬╪╣┘ä┘æ┘à ┘è┘å╪¡┘å┘è ╪¡┘ê┘ä ╪º┘ä┘à╪¬╪╣┘ä┘à╪î ╪¿╪»┘ä┘ï╪º ┘à┘å ╪Ñ╪¼╪¿╪º╪▒ ╪º┘ä┘à╪¬╪╣┘ä┘à ╪╣┘ä┘ë ╪º┘ä╪º┘å╪¡┘å╪º╪í ╪¡┘ê┘ä ╪º┘ä╪¼╪»┘ê┘ä ╪º┘ä╪▓┘à┘å┘è.</p> <p>╪│╪ñ╪º┘ä ╪º┘ä╪¿╪¡╪½ ╪º┘ä┘à╪▒┘â╪▓┘è ╪º┘ä╪░┘è ┘è┘ê╪¼┘æ┘ç ┘ç╪░╪º ╪º┘ä╪╣┘à┘ä ╪¿╪│┘è╪╖ ┘ü┘è ╪╡┘è╪º╪║╪¬┘ç ┘ê╪╡╪╣╪¿ ┘ü┘è ╪Ñ╪¼╪º╪¿╪¬┘ç: ┘ç┘ä ┘è┘à┘â┘å ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪ú┘å ┘è╪¼╪╣┘ä ╪º┘ä╪¬╪╣┘ä┘è┘à ╪ú┘â╪½╪▒ ┘ü╪╣╪º┘ä┘è╪⌐ ┘ê╪╣╪»╪º┘ä╪⌐╪î ┘ê╪Ñ╪░╪º ┘â╪º┘å ╪º┘ä╪ú┘à╪▒ ┘â╪░┘ä┘â╪î ┘ü╪¬╪¡╪¬ ╪ú┘è┘æ ╪┤╪▒┘ê╪╖╪ƒ</p> <p>┘ä┘ä╪Ñ╪¼╪º╪¿╪⌐╪î ┘è╪¼╪¿ ╪╣┘ä┘è┘å╪º ╪ú┘ê┘ä┘ï╪º ╪ú┘å ┘å┘ü┘ç┘à ┘à╪º ╪¬┘ü╪╣┘ä┘ç ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘ü╪╣┘ä┘ï╪º. ┘è╪¼╪¿ ╪ú┘å ┘å┘ü╪╡┘ä ╪º┘ä╪¬┘é╪»┘à ╪º┘ä╪¡┘é┘è┘é┘è ╪╣┘å ╪º┘ä╪╢╪¼┘è╪¼ ╪º┘ä╪¬╪│┘ê┘è┘é┘è. ┘è╪¼╪¿ ╪ú┘å ┘å┘ê╪º╪¼┘ç ╪¡┘é╪º╪ª┘é ┘à╪▓╪╣╪¼╪⌐ ╪╣┘å ╪º┘ä╪¿┘è╪º┘å╪º╪¬ ┘ê╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐╪î ┘ê╪«╪╖╪▒ ╪ú┘å ╪º┘ä╪ú╪»┘ê╪º╪¬ ╪¡╪│┘å╪⌐ ╪º┘ä┘å┘è╪⌐ ┘é╪» ╪¬╪╣┘à┘æ┘é ╪ú┘ê╪¼┘ç ╪╣╪»┘à ╪º┘ä┘à╪│╪º┘ê╪º╪⌐ ╪º┘ä╪¬┘è ╪¬╪»┘æ╪╣┘è ╪º┘ä┘é╪╢╪º╪í ╪╣┘ä┘è┘ç╪º.</p> <p>╪¬┘à╪╢┘è ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ┘ü┘è ╪½┘ä╪º╪½ ╪¡╪▒┘â╪º╪¬.</p> <p>╪ú┘ê┘ä┘ï╪º╪î ┘å╪¿┘å┘è ╪º┘ä╪│┘è╪º┘é. ┘å┘å╪╕╪▒ ┘à┘å ╪ú┘è┘å ╪¼╪º╪í╪¬ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐ ┘ê┘ä┘à╪º╪░╪º ┘é╪╡┘æ╪▒╪¬ ╪º┘ä╪½┘ê╪▒╪º╪¬ ╪º┘ä╪│╪º╪¿┘é╪⌐ ┘ü┘è ┘ê╪╣┘ê╪»┘ç╪º. ╪½╪º┘å┘è┘ï╪º╪î ┘å┘ü╪¡╪╡ ╪º┘ä╪¡╪º╪╢╪▒. ┘å╪│╪¬┘â╪┤┘ü ╪º┘ä╪╖╪▒┘é ╪º┘ä┘à┘ä┘à┘ê╪│╪⌐ ╪º┘ä╪¬┘è ┘è╪╣┘à┘ä ╪¿┘ç╪º ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪¿╪º┘ä┘ü╪╣┘ä ╪»╪º╪«┘ä ╪º┘ä┘ü╪╡┘ê┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪⌐╪î ┘ê┘å╪▓┘å ╪º┘ä╪ú╪»┘ä╪⌐ ╪╣┘ä┘ë ╪ó╪½╪º╪▒┘ç. ╪½╪º┘ä╪½┘ï╪º╪î ┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ╪º┘ä╪ú┘à╪º┘à. ┘å┘ê╪º╪¼┘ç ╪º┘ä╪ú╪«┘ä╪º┘é ┘ê╪º┘ä╪│┘è╪º╪│╪º╪¬ ┘ê╪º┘ä╪«┘è╪º╪▒╪º╪¬ ╪º┘ä╪¬┘è ╪│╪¬╪¡╪»┘æ╪» ┘à╪º ╪Ñ╪░╪º ┘â╪º┘å╪¬ ┘ç╪░┘ç ╪º┘ä╪½┘ê╪▒╪⌐ ╪¬╪«╪»┘à ┘â┘ä ╪╖╪º┘ä╪¿ ╪ú┘à ┘é┘ä╪⌐ ┘à┘à┘è╪▓╪⌐ ┘ü┘é╪╖.</p> <p>╪º┘ä╪▒┘ç╪º┘å╪º╪¬ ╪╣╪º┘ä┘è╪⌐. ╪º┘ä╪¬╪╣┘ä┘è┘à ┘ç┘ê ╪º┘ä┘à╪¡╪▒┘â ╪º┘ä╪╣╪╕┘è┘à ┘ä┘ä┘ü╪▒╪╡╪⌐╪î ╪º┘ä┘é┘ê╪⌐ ╪º┘ä╪¬┘è ╪¬╪▒┘ü╪╣ ╪º┘ä╪╣╪º╪ª┘ä╪º╪¬ ╪╣╪¿╪▒ ╪º┘ä╪ú╪¼┘è╪º┘ä. ╪Ñ╪░╪º ╪¼╪╣┘ä┘ç ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪ú┘é┘ê┘ë╪î ┘å┘â╪│╪¿ ╪┤┘è╪ª┘ï╪º ┘ä╪º ┘è┘Å┘é╪»┘æ╪▒ ╪¿╪½┘à┘å. ╪Ñ╪░╪º ╪¼╪╣┘ä┘ç ╪ú╪╢┘è┘é╪î ┘å╪«╪│╪▒ ╪┤┘è╪ª┘ï╪º ┘é╪» ┘ä╪º ┘å╪│╪¬╪╣┘è╪»┘ç ╪ú╪¿╪»┘ï╪º. ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ┘à╪¡╪º┘ê┘ä╪⌐ ┘ä┘ü┘ç┘à ╪ú┘è┘æ ┘à╪│╪¬┘é╪¿┘ä ┘å╪¿┘å┘è.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪º┘ä┘à╪│╪¬┘é╪¿┘ä ╪º┘ä╪ú┘ê┘ä ╪º┘ä┘à╪│╪¬┘é╪¿┘ä ╪º┘ä╪½╪º┘å┘è (┘ü╪▒╪╡╪⌐) (╪╣╪»┘à ┘à╪│╪º┘ê╪º╪⌐) | | | ┘â┘ä ╪╣┘é┘ä ┘é╪º╪»╪▒ ┘è╪¬┘ê╪│┘æ╪╣ | ╪ú┘ü╪╢┘ä ╪º┘ä╪ú╪»┘ê╪º╪¬ ┘ä┘ä┘é┘ä╪⌐ | ┘ä╪º ╪ú╪¡╪» ┘è╪│┘é╪╖ ╪¿╪╡┘à╪¬ | ┘ü╪▒╪▓ ╪║╪º┘à╪╢ ┘ä┘ä╪ú╪╖┘ü╪º┘ä | ╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘ü┘ê╪▒┘è╪⌐ | ┘è┘ü┘é╪» ╪º┘ä╪¬╪╣┘ä┘à ╪▒┘ê╪¡┘ç | | \____________ ____________/ \/ / \ / ╪ú┘å╪¬ \ / ╪¬┘é╪▒╪▒ \ /___________\ </code></pre> </div> <h2> 2. ┘å╪¿╪░╪⌐ ╪¬╪º╪▒┘è╪«┘è╪⌐ ┘à┘ê╪¼╪▓╪⌐ ╪╣┘å ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐ </h2> <p>┘ä┘ü┘ç┘à ╪ú┘è┘å ┘å╪¡┘å ╪░╪º┘ç╪¿┘ê┘å╪î ┘è╪¼╪¿ ╪ú┘ê┘ä┘ï╪º ╪ú┘å ┘å┘ü┘ç┘à ┘à┘å ╪ú┘è┘å ╪ú╪¬┘è┘å╪º.</p> <p>┘é╪╡╪⌐ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ┘ç┘è ┘é╪╡╪⌐ ╪»┘ê╪▒╪º╪¬. ┘à╪▒╪º╪▒┘ï╪º ┘ê╪¬┘â╪▒╪º╪▒┘ï╪º╪î ┘è╪╡┘ä ╪º╪«╪¬╪▒╪º╪╣ ╪¼╪»┘è╪» ┘à┘ä┘ü┘ê┘ü┘ï╪º ╪¿┘ê╪╣┘ê╪» ╪╣╪╕┘è┘à╪⌐ ╪¿╪º┘ä╪¬╪¡┘ê┘ä. ┘ê┘à╪▒╪º╪▒┘ï╪º ┘ê╪¬┘â╪▒╪º╪▒┘ï╪º╪î ┘è╪│╪¬┘é╪▒ ┘ü┘è ╪»┘ê╪▒ ╪½╪º┘å┘ê┘è ΓÇö ┘à┘ü┘è╪»╪î ┘ä┘â┘å┘ç ┘å╪º╪»╪▒┘ï╪º ┘à╪º ┘è┘â┘ê┘å ╪½┘ê╪▒┘è┘ï╪º.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ┘ê╪╣╪» ╪╣╪╕┘è┘à | v ╪ú┘à┘ä ┘ê┘å╪┤┘ê╪⌐ | v ┘è┘ü╪▒╪╢ ╪º┘ä┘ê╪º┘é╪╣ ┘å┘ü╪│┘ç | v ╪»┘ê╪▒ ╪½╪º┘å┘ê┘è &lt;---- ╪º┘ä╪ó┘ä╪⌐ ┘ä╪º ╪¬╪¡┘â┘à ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è </code></pre> </div> <p>┘ä┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ╪º┘ä╪▒╪º╪»┘è┘ê. ╪╣┘å╪»┘à╪º ┘ê╪╡┘ä╪¬ ╪º┘ä╪¿╪½ ╪º┘ä╪Ñ╪░╪º╪╣┘è ╪Ñ┘ä┘ë ╪º┘ä╪¿┘è┘ê╪¬ ╪º┘ä╪ú┘à╪▒┘è┘â┘è╪⌐ ┘ü┘è ╪╣╪┤╪▒┘è┘å┘è╪º╪¬ ╪º┘ä┘é╪▒┘å ╪º┘ä╪╣╪┤╪▒┘è┘å╪î ╪¬┘å╪¿╪ú ╪º┘ä┘à╪¬╪¡┘à╪│┘ê┘å ╪¿╪ú┘å ┘â┘ä ╪╖┘ü┘ä ┘ü┘è ╪º┘ä╪¿┘ä╪º╪» ╪│┘è╪╣┘ä┘æ┘à┘ç ┘é╪▒┘è╪¿┘ï╪º ╪╣╪»╪» ┘é┘ä┘è┘ä ┘à┘å ╪º┘ä┘à╪¡╪º╪╢╪▒┘è┘å ╪º┘ä┘ä╪º┘à╪╣┘è┘å╪î ╪¬╪╡┘ä ╪ú╪╡┘ê╪º╪¬┘ç┘à ╪Ñ┘ä┘ë ┘à╪╖╪º╪¿╪« ╪º┘ä┘à╪▓╪º╪▒╪╣ ┘ê╪┤┘é┘é ╪º┘ä┘à╪»┘å ╪╣┘ä┘ë ╪¡╪» ╪│┘ê╪º╪í. ┘ä┘à ┘è╪¡╪»╪½ ╪░┘ä┘â. ╪ú╪╡╪¿╪¡ ╪º┘ä╪▒╪º╪»┘è┘ê ┘à┘â┘à┘æ┘ä┘ï╪º╪î ┘ä╪º ╪¿╪»┘è┘ä┘ï╪º.</p> <p>┘ä┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ╪º┘ä╪¬┘ä┘ü╪º╪▓. ┘ü┘è ╪«┘à╪│┘è┘å┘è╪º╪¬ ┘ê╪│╪¬┘è┘å┘è╪º╪¬ ╪º┘ä┘é╪▒┘å ╪º┘ä┘à╪º╪╢┘è╪î ┘ê╪╣╪» ╪º┘ä╪¬┘ä┘ü╪º╪▓ ╪º┘ä╪¬╪╣┘ä┘è┘à┘è ╪¿╪Ñ╪»╪«╪º┘ä ╪ú┘ü╪╢┘ä ╪º┘ä┘à╪╣┘ä┘æ┘à┘è┘å ┘ü┘è ╪º┘ä╪╣╪º┘ä┘à ╪Ñ┘ä┘ë ┘â┘ä ╪║╪▒┘ü╪⌐ ┘à╪╣┘è╪┤╪⌐. ┘ç┘ê ╪º┘ä╪ó╪«╪▒ ╪¬┘ä╪º╪┤┘ë ╪Ñ┘ä┘ë ╪º┘ä╪«┘ä┘ü┘è╪⌐╪î ╪╣╪▒╪╢┘ï╪º ┘à╪¬╪«╪╡╪╡┘ï╪º ┘ä╪º ╪ú╪│╪º╪│┘Ä ┘ä┘å╪╕╪º┘à ╪¼╪»┘è╪».</p> <p>╪½┘à ╪¼╪º╪í ╪º┘ä╪¡╪º╪│┘ê╪¿╪î ┘ê┘à╪╣ ┘à┘ê╪¼╪⌐ ╪¼╪»┘è╪»╪⌐ ┘à┘å ╪º┘ä╪¬┘ü╪º╪ñ┘ä.</p> <p>┘é╪»┘æ┘à ╪º┘ä╪¬╪╣┘ä┘è┘à ╪º┘ä┘à╪¿╪▒┘à╪¼╪î ┘ê┘ç┘ê ┘à╪╡╪╖┘ä╪¡ ╪╡╪º╪║┘ç ╪╣╪º┘ä┘à ╪º┘ä┘å┘ü╪│ ╪¿. ┘ü. ╪│┘â┘è┘å╪▒ ┘ü┘è ╪«┘à╪│┘è┘å┘è╪º╪¬ ╪º┘ä┘é╪▒┘å ╪º┘ä┘à╪º╪╢┘è╪î ╪▒╪ñ┘è╪⌐ ┘à╪║╪▒┘è╪⌐: ┘à╪¡╪¬┘ê┘ë ┘à┘é╪│┘æ┘à ╪Ñ┘ä┘ë ╪«╪╖┘ê╪º╪¬ ╪╡╪║┘è╪▒╪⌐╪î ┘ê┘â┘ä ╪╖╪º┘ä╪¿ ┘è╪¬┘é╪»┘à ╪¿┘ê╪¬┘è╪▒╪¬┘ç ╪º┘ä╪«╪º╪╡╪⌐╪î ┘à╪╣ ╪¬╪║╪░┘è╪⌐ ╪▒╪º╪¼╪╣╪⌐ ┘ü┘ê╪▒┘è╪⌐ ╪╣┘å╪» ┘â┘ä ┘à┘å╪╣╪╖┘ü. ┘â╪º┘å╪¬ "╪ó┘ä╪º╪¬ ╪º┘ä╪¬╪╣┘ä┘è┘à" ╪º┘ä╪«╪º╪╡╪⌐ ╪¿╪│┘â┘è┘å╪▒ ┘à┘è┘â╪º┘å┘è┘â┘è╪⌐ ┘ê╪«╪▒┘é╪º╪í ┘ê┘à╪¡╪»┘ê╪»╪⌐. ┘ä┘â┘å ╪º┘ä┘ü┘â╪▒ ╪º┘ä╪░┘è ╪«┘ä┘ü┘ç╪º ΓÇö ╪ú┘å ╪º┘ä╪¬╪╣┘ä┘à ┘è┘à┘â┘å ╪¬╪«╪╡┘è╪╡┘ç ╪╣╪¿╪▒ ╪¬╪│┘ä╪│┘ä ╪»┘é┘è┘é ┘ê╪¬╪╣╪▓┘è╪▓ ┘à╪│╪¬┘à╪▒ ΓÇö ╪▓╪▒╪╣ ╪¿╪░╪▒╪⌐ ╪│╪¬┘å┘à┘ê ┘ä╪╣┘é┘ê╪».</p> <p>╪¼┘ä╪¿ ╪╕┘ç┘ê╪▒ ╪º┘ä╪¡╪º╪│┘ê╪¿ ╪º┘ä╪┤╪«╪╡┘è ┘ü┘è ╪½┘à╪º┘å┘è┘å┘è╪º╪¬ ╪º┘ä┘é╪▒┘å ╪º┘ä┘à╪º╪╢┘è ╪º┘ä╪¡┘ê╪º╪│┘è╪¿ ╪Ñ┘ä┘ë ╪º┘ä┘à╪»╪º╪▒╪│ ╪╣┘ä┘ë ┘å╪╖╪º┘é ┘ê╪º╪│╪╣. ╪╕┘ç╪▒ ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¿┘à╪│╪º╪╣╪»╪⌐ ╪º┘ä╪¡╪º╪│┘ê╪¿ ┘ü┘è ╪º┘ä┘à╪«╪¬╪¿╪▒╪º╪¬ ┘ê╪º┘ä┘ü╪╡┘ê┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪⌐ ┘ü┘è ╪¼┘à┘è╪╣ ╪ú┘å╪¡╪º╪í ╪º┘ä╪╣╪º┘ä┘à ╪º┘ä┘à╪¬┘é╪»┘à. ┘ê┘à╪╣ ╪░┘ä┘â╪î ┘â╪º┘å╪¬ ╪º┘ä┘å╪¬╪º╪ª╪¼╪î ╪¿╪¡╪│╪¿ ┘à╪╣╪╕┘à ╪º┘ä┘à┘é╪º┘è┘è╪│╪î ┘à╪¬┘ê╪º╪╢╪╣╪⌐. ╪¬╪▒╪º┘â┘à ╪º┘ä╪║╪¿╪º╪▒ ╪╣┘ä┘ë ╪ó┘ä╪º╪¬ ┘â╪½┘è╪▒╪⌐. ╪¿┘é┘è ╪º┘ä┘â╪½┘è╪▒ ┘à┘å ╪º┘ä╪¿╪▒╪º┘à╪¼ ╪¿┘ä╪º ╪º╪│╪¬╪«╪»╪º┘à.</p> <p>┘é╪»┘æ┘à ╪º┘ä╪¿╪º╪¡╪½┘ê┘å ╪╖╪▒┘é┘ï╪º ┘ä╪¬┘ü╪│┘è╪▒ ┘ç╪░╪º ╪º┘ä┘å┘à╪╖ ┘à┘å ╪º┘ä╪╢╪¼┘è╪¼ ┘ê╪«┘è╪¿╪⌐ ╪º┘ä╪ú┘à┘ä.</p> <p>┘ä╪º╪▒┘è ┘â┘ê╪¿╪º┘å╪î ┘à╪ñ╪▒╪« ╪º┘ä╪¬╪╣┘ä┘è┘à ┘ü┘è ╪¼╪º┘à╪╣╪⌐ ╪│╪¬╪º┘å┘ü┘ê╪▒╪»╪î ╪▒╪¿┘à╪º ┘è┘â┘ê┘å ╪º┘ä╪╡┘ê╪¬ ╪º┘ä╪ú╪¿╪▒╪▓. ┘è┘ê╪½┘æ┘é ╪╣┘à┘ä┘ç ╪¡┘é┘è┘é╪⌐ ┘à╪¬┘â╪▒╪▒╪⌐: ╪º┘ä┘à╪»╪º╪▒╪│ ┘à┘é╪º┘ê┘à╪⌐ ╪¿╪┤┘â┘ä ┘à┘ä╪¡┘ê╪╕ ┘ä┘ä╪¬╪║┘è┘è╪▒ ╪º┘ä╪¼╪░╪▒┘è╪î ╪¬┘à╪¬╪╡ ╪º┘ä╪¬┘é┘å┘è╪º╪¬ ╪º┘ä╪¼╪»┘è╪»╪⌐ ╪»┘ê┘å ╪ú┘å ╪¬╪¬╪¡┘ê┘ä ╪¿┘ç╪º. ┘è╪┤┘è╪▒ ╪¬╪¡┘ä┘è┘ä ┘â┘ê╪¿╪º┘å ╪Ñ┘ä┘ë ╪ú┘å ╪º┘ä╪º╪¿╪¬┘â╪º╪▒ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è ┘å╪º╪»╪▒┘ï╪º ┘à╪º ┘è╪┤┘â┘æ┘ä ╪º┘ä╪¬╪╣┘ä┘è┘à ╪º┘ä┘à╪»╪▒╪│┘è╪¢ ╪¿┘ä ╪Ñ┘å ╪º┘ä╪¬╪╣┘ä┘è┘à ╪º┘ä┘à╪»╪▒╪│┘è ┘ç┘ê ╪º┘ä╪░┘è ┘è╪┤┘â┘æ┘ä ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º╪î ╪¿╪╖┘è┘ç╪º ┘ä╪¬┘å╪º╪│╪¿ ╪º┘ä╪Ñ╪¼╪▒╪º╪í╪º╪¬ ┘ê╪º┘ä╪¼╪»╪º┘ê┘ä ┘ê┘ç┘è╪º┘â┘ä ╪º┘ä╪│┘ä╪╖╪⌐ ╪º┘ä┘é╪º╪ª┘à╪⌐.</p> <p>┘ü┘è ┘ç╪░╪º ╪º┘ä┘à╪┤┘ç╪» ┘à┘å ╪º┘ä╪¬┘ê┘é╪╣╪º╪¬ ╪º┘ä┘à╪╣╪¬╪»┘ä╪⌐╪î ┘ê╪╡┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è.</p> <p>┘è╪«╪¬┘ä┘ü ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪╣┘å ╪ú╪│┘ä╪º┘ü┘ç ┘ü┘è ╪¼╪º┘å╪¿ ╪¡╪º╪│┘à. ┘â╪º┘å╪¬ ╪º┘ä╪¬┘é┘å┘è╪º╪¬ ╪º┘ä╪│╪º╪¿┘é╪⌐ ╪ú┘å╪╕┘à╪⌐ ╪¬┘ê╪▓┘è╪╣ ┘ü┘è ╪¼┘ê┘ç╪▒┘ç╪º ΓÇö ╪¬┘å┘é┘ä ╪º┘ä┘à╪¡╪¬┘ê┘ë ┘à┘å ┘à╪╡╪»╪▒ ╪Ñ┘ä┘ë ┘à╪¬╪╣┘ä┘à. ╪ú┘à╪º ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è┘à┘â┘å┘ç ╪ú┘å ┘è╪│╪¬╪¼┘è╪¿. ┘è┘à┘â┘å┘ç ╪ú┘å ┘è╪¬┘â┘è┘ü. ┘è┘à┘â┘å┘ç ╪ú┘å ┘è╪¡╪º┘ê╪▒. ┘è┘à┘â┘å┘ç ╪ú┘å ┘è┘ê┘ä┘æ╪». ┘ç╪░┘ç ┘ä┘è╪│╪¬ ╪º╪«╪¬┘ä╪º┘ü╪º╪¬ ╪¬╪º┘ü┘ç╪⌐. ┘ä╪º ┘è┘à┘â┘å ┘ä┘â╪¬╪º╪¿ ┘à╪»╪▒╪│┘è ╪ú┘å ┘è╪▒┘ë ╪ú┘å┘â ╪¬╪╣╪º┘å┘è ┘ê┘è╪║┘è┘æ╪▒ ╪┤╪▒╪¡┘ç. ┘ê┘è┘à┘â┘å ┘ä┘å╪╕╪º┘à ╪░┘â┘è ╪░┘ä┘â. ╪Ñ┘å ┘ç╪░┘ç ╪º┘ä┘é╪»╪▒╪⌐ ╪º┘ä┘ê╪º╪¡╪»╪⌐ ┘ç┘è ╪º┘ä╪│╪¿╪¿ ╪º┘ä╪░┘è ┘è╪¼╪╣┘ä ┘â╪½┘è╪▒┘ï╪º ┘à┘å ╪º┘ä┘à╪▒╪º┘é╪¿┘è┘å ┘è╪╣╪¬┘é╪»┘ê┘å ╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪│┘è┘â╪│╪▒ ╪º┘ä╪»┘ê╪▒╪⌐ ╪º┘ä╪¬╪º╪▒┘è╪«┘è╪⌐ ╪º┘ä╪¬┘è ┘ä┘à ╪¬╪│╪¬╪╖╪╣ ╪º┘ä╪▒╪º╪»┘è┘ê ┘ê┘ä╪º ╪º┘ä╪¬┘ä┘ü╪º╪▓ ┘ê┘ä╪º ╪º┘ä╪¡╪º╪│┘ê╪¿ ╪º┘ä╪┤╪«╪╡┘è ┘â╪│╪▒┘ç╪º.</p> <p>┘à╪º ╪Ñ╪░╪º ┘â╪º┘å┘ê╪º ╪╣┘ä┘ë ╪¡┘é╪î ┘ç┘ê ╪º┘ä╪│╪ñ╪º┘ä ╪º┘ä╪░┘è ╪¬┘ê╪¼╪» ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ┘ä┘ü╪¡╪╡┘ç.</p> <h2> 3. ┘â┘è┘ü ┘è╪╣┘à┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è </h2> <p>┘à┘å ╪º┘ä┘à╪║╪▒┘è ╪ú┘å ┘å╪¬╪╡┘ê┘æ╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è ┘â╪┤┘è╪í ┘ê╪º╪¡╪» ┘à╪¬╪▒╪º╪╡. ┘ä┘è╪│ ┘â╪░┘ä┘â.</p> <p>┘à┘å ╪º┘ä┘à╪▒╪¼╪¡ ╪ú┘å ┘è╪¡╪¬┘ê┘è ┘ü╪╡┘ä ╪º┘ä┘è┘ê┘à ╪╣┘ä┘ë ╪ú┘å╪╕┘à╪⌐ ╪░┘â╪º╪í ╪º╪╡╪╖┘å╪º╪╣┘è ┘à╪«╪¬┘ä┘ü╪⌐ ┘â╪½┘è╪▒╪⌐╪î ┘â┘ä ┘à┘å┘ç╪º ┘è┘é┘ê┘à ╪¿╪╣┘à┘ä ┘à╪«╪¬┘ä┘ü. ╪¿╪╣╪╢┘ç╪º ┘à╪▒╪ª┘è. ┘ê╪¿╪╣╪╢┘ç╪º ┘ä┘è╪│ ┘â╪░┘ä┘â.</p> <p>┘ä┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ┘å╪╕╪º┘à ╪Ñ╪»╪º╪▒╪⌐ ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪░┘è ┘è╪¬╪¬╪¿╪╣ ╪º┘ä╪¡╪╢┘ê╪▒ ┘ê╪º┘ä╪»╪▒╪¼╪º╪¬. ╪¬╪¡╪¬ ╪│╪╖╪¡┘ç╪î ╪¬╪¬┘å╪¿╪ú ╪º┘ä╪«┘ê╪º╪▒╪▓┘à┘è╪º╪¬ ╪¿╪╡┘à╪¬ ╪¿╪ú┘è ╪º┘ä╪╖┘ä╪º╪¿ ┘à╪╣╪▒╪╢┘ê┘å ┘ä╪«╪╖╪▒ ╪º┘ä╪¬╪│╪▒╪¿╪î ┘ê╪¬╪▒┘ü╪╣ ╪ú╪│┘à╪º╪í┘ç┘à ┘ä┘ä╪¬╪»╪«┘ä ┘é╪¿┘ä ╪ú┘å ╪¬╪╡╪¿╪¡ ╪º┘ä┘à╪┤┘â┘ä╪⌐ ┘ê╪º╪╢╪¡╪⌐. ┘ç╪░╪º ╪░┘â╪º╪í ╪º╪╡╪╖┘å╪º╪╣┘è╪î ┘ê╪Ñ┘å ┘â╪º┘å ┘é┘ä┘è┘ä┘ê┘å ┘è╪│┘à┘ê┘å┘ç ╪¿╪░┘ä┘â.</p> <p>┘ä┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ┘à╪│╪º╪╣╪» ╪º┘ä┘â╪¬╪º╪¿╪⌐ ╪º┘ä╪░┘è ┘è╪╢╪╣ ╪«╪╖┘ï╪º ╪¬╪¡╪¬ ╪ú╪«╪╖╪º╪í ╪º┘ä┘é┘ê╪º╪╣╪» ┘ä╪»┘ë ╪╖╪º┘ä╪¿ ┘ê┘è┘é╪¬╪▒╪¡ ╪¬╪¡╪│┘è┘å╪º╪¬. ┘è╪│╪¬╪«╪»┘à ╪┤╪¿┘â╪º╪¬ ╪╣╪╡╪¿┘è╪⌐ ┘à╪»╪▒┘æ╪¿╪⌐ ╪╣┘ä┘ë ┘à┘ä╪º┘è┘è┘å ╪º┘ä╪ú┘à╪½┘ä╪⌐ ┘à┘å ╪º┘ä┘â╪¬╪º╪¿╪⌐ ╪º┘ä╪¿╪┤╪▒┘è╪⌐. ╪Ñ┘å┘ç ╪░┘â╪º╪í ╪º╪╡╪╖┘å╪º╪╣┘è╪î ┘à╪»┘à╪¼ ╪¿┘à┘ç╪º╪▒╪⌐ ┘ä╪»╪▒╪¼╪⌐ ╪ú┘å┘ç ┘è╪¿╪»┘ê ╪╣╪º╪»┘è┘ï╪º.</p> <p>┘ä┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ┘à┘å╪╡╪⌐ ╪º┘ä╪¬╪»╪▒┘è╪│ ╪º┘ä╪¬┘è ╪¬╪╣╪▒╪╢ ┘à╪│╪ú┘ä╪⌐╪î ┘ê╪¬┘å╪¬╪╕╪▒ ╪Ñ╪¼╪º╪¿╪⌐╪î ╪½┘à ΓÇö ╪¿┘å╪º╪í┘ï ╪╣┘ä┘ë ╪¬┘ä┘â ╪º┘ä╪Ñ╪¼╪º╪¿╪⌐ ╪º┘ä┘ê╪¡┘è╪»╪⌐ ΓÇö ╪¬┘é╪▒╪▒ ┘à╪º ╪¬╪╣╪▒╪╢┘ç ╪¿╪╣╪» ╪░┘ä┘â. ┘ç╪░╪º ┘ç┘ê ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪ú┘â╪½╪▒ ╪ú╪┤┘â╪º┘ä┘ç ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐ ┘à╪¿╪º╪┤╪▒╪⌐: ┘å╪╕╪º┘à ┘è╪¬╪«╪░ ┘é╪▒╪º╪▒╪º╪¬ ╪¬╪▒╪¿┘ê┘è╪⌐ ┘ü┘è ╪º┘ä┘ê┘é╪¬ ╪º┘ä╪¡┘é┘è┘é┘è.</p> <p>┘ç┘å╪º┘â ┘å┘é╪╖╪⌐ ╪¬┘é┘å┘è╪⌐ ┘à┘ç┘à╪⌐ ╪¬╪│╪¬╪¡┘é ╪ú┘å ╪¬┘Å╪░┘â╪▒ ╪¿┘ê╪╢┘ê╪¡.</p> <p>╪¬┘é┘ê┘à ╪º┘ä┘à┘ê╪¼╪⌐ ╪º┘ä╪¡╪º┘ä┘è╪⌐ ┘à┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪Ñ┘ä┘ë ╪¡╪» ┘â╪¿┘è╪▒ ╪╣┘ä┘ë ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪ó┘ä┘è╪î ┘ê┘ä╪º ╪│┘è┘à╪º ╪╣┘ä┘ë ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪╣┘à┘è┘é. ┘ä╪º ╪¬╪¬╪¿╪╣ ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘é┘ê╪º╪╣╪» ┘à┘â╪¬┘ê╪¿╪⌐ ┘è╪»┘ê┘è┘ï╪º. ╪¿┘ä ╪¬╪¬╪╣┘ä┘à ╪ú┘å┘à╪º╪╖┘ï╪º ┘à┘å ┘â┘à┘è╪º╪¬ ┘ç╪º╪ª┘ä╪⌐ ┘à┘å ╪º┘ä╪¿┘è╪º┘å╪º╪¬. ┘å┘à┘ê╪░╪¼ ╪º┘ä┘ä╪║╪⌐ ╪º┘ä╪░┘è ┘è┘ê┘ä┘æ╪» ╪¬╪║╪░┘è╪⌐ ╪▒╪º╪¼╪╣╪⌐ ┘ä┘à┘é╪º┘ä╪º╪¬ ╪º┘ä╪╖┘ä╪º╪¿ ┘ä┘à ┘è╪¬╪╣┘ä┘à ╪º┘ä┘é┘ê╪º╪╣╪» ┘â┘à╪¼┘à┘ê╪╣╪⌐ ┘à╪¿╪º╪»╪ª╪¢ ╪¿┘ä ╪º┘à╪¬╪╡ ╪º┘ä┘é┘ê╪º╪╣╪» ╪¿┘é╪▒╪º╪í╪⌐ ┘à┘ä┘è╪º╪▒╪º╪¬ ╪º┘ä╪¼┘à┘ä ┘ê╪º╪│╪¬┘è╪╣╪º╪¿ ╪º┘å╪¬╪╕╪º┘à╪º╪¬┘ç╪º ╪º┘ä╪Ñ╪¡╪╡╪º╪ª┘è╪⌐.</p> <p>┘ä┘ç╪░╪º ┘å╪¬┘è╪¼╪¬╪º┘å ╪╣┘à┘è┘é╪¬╪º┘å.</p> <p>╪ú┘ê┘ä┘ï╪º╪î ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪¿╪º╪▒╪╣╪⌐ ╪¿╪┤┘â┘ä ┘à┘ä╪¡┘ê╪╕. ┘è┘à┘â┘å┘ç╪º ╪Ñ┘å╪¬╪º╪¼ ┘å╪½╪▒ ╪│┘ä╪│╪î ┘ê╪¡┘ä ┘à╪│╪º╪ª┘ä ╪▒┘è╪º╪╢┘è╪⌐╪î ┘ê╪º┘ä╪Ñ╪¼╪º╪¿╪⌐ ╪╣┘å ╪ú╪│╪ª┘ä╪⌐ ┘ü┘è ┘å╪╖╪º┘é ┘à╪░┘ç┘ä ┘à┘å ╪º┘ä┘à┘ê╪╢┘ê╪╣╪º╪¬╪î ┘ê╪«┘ê╪╢ ╪¡┘ê╪º╪▒╪º╪¬ ╪╖┘ê┘è┘ä╪⌐ ╪¬╪¿╪»┘ê ╪║╪º┘ä╪¿┘ï╪º ╪╖╪¿┘è╪╣┘è╪⌐.</p> <p>╪½╪º┘å┘è┘ï╪º╪î ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪║╪▒┘è╪¿╪⌐ ╪¿╪╖╪▒┘è┘é╪⌐ ┘à┘é┘ä┘é╪⌐. ┘ê┘ä╪ú┘å┘ç╪º ╪¬╪¬╪╣┘ä┘à ╪¿╪»┘ä┘ï╪º ┘à┘å ╪ú┘å ╪¬┘Å╪¿╪▒┘à╪¼╪î ┘ü╪Ñ┘å ╪│┘ä┘ê┘â┘ç╪º ┘ä┘è╪│ ╪»╪º╪ª┘à┘ï╪º ┘é╪º╪¿┘ä╪º┘ï ┘ä┘ä╪¬┘å╪¿╪ñ ╪ú┘ê ┘ä┘ä╪¬╪ú┘ê┘è┘ä. ┘é╪» ┘è┘å╪¬╪¼ ╪º┘ä┘å╪╕╪º┘à ╪┤╪▒╪¡┘ï╪º ╪▒╪º╪ª╪╣┘ï╪º ┘ü┘è ┘ä╪¡╪╕╪⌐ ┘ê┘è╪▒╪¬┘â╪¿ ╪«╪╖╪ú┘ï ┘ê╪º╪½┘é┘ï╪º ┘ü┘è ╪º┘ä╪¬╪º┘ä┘è╪⌐. ┘ä╪º ┘è┘à┘ä┘â ┘ü┘ç┘à┘ï╪º ╪¡┘é┘è┘é┘è┘ï╪º ╪¿╪º┘ä┘à╪╣┘å┘ë ╪º┘ä╪¿╪┤╪▒┘è╪¢ ┘è┘à┘ä┘â ╪ú┘å┘à╪º╪╖┘ï╪º╪î ┘ê╪º┘ä╪ú┘å┘à╪º╪╖ ┘é┘ê┘è╪⌐ ┘ä┘â┘å┘ç╪º ┘ä┘è╪│╪¬ ┘ü┘ç┘à┘ï╪º.</p> <p>┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪î ╪¬╪¬╪▒╪¼┘à ┘ç╪░┘ç ╪º┘ä┘é┘ê┘ë ┘ê┘å┘é╪º╪╖ ╪º┘ä╪╢╪╣┘ü ╪Ñ┘ä┘ë ┘ê┘é╪º╪ª╪╣ ┘à┘ä┘à┘ê╪│╪⌐.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è ________________________________________________________ | | | ┘å╪╕╪º┘à ╪Ñ╪»╪º╪▒╪⌐ ╪º┘ä╪¬╪╣┘ä┘à ┘à╪│╪º╪╣╪» ╪º┘ä┘â╪¬╪º╪¿╪⌐ | | ┘è╪¬┘å╪¿╪ú ╪¿╪º┘ä╪¬╪│╪▒╪¿ ┘è╪╡╪¡╪¡ ╪º┘ä┘é┘ê╪º╪╣╪» | | | | | | ┘à┘å╪╡╪⌐ ╪º┘ä╪¬╪»╪▒┘è╪│ ┘à╪¡╪▒┘â ╪º┘ä╪¬┘é┘è┘è┘à | | ┘è┘â┘è┘æ┘ü ┘â┘ä ┘à╪│╪ú┘ä╪⌐ ┘è┘é┘è┘æ┘à ╪º┘ä┘à┘é╪º┘ä╪º╪¬ | | | | ┘è╪│╪¬╪«╪»┘à ╪º┘ä┘â┘ä: ╪¿┘è╪º┘å╪º╪¬ --&gt; ╪ú┘å┘à╪º╪╖ --&gt; ╪¬┘å╪¿╪ñ | |________________________________________________________| </code></pre> </div> <p>╪¬┘à┘â┘æ┘å ╪º┘ä┘é┘ê┘ë ╪º┘ä┘à┘à╪º╪▒╪│╪⌐ ╪º┘ä┘à╪«╪╡╪╡╪⌐╪î ┘ê╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪º┘ä┘ü┘ê╪▒┘è╪⌐╪î ┘ê╪»╪▒╪¼╪⌐ ┘à┘å ╪º┘ä╪º┘ç╪¬┘à╪º┘à ╪º┘ä┘ü╪▒╪»┘è ┘ä╪º ┘è┘à┘â┘å ┘ä╪ú┘è ┘à╪╣┘ä┘æ┘à ╪¿╪┤╪▒┘è ╪ú┘å ┘è┘é╪»┘à┘ç╪º ┘ä╪½┘ä╪º╪½┘è┘å ╪╖╪º┘ä╪¿┘ï╪º ┘ü┘è ┘ê┘é╪¬ ┘ê╪º╪¡╪». ╪ú┘à╪º ┘å┘é╪º╪╖ ╪º┘ä╪╢╪╣┘ü ┘ü╪¬╪╡┘å╪╣ ╪º┘ä┘à╪«╪º╪╖╪▒: ╪º╪¡╪¬┘à╪º┘ä ╪¬╪╣╪▓┘è╪▓ ╪º┘ä╪ú╪«╪╖╪º╪í╪î ┘ê╪╣╪»┘à ╪┤┘ü╪º┘ü┘è╪⌐ ╪º╪¬╪«╪º╪░ ╪º┘ä┘é╪▒╪º╪▒╪î ┘ê╪º┘ä╪«╪╖╪▒ ╪º┘ä╪»╪º╪ª┘à ╪º┘ä┘à╪¬┘à╪½┘ä ┘ü┘è ╪ú┘å ┘è╪«╪╖╪ª ╪º┘ä┘å╪╕╪º┘à ╪¿┘â┘ä ╪º┘ä╪½┘é╪⌐ ╪║┘è╪▒ ╪º┘ä┘à╪¿╪▒╪▒╪⌐ ┘ä╪ó┘ä╪⌐ ╪▒╪ú╪¬ ╪º┘ä╪Ñ╪¼╪º╪¿╪⌐ ╪º┘ä╪╡╪¡┘è╪¡╪⌐ ┘à╪▒╪⌐ ┘ê╪º╪¡╪»╪⌐ ┘ê╪¿╪º╪¬╪¬ ╪¬╪╣╪¬┘é╪» ╪ú┘å┘ç╪º ╪╡╪¡┘è╪¡╪⌐ ╪»╪º╪ª┘à┘ï╪º.</p> <p>┘è╪¼╪¿ ╪ú┘å ┘è┘à╪│┘â ╪º┘ä╪│╪▒╪» ╪º┘ä╪╡╪º╪»┘é ╪╣┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¿╪º┘ä╪¡┘é┘è┘é╪¬┘è┘å ┘à╪╣┘ï╪º. ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘ê╪º╪╣╪»╪⌐ ╪¡┘é┘ï╪º. ┘ê┘ç┘è ╪║┘è╪▒ ┘â╪º┘à┘ä╪⌐ ╪¡┘é┘ï╪º.</p> <h2> 4. ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪┤╪«╪╡┘è ┘ê╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ </h2> <p>┘é┘ä╪⌐ ┘à┘å ╪º┘ä╪╣╪¿╪º╪▒╪º╪¬ ┘ü┘è ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐ ╪¬┘Å╪│╪¬╪»╪╣┘ë ╪¿┘é╪»╪▒ ╪ú┘â╪¿╪▒ ┘à┘å ╪º┘ä╪¬╪¿╪¼┘è┘ä ┘ê╪¿┘é╪»╪▒ ╪ú┘é┘ä ┘à┘å ╪º┘ä╪»┘é╪⌐ ┘à┘å ╪╣╪¿╪º╪▒╪⌐ "╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪┤╪«╪╡┘è".</p> <p>╪║┘è╪▒ ╪ú┘å ╪º┘ä┘ü┘â╪▒╪⌐ ╪º┘ä┘à╪▒┘â╪▓┘è╪⌐ ╪¿╪│┘è╪╖╪⌐ ┘ê╪¼╪░╪º╪¿╪⌐. ┘â┘ä ╪╖╪º┘ä╪¿ ┘è╪¬╪╣┘ä┘à ╪¿╪┤┘â┘ä ┘à╪«╪¬┘ä┘ü. ┘â┘ä ╪╖╪º┘ä╪¿ ┘è╪ú╪¬┘è ╪¿┘à╪╣╪▒┘ü╪⌐ ╪│╪º╪¿┘é╪⌐ ┘à╪«╪¬┘ä┘ü╪⌐╪î ┘ê┘å┘é╪º╪╖ ┘é┘ê╪⌐ ┘à╪«╪¬┘ä┘ü╪⌐╪î ┘ê┘å┘é╪º╪╖ ╪╢╪╣┘ü ┘à╪«╪¬┘ä┘ü╪⌐╪î ┘ê┘à╪╣╪»┘ä╪º╪¬ ╪¬┘é╪»┘à ┘à╪«╪¬┘ä┘ü╪⌐. ╪º┘ä┘å╪╕╪º┘à ╪º┘ä╪░┘è ┘è┘à┘â┘å┘ç ╪Ñ╪»╪▒╪º┘â ┘ç╪░┘ç ╪º┘ä╪º╪«╪¬┘ä╪º┘ü╪º╪¬ ┘ê╪º┘ä╪º╪│╪¬╪¼╪º╪¿╪⌐ ┘ä┘ç╪º ╪¿╪┤┘â┘ä ┘ü╪▒╪»┘è ┘ä╪»┘è┘ç ╪º┘ä┘é╪»╪▒╪⌐ ╪╣┘ä┘ë ╪«╪»┘à╪⌐ ┘â┘ä ╪╖╪º┘ä╪¿ ╪ú┘ü╪╢┘ä ╪¿┘â╪½┘è╪▒ ┘à┘å ┘à┘å┘ç╪¼ ┘à┘ê╪¡┘æ╪».</p> <p>╪¬╪¬╪¼╪│╪» ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ┘ç╪░┘ç ╪º┘ä┘ü┘â╪▒╪⌐ ┘ü┘è ╪¿╪▒┘à╪¼┘è╪º╪¬.</p> <p>╪¬╪╣┘à┘ä ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪╣╪º╪»╪⌐ ┘ü┘è ╪»┘ê╪▒╪⌐. ╪¬╪╣╪▒╪╢ ┘à╪│╪ú┘ä╪⌐╪¢ ╪¬╪▒╪º┘é╪¿ ╪º╪│╪¬╪¼╪º╪¿╪⌐ ╪º┘ä╪╖╪º┘ä╪¿╪¢ ╪¬╪¡╪»┘æ╪½ ┘å┘à┘ê╪░╪¼┘ï╪º ╪»╪º╪«┘ä┘è┘ï╪º ┘ä┘à╪╣╪▒┘ü╪⌐ ╪░┘ä┘â ╪º┘ä╪╖╪º┘ä╪¿╪¢ ┘ê╪¬╪«╪¬╪º╪▒ ╪º┘ä┘à╪│╪ú┘ä╪⌐ ╪º┘ä╪¬╪º┘ä┘è╪⌐ ╪¿┘å╪º╪í┘ï ╪╣┘ä┘ë ╪░┘ä┘â ╪º┘ä┘å┘à┘ê╪░╪¼. ╪Ñ╪░╪º ╪╣╪º┘å┘ë ╪╖╪º┘ä╪¿ ┘à╪╣ ╪º┘ä┘â╪│┘ê╪▒╪î ┘è┘à┘å╪¡┘ç ╪º┘ä┘å╪╕╪º┘à ┘à╪▓┘è╪»┘ï╪º ┘à┘å ╪º┘ä╪¬╪»╪▒┘è╪¿ ╪╣┘ä┘ë ╪º┘ä┘â╪│┘ê╪▒ ┘é╪¿┘ä ╪º┘ä┘à╪╢┘è ┘é╪»┘à┘ï╪º. ╪Ñ╪░╪º ╪ú╪¬┘é┘å┘ç╪º ╪º┘ä╪╖╪º┘ä╪¿╪î ┘è╪¬┘é╪»┘à ╪º┘ä┘å╪╕╪º┘à. ╪¡┘ä┘é╪⌐ ╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘ç┘è ┘é┘ä╪¿ ╪º┘ä╪¬╪╡┘à┘è┘à.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪¬╪╣╪▒╪╢ ╪¬╪▒╪º┘é╪¿ ╪¬╪¡╪»┘æ╪½ ┘à╪│╪ú┘ä╪⌐ ╪º┘ä╪º╪│╪¬╪¼╪º╪¿╪⌐ ┘å┘à┘ê╪░╪¼ ╪º┘ä╪╖╪º┘ä╪¿ \ / | \ / | \________/______ | | | ╪¬╪«╪¬╪º╪▒ ╪º┘ä┘à╪│╪ú┘ä╪⌐ ╪º┘ä╪¬╪º┘ä┘è╪⌐ &lt;-------------+ </code></pre> </div> <p>┘è┘à┘â┘å ╪ú┘å ╪¬┘â┘ê┘å ╪º┘ä┘å╪¬╪º╪ª╪¼ ┘à╪░┘ç┘ä╪⌐ ┘ü┘è ╪│┘è╪º┘é╪º╪¬ ┘à╪¡╪»╪»╪⌐.</p> <p>┘ü┘è ╪º┘ä╪▒┘è╪º╪╢┘è╪º╪¬╪î ╪ú╪╕┘ç╪▒╪¬ ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ╪¿╪╣╪╢┘ï╪º ┘à┘å ╪ú┘é┘ê┘ë ╪º┘ä╪ú╪»┘ä╪⌐. ╪ú╪¿┘ä╪║╪¬ ╪»╪▒╪º╪│╪º╪¬ ┘ä┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä┘à╪│╪¬╪«╪»┘à╪⌐ ┘ü┘è ╪º┘ä╪¬╪»╪▒┘è╪¿ ╪º┘ä╪¡╪│╪º╪¿┘è ┘ê╪º┘ä╪¬╪╣┘ä┘è┘à ╪º┘ä╪¼╪¿╪▒┘è ╪╣┘å ┘à┘â╪º╪│╪¿ ┘à╪╣┘å┘ê┘è╪⌐ ┘ü┘è ╪¬╪¡╪╡┘è┘ä ╪º┘ä╪╖┘ä╪º╪¿╪î ┘ä╪º ╪│┘è┘à╪º ┘ä┘ä┘à╪¬╪╣┘ä┘à┘è┘å ╪º┘ä╪░┘è┘å ┘è╪¿╪»╪ú┘ê┘å ┘à╪¬╪ú╪«╪▒┘è┘å ╪╣┘å ╪ú┘é╪▒╪º┘å┘ç┘à. ╪º┘ä╪ó┘ä┘è╪⌐ ┘à┘å╪╖┘é┘è╪⌐ ╪¿╪»┘è┘ç┘è┘ï╪º: ╪º┘ä╪╖╪º┘ä╪¿ ╪º┘ä╪░┘è ┘ä┘ê┘ä╪º ╪░┘ä┘â ╪│┘è╪¬╪«┘ä┘ü ╪╣┘å ╪º┘ä╪▒┘â╪¿ ┘ê┘è╪│╪¬╪│┘ä┘à ΓÇö ╪ú┘ê ╪│┘è╪¼┘ä╪│ ┘à┘ä┘ä┘ï╪º ╪¿┘è┘å┘à╪º ┘è┘ä╪¡┘é ╪¿┘ç ╪▓┘à┘ä╪º╪ñ┘ç ΓÇö ┘è╪¡╪╡┘ä ╪¿╪»┘ä┘ï╪º ┘à┘å ╪░┘ä┘â ╪╣┘ä┘ë ╪╣┘à┘ä ┘à╪╣╪º┘è╪▒ ╪¬┘à╪º┘à┘ï╪º ╪╣┘å╪» ┘à╪│╪¬┘ê╪º┘ç. ╪¬╪¼╪▒╪¿╪⌐ ╪º┘ä╪º┘å╪│┘è╪º╪¿╪î ╪¬╪¼╪▒╪¿╪⌐ ╪º┘ä╪╣┘à┘ä ╪º┘ä╪░┘è ┘ä┘è╪│ ╪│┘ç┘ä╪º┘ï ╪¼╪»┘ï╪º ┘ê┘ä╪º ╪╡╪╣╪¿┘ï╪º ╪¼╪»┘ï╪º╪î ┘à╪¡┘ü┘ê╪╕╪⌐.</p> <p>╪║┘è╪▒ ╪ú┘å ╪º┘ä╪ú╪»┘ä╪⌐ ┘ä┘è╪│╪¬ ╪Ñ┘è╪¼╪º╪¿┘è╪⌐ ╪¿╪┤┘â┘ä ┘à┘ê╪¡╪»╪î ┘ê╪º┘ä╪ú╪│╪¿╪º╪¿ ┘à┘ü┘è╪»╪⌐.</p> <p>╪ú┘ê┘ä┘ï╪º╪î ┘ä╪º ╪¬╪│╪º┘ê┘è ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ╪Ñ┘ä╪º ╪¿┘é╪»╪▒ ┘å┘à┘ê╪░╪¼┘ç╪º ╪º┘ä╪ú╪│╪º╪│┘è ┘ä┘ä┘à╪╣╪▒┘ü╪⌐. ┘é╪» ┘è╪¡┘ä ╪╖╪º┘ä╪¿ ┘à╪│╪ú┘ä╪⌐ ╪¿╪┤┘â┘ä ╪╡╪¡┘è╪¡ ┘ä╪ú╪│╪¿╪º╪¿ ╪«╪º╪╖╪ª╪⌐╪î ╪ú┘ê ╪¿╪┤┘â┘ä ╪«╪º╪╖╪ª ┘ä╪ú╪│╪¿╪º╪¿ ╪╡╪¡┘è╪¡╪⌐. ┘è╪▒┘ë ╪º┘ä┘å╪╕╪º┘à ╪º┘ä┘å╪¬╪º╪ª╪¼ ┘ü┘é╪╖╪î ┘ä╪º ╪º┘ä╪º╪│╪¬╪»┘ä╪º┘ä ╪«┘ä┘ü┘ç╪º. ╪╣┘å╪»┘à╪º ┘è┘â┘ê┘å ╪º┘ä┘ç╪»┘ü ┘à┘ç╪º╪▒╪⌐ ╪Ñ╪¼╪▒╪º╪ª┘è╪⌐╪î ┘é╪» ┘ä╪º ┘è┘ç┘à ┘ç╪░╪º ┘â╪½┘è╪▒┘ï╪º. ╪╣┘å╪»┘à╪º ┘è┘â┘ê┘å ╪º┘ä┘ç╪»┘ü ┘ü┘ç┘à┘ï╪º ┘à┘ü╪º┘ç┘è┘à┘è┘ï╪º╪î ┘ü╪Ñ┘å┘ç ┘è┘ç┘à ┘â╪½┘è╪▒┘ï╪º.</p> <p>╪½╪º┘å┘è┘ï╪º╪î ╪¬╪«╪╡┘è╪╡ ╪º┘ä┘ê╪¬┘è╪▒╪⌐ ┘ä┘è╪│ ┘ç┘ê ┘å┘ü╪│┘ç ╪¬╪«╪╡┘è╪╡ ╪º┘ä┘à╪╣┘å┘ë. ┘è┘à┘â┘å ┘ä┘ä┘å╪╕╪º┘à ╪╢╪¿╪╖ ╪╡╪╣┘ê╪¿╪⌐ ╪º┘ä┘à╪│╪º╪ª┘ä╪î ┘ä┘â┘å┘ç ┘ä╪º ┘è╪│╪¬╪╖┘è╪╣ ╪¿╪│┘ç┘ê┘ä╪⌐ ╪Ñ╪»╪▒╪º┘â ╪│╪¿╪¿ ┘ü┘é╪»╪º┘å ╪º┘ä╪╖╪º┘ä╪¿ ┘ä┘ä╪¬╪¡┘ü┘è╪▓╪î ╪ú┘ê ╪º┘ä╪¬╪╣╪▒┘ü ╪╣┘ä┘ë ┘ä╪¡╪╕╪⌐ ╪╡╪▒╪º╪╣ ╪┤╪«╪╡┘è╪î ╪ú┘ê ╪¬┘é╪»┘è┘à ╪º┘ä╪¬╪┤╪¼┘è╪╣ ╪º┘ä╪░┘è ┘ä╪º ┘è┘é╪»┘à┘ç ╪Ñ┘ä╪º ╪Ñ┘å╪│╪º┘å ┘à┘ç╪¬┘à. ╪ú╪¿╪╣╪º╪» ╪º┘ä╪¬╪╣┘ä┘à ┘ç╪░┘ç ┘à┘ç┘à╪⌐╪î ┘ê╪º┘ä╪ó┘ä╪º╪¬ ┘à╪º ╪¬╪▓╪º┘ä ╪╣┘à┘è╪º╪í ╪╣┘å┘ç╪º ╪Ñ┘ä┘ë ╪¡╪» ┘â╪¿┘è╪▒.</p> <p>╪½╪º┘ä╪½┘ï╪º╪î ┘ç┘å╪º┘â ╪«╪╖╪▒ ╪ú┘å ┘è╪╢┘è┘æ┘é ╪º┘ä╪¬╪«╪╡┘è╪╡ ╪º┘ä╪╖┘à┘ê╪¡ ╪¿╪»┘ä┘ï╪º ┘à┘å ╪ú┘å ┘è┘ê╪│┘æ╪╣┘ç. ╪º┘ä┘å╪╕╪º┘à ╪º┘ä╪░┘è ┘è╪¡╪│┘æ┘å ╪Ñ╪¬┘é╪º┘å ┘à┘å┘ç╪¼ ╪½╪º╪¿╪¬ ┘é╪» ┘è╪¿┘é┘è ╪╖╪º┘ä╪¿┘ï╪º ┘à╪¬╪╣╪½╪▒┘ï╪º ┘è╪¬╪»╪▒╪¿ ╪Ñ┘ä┘ë ╪º┘ä╪ú╪¿╪» ╪╣┘ä┘ë ╪º┘ä┘à┘ç╪º╪▒╪º╪¬ ╪º┘ä╪Ñ╪¼╪▒╪º╪ª┘è╪⌐ ┘å┘ü╪│┘ç╪º╪î ╪»┘ê┘å ╪ú┘å ┘è╪╣╪▒╪╢┘ç ╪ú╪¿╪»┘ï╪º ┘ä┘ä╪¼┘ê╪º┘å╪¿ ╪º┘ä╪ú┘â╪½╪▒ ╪½╪▒╪º╪í┘ï ┘ê╪Ñ╪¿╪»╪º╪╣┘ï╪º ┘à┘å ╪º┘ä┘à╪º╪»╪⌐ ╪º┘ä╪¬┘è ┘é╪» ╪¬┘Å╪┤╪╣┘ä ╪º┘ç╪¬┘à╪º┘à┘ç ┘ü╪╣┘ä┘ï╪º. ╪º┘ä╪¬╪«╪╡┘è╪╡╪î ╪Ñ╪░╪º ╪ú┘Å╪│╪º╪í ╪¬┘å┘ü┘è╪░┘ç╪î ┘é╪» ┘è╪╡╪¿╪¡ ┘é┘ü╪╡┘ï╪º.</p> <p>┘ä╪░┘ä┘â ╪¬┘ê╪╡┘ä ╪º┘ä╪ú╪»╪¿┘è╪º╪¬ ╪¡┘ê┘ä ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä╪¬┘â┘è┘ü┘è ╪Ñ┘ä┘ë ╪º╪│╪¬┘å╪¬╪º╪¼ ┘à╪¬┘ê╪º╪▓┘å. ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪¡┘é┘è┘é┘è╪⌐╪î ┘ê┘ü┘è ╪º┘ä╪╕╪▒┘ê┘ü ╪º┘ä┘à┘å╪º╪│╪¿╪⌐ ╪¬┘å╪¬╪¼ ┘ü┘ê╪º╪ª╪» ┘é╪º╪¿┘ä╪⌐ ┘ä┘ä┘é┘è╪º╪│. ┘ä┘â┘å┘ç╪º ╪ú╪»╪º╪⌐ ╪¿┘è┘å ╪ú╪»┘ê╪º╪¬╪î ┘ä╪º ╪¿╪»┘è┘ä ╪╣┘å ╪ú╪╡┘ê┘ä ╪¬╪»╪▒┘è╪│ ┘à╪»╪▒┘ê╪│╪⌐. ┘é┘è┘à╪¬┘ç╪º ╪¬╪╣╪¬┘à╪»╪î ┘â┘à╪º ┘ç┘ê ╪º┘ä╪¡╪º┘ä ╪»╪º╪ª┘à┘ï╪º╪î ╪╣┘ä┘ë ╪º┘ä┘à┘ç╪º╪▒╪⌐ ╪º┘ä╪¬┘è ╪¬┘Å╪│╪¬╪«╪»┘à ╪¿┘ç╪º.</p> <h2> 5. ╪º┘ä╪¬┘é┘è┘è┘à ┘ê╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪º┘ä╪ó┘ä┘è╪º┘å </h2> <p>┘è╪╣╪» ╪º┘ä╪¬╪╡╪¡┘è╪¡ ┘à┘å ╪ú┘â╪½╪▒ ╪º┘ä┘à┘ç╪º┘à ╪º╪│╪¬┘ç┘ä╪º┘â┘ï╪º ┘ä┘ä┘ê┘é╪¬ ┘ê╪ú┘é┘ä┘ç╪º ┘à╪¡╪¿╪⌐ ┘ü┘è ╪º┘ä╪¬╪»╪▒┘è╪│.</p> <p>┘é╪» ┘è┘é╪╢┘è ┘à╪╣┘ä┘æ┘à ┘ä╪»┘è┘ç ┘à╪ª╪⌐ ╪╖╪º┘ä╪¿ ╪ú┘à╪│┘è╪⌐ ┘ü┘è ┘à╪▒╪º╪¼╪╣╪⌐ ┘à╪ª╪⌐ ┘à╪¼┘à┘ê╪╣╪⌐ ┘à┘å ╪º┘ä╪ú╪╣┘à╪º┘ä ╪º┘ä┘à┘â╪¬┘ê╪¿╪⌐. ╪¬╪¬┘å╪º┘ü╪│ ┘à╪╖╪º┘ä╪¿ ╪º┘ä╪¬╪╡╪¡┘è╪¡ ┘à╪¿╪º╪┤╪▒╪⌐ ┘à╪╣ ┘à╪╖╪º┘ä╪¿ ╪º┘ä╪¬╪«╪╖┘è╪╖ ┘ê╪º┘ä╪¬╪¡╪╢┘è╪▒ ┘ê╪º┘ä╪¬┘ü╪º╪╣┘ä╪º╪¬ ╪º┘ä╪¿╪┤╪▒┘è╪⌐ ╪º┘ä╪¬┘è ╪¬┘à┘å╪¡ ╪º┘ä╪¬╪»╪▒┘è╪│ ┘à╪╣┘å╪º┘ç. ┘ä╪º ╪╣╪¼╪¿ ╪ú┘å ╪º┘ä╪ú╪¬┘à╪¬╪⌐ ┘ê╪╡┘ä╪¬ ╪Ñ┘ä┘ë ┘ç┘å╪º ╪¿┘é┘ê╪⌐.</p> <p>┘è╪ú╪¬┘è ╪º┘ä╪¬┘é┘è┘è┘à ╪º┘ä╪ó┘ä┘è ┘ü┘è ╪ú╪┤┘â╪º┘ä ╪╣╪»╪⌐.</p> <p>╪ú╪¿╪│╪╖┘ç╪º ┘è╪¬╪╢┘à┘å ╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪º╪«╪¬┘è╪º╪▒ ┘à┘å ┘à╪¬╪╣╪»╪» ┘ê╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪Ñ╪¼╪º╪¿╪⌐ ╪º┘ä┘é╪╡┘è╪▒╪⌐╪î ╪¡┘è╪½ ╪╡╪¡┘æ╪¡╪¬ ╪º┘ä╪¡┘ê╪º╪│┘è╪¿ ╪º┘ä╪Ñ╪¼╪º╪¿╪º╪¬ ╪¿╪»┘é╪⌐ ┘ä╪╣┘é┘ê╪». ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘à┘ê╪½┘ê┘é╪⌐ ┘ê╪│╪▒┘è╪╣╪⌐ ┘ê╪║┘è╪▒ ┘é╪º╪¿┘ä╪⌐ ┘ä┘ä╪º╪╣╪¬╪▒╪º╪╢ ┘ü┘è ╪│┘è╪º┘é╪º╪¬ ┘â╪½┘è╪▒╪⌐╪î ┘ê╪Ñ┘å ┘â╪º┘å╪¬ ╪¬╪«╪¬╪¿╪▒ ╪┤╪▒┘è╪¡╪⌐ ╪╢┘è┘é╪⌐ ┘à┘å ╪º┘ä╪¬╪╣┘ä┘à ┘ê╪¬┘â╪º┘ü╪ª ╪º┘ä╪¬╪╣╪▒┘æ┘ü ╪╣┘ä┘ë ╪º┘ä╪º╪│╪¬╪»┘ä╪º┘ä.</p> <p>╪ú┘à╪º ╪º┘ä╪ú┘â╪½╪▒ ╪╖┘à┘ê╪¡┘ï╪º ┘ü┘ç┘è ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä┘à╪╡┘à┘à╪⌐ ┘ä╪¬┘é┘è┘è┘à ╪º┘ä┘â╪¬╪º╪¿╪⌐ ╪º┘ä┘à╪╖┘ê┘æ┘ä╪⌐.</p> <p>╪¬╪│╪¬╪«╪»┘à ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘à╪╣╪º┘ä╪¼╪⌐ ╪º┘ä┘ä╪║╪⌐ ╪º┘ä╪╖╪¿┘è╪╣┘è╪⌐ ┘ä╪¬┘é┘è┘è┘à ╪º┘ä┘à┘é╪º┘ä╪º╪¬ ╪╣┘ä┘ë ╪ú╪¿╪╣╪º╪» ┘à╪½┘ä ╪º┘ä┘ê╪╢┘ê╪¡ ┘ê╪º┘ä╪¬┘å╪╕┘è┘à ┘ê╪º┘ä┘é┘ê╪º╪╣╪» ┘ê╪º┘ä╪º┘ä╪¬╪▓╪º┘à ╪¿┘à┘ê╪╢┘ê╪╣. ┘ü┘è ╪º┘ä╪º╪«╪¬╪¿╪º╪▒╪º╪¬ ╪º┘ä┘à┘ê╪¡╪»╪⌐╪î ╪½╪¿╪¬ ╪ú┘å ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪¬┘å╪¬╪¼ ╪»╪▒╪¼╪º╪¬ ╪¬╪▒╪¬╪¿╪╖ ╪¿╪┤┘â┘ä ┘à╪╣┘é┘ê┘ä ┘à╪╣ ╪º┘ä┘à╪╡╪¡╪¡┘è┘å ╪º┘ä╪¿╪┤╪▒┘è┘è┘å. ╪¿╪º┘ä┘å╪│╪¿╪⌐ ┘ä┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪º┘ä╪¬┘â┘ê┘è┘å┘è╪⌐ ┘à┘å╪«┘ü╪╢╪⌐ ╪º┘ä┘à╪«╪º╪╖╪▒ ΓÇö ┘à╪│╪º╪╣╪»╪⌐ ╪╖╪º┘ä╪¿ ╪╣┘ä┘ë ╪▒╪ñ┘è╪⌐ ╪ú┘å ┘ü┘é╪▒╪⌐ ╪«╪º╪▒╪¼ ╪º┘ä┘à┘ê╪╢┘ê╪╣╪î ┘ê╪ú┘å ╪ú╪╖╪▒┘ê╪¡╪⌐ ┘à╪»┘ü┘ê┘å╪⌐╪î ┘ê╪ú┘å ╪º┘å╪¬┘é╪º┘ä╪º╪¬ ┘à┘ü┘é┘ê╪»╪⌐ ΓÇö ┘è┘à┘â┘å ╪ú┘å ╪¬┘â┘ê┘å ╪º┘ä╪ú╪»┘ê╪º╪¬ ┘à┘ü┘è╪»╪⌐ ╪¡┘é┘ï╪º╪î ╪Ñ╪░ ╪¬┘é╪»┘à ╪¬┘ê╪¼┘è┘ç┘ï╪º ╪¿┘à┘é┘è╪º╪│ ┘ä╪º ┘è┘à┘â┘å ┘ä╪ú┘è ┘à╪╣┘ä┘æ┘à ┘à╪¼╪º╪▒╪º╪¬┘ç.</p> <p>┘ä┘â┘å ╪º┘ä┘à╪«╪º╪╖╪▒ ┘â╪¿┘è╪▒╪⌐ ┘ê╪¬╪│╪¬╪¡┘é ╪º┘ä┘ü╪¡╪╡.</p> <p>┘å╪╕╪º┘à ╪º┘ä┘â╪¬╪º╪¿╪⌐ ╪º┘ä┘à╪»╪▒┘æ╪¿ ╪╣┘ä┘ë ┘à┘é╪º┘ä╪º╪¬ ╪│╪º╪¿┘é╪⌐ ┘è╪¬╪╣┘ä┘à ┘â┘è┘ü ┘â╪º┘å╪¬ ╪º┘ä┘à┘é╪º┘ä╪º╪¬ ╪º┘ä╪│╪º╪¿┘é╪⌐. ┘è┘à┘â┘å┘ç ┘à┘â╪º┘ü╪ú╪⌐ ╪º┘ä╪¿┘å┘è╪⌐ ╪º┘ä┘å┘à╪╖┘è╪⌐╪î ┘ê┘à╪╣╪º┘é╪¿╪⌐ ╪º┘ä┘ü╪▒╪»┘è╪⌐ ╪º┘ä╪ú╪│┘ä┘ê╪¿┘è╪⌐╪î ┘ê╪ú┘å ┘è╪«╪»╪╣┘ç ╪º┘ä╪╖┘ä╪º╪¿ ╪º┘ä╪░┘è┘å ┘è╪¬╪╣┘ä┘à┘ê┘å ╪º┘ä┘â╪¬╪º╪¿╪⌐ ┘ê┘ü┘é ╪º┘ä┘å┘à╪╖ ╪º┘ä╪░┘è ╪¬┘ü╪╢┘æ┘ä┘ç ╪º┘ä╪«┘ê╪º╪▒╪▓┘à┘è╪⌐. ┘ê╪º┘ä╪ú┘â╪½╪▒ ╪Ñ╪▓╪╣╪º╪¼┘ï╪º╪î ┘ä╪º ┘è┘à┘â┘å ┘ä┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪ú┘å ╪¬┘é╪▒╪ú ┘ü╪╣┘ä┘ï╪º. ╪¬┘â╪¬╪┤┘ü ╪Ñ╪┤╪º╪▒╪º╪¬ ╪Ñ╪¡╪╡╪º╪ª┘è╪⌐╪î ┘ä╪º ┘à╪╣┘å┘ë. ┘à┘é╪º┘ä ╪¡┘Å╪¼╪¼┘ç ╪¼┘à┘è┘ä╪⌐ ┘ä┘â┘å┘ç ┘è┘å╪¡╪▒┘ü ╪╣┘å ╪º┘ä┘à╪╣╪¬╪º╪» ┘é╪» ┘è┘Å┘é┘è┘æ┘à ╪¿┘é╪│┘ê╪⌐╪¢ ┘ê┘à┘é╪º┘ä ╪ú╪¼┘ê┘ü ┘è╪╣┘â╪│ ╪º┘ä╪┤┘â┘ä ╪º┘ä┘à╪¬┘ê┘é╪╣ ┘é╪» ┘è╪¡╪╕┘ë ╪¿╪»╪▒╪¼╪⌐ ╪│╪«┘è╪⌐.</p> <p>╪½┘à ┘ç┘å╪º┘â ╪│╪ñ╪º┘ä ┘à╪º ╪¬┘ü╪╣┘ä┘ç ╪º┘ä╪ú╪¬┘à╪¬╪⌐ ╪¿╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘å┘ü╪│┘ç╪º.</p> <p>╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘ç┘è ╪º┘ä╪ú┘é┘ê┘ë ╪╣┘å╪»┘à╪º ╪¬┘â┘ê┘å ┘ü┘ê╪▒┘è╪⌐ ┘ê┘à╪¡╪»╪»╪⌐ ┘ê┘à╪┤╪¼╪╣╪⌐. ┘è┘à┘â┘å ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪¬┘é╪»┘è┘à ╪ú┘ê┘ä ╪º╪½┘å╪¬┘è┘å╪î ┘ê╪Ñ┘ä┘ë ╪¡╪» ┘à╪º ╪º┘ä╪½╪º┘ä╪½╪⌐. ┘ä┘â┘å ╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪╣┘ä╪º┘é╪⌐ ╪ú┘è╪╢┘ï╪º. ╪╖╪º┘ä╪¿ ┘è╪¬┘ä┘é┘ë ╪»╪▒╪¼╪⌐ ┘ê╪¬╪╣┘ä┘è┘é╪º╪¬ ┘é┘ä┘è┘ä╪⌐ ┘à┘å ╪ó┘ä╪⌐ ╪¬┘ä┘é┘ë ┘à╪╣┘ä┘ê┘à╪º╪¬╪¢ ╪ú┘à╪º ╪╖╪º┘ä╪¿ ┘è┘å╪º┘é╪┤ ╪╣┘à┘ä┘ç ┘à╪╣ ┘à╪╣┘ä┘æ┘à ┘ü┘é╪» ╪¬┘ä┘é┘ë ╪┤┘è╪ª┘ï╪º ╪ú┘â╪½╪▒. ╪º┘ä╪ú┘ê┘ä ┘ü╪╣┘æ╪º┘ä. ┘ê╪º┘ä╪½╪º┘å┘è ╪¬╪¡┘ê┘è┘ä┘è. ┘è╪¡╪¬╪º╪¼ ╪º┘ä┘å╪╕╪º┘à ╪º┘ä╪¬╪╣┘ä┘è┘à┘è ╪º┘ä╪╡╪¡┘è ╪Ñ┘ä┘ë ┘â┘ä┘è┘ç┘à╪º╪î ┘ê┘è╪¼╪¿ ╪ú┘å ┘è┘é╪º┘ê┘à ╪Ñ╪║╪▒╪º╪í ╪¬╪▒┘â ╪º┘ä┘ü╪╣╪º┘ä┘è╪⌐ ╪¬╪╖╪▒╪» ╪º┘ä╪¬╪¡┘ê┘ä.</p> <p>╪ú┘â╪½╪▒ ┘à╪│╪º╪▒ ╪¡┘â┘è┘à ┘è╪¿╪»┘ê ┘ç╪¼┘è┘å┘ï╪º.</p> <p>╪»╪╣ ╪º┘ä╪ó┘ä╪º╪¬ ╪¬╪»┘è╪▒ ╪º┘ä╪▒┘ê╪¬┘è┘å┘è╪î ┘ê╪º┘ä┘é╪º╪¿┘ä ┘ä┘ä╪¬┘â╪▒╪º╪▒╪î ┘ê╪º┘ä╪ú╪¼╪▓╪º╪í ┘à┘å ╪º┘ä╪¬┘é┘è┘è┘à ╪º┘ä╪¬┘è ╪¬╪▒┘ç┘é ╪º┘ä┘à╪╣┘ä┘à┘è┘å. ┘ê╪º╪¡╪¬┘ü╪╕ ┘ä┘ä╪¿╪┤╪▒ ╪¿╪º┘ä╪¡┘â┘à ╪º┘ä╪░┘è ┘è╪¬╪╖┘ä╪¿ ╪º┘ä┘ü┘ç┘à╪î ┘ê╪º┘ä╪¬╪┤╪¼┘è╪╣ ╪º┘ä╪░┘è ┘è╪¬╪╖┘ä╪¿ ╪º┘ä╪¬╪╣╪º╪╖┘ü╪î ┘ê╪º┘ä┘à╪¡╪º╪»╪½╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¬╪╖┘ä╪¿ ╪º┘ä╪¡╪╢┘ê╪▒. ┘ç╪░╪º ╪º┘ä╪¬┘é╪│┘è┘à ┘ä┘ä╪╣┘à┘ä ┘ä┘è╪│ ╪¡┘ä┘ï╪º ┘ê╪│╪╖┘ï╪º╪¢ ╪Ñ┘å┘ç ┘à╪¿╪»╪ú ╪¬╪╡┘à┘è┘à.</p> <h2> 6. ╪»┘ê╪▒ ╪º┘ä┘à╪╣┘ä┘à ┘ü┘è ╪╣╪╡╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è </h2> <p>┘â┘ä ┘à╪¡╪º╪»╪½╪⌐ ╪╣┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¬╪╡┘ä ┘ü┘è ╪º┘ä┘å┘ç╪º┘è╪⌐ ╪Ñ┘ä┘ë ╪º┘ä╪│╪ñ╪º┘ä ╪º┘ä┘é┘ä┘é ┘å┘ü╪│┘ç: ┘ç┘ä ╪│╪¬╪¡┘ä ╪º┘ä╪ó┘ä╪º╪¬ ┘à╪¡┘ä ╪º┘ä┘à╪╣┘ä┘à┘è┘å╪ƒ</p> <p>╪º┘ä╪¼┘ê╪º╪¿ ╪º┘ä┘é╪╡┘è╪▒ ┘ç┘ê ┘ä╪º. ┘ê╪º┘ä╪¼┘ê╪º╪¿ ╪º┘ä╪ú╪╖┘ê┘ä ╪ú┘â╪½╪▒ ╪Ñ╪½╪º╪▒╪⌐ ┘ä┘ä╪º┘ç╪¬┘à╪º┘à.</p> <p>╪º┘ä╪¬╪»╪▒┘è╪│ ┘ä┘è╪│ ┘à╪¼╪▒╪» ┘å┘é┘ä ┘ä┘ä┘à╪╣┘ä┘ê┘à╪º╪¬. ┘ä┘ê ┘â╪º┘å ┘â╪░┘ä┘â╪î ┘ä┘â╪º┘å╪¬ ╪¡╪¼╪⌐ ╪º┘ä╪º╪│╪¬╪¿╪»╪º┘ä ┘é┘ê┘è╪⌐╪î ┘ä╪ú┘å ╪º┘ä╪ó┘ä╪º╪¬ ┘à┘à╪¬╪º╪▓╪⌐ ┘ü┘è ┘å┘é┘ä ╪º┘ä┘à╪╣┘ä┘ê┘à╪º╪¬. ┘ä┘â┘å ╪º┘ä╪¬╪»╪▒┘è╪│ ┘ç┘ê ╪ú┘è╪╢┘ï╪º ╪¬┘å┘à┘è╪⌐ ╪º┘ä╪╣┘ä╪º┘é╪º╪¬╪î ┘ê┘å┘à╪░╪¼╪⌐ ╪º┘ä┘ü╪╢┘ê┘ä╪î ┘ê╪º┘ä╪¬╪╣╪▒┘ü ╪╣┘ä┘ë ╪╖╪º┘ä╪¿ ┘è╪╣╪º┘å┘è ╪¿╪╡┘à╪¬╪î ┘ê╪º┘ä╪º╪¡╪¬┘ü╪º┘ä ╪¿╪╖╪º┘ä╪¿ ┘è┘ü┘ç┘à ╪ú╪«┘è╪▒┘ï╪º. ┘ç╪░┘ç ╪º┘ä╪ú┘ü╪╣╪º┘ä ╪¬╪¬╪╖┘ä╪¿ ╪¡╪╢┘ê╪▒┘ï╪º ╪¿╪┤╪▒┘è┘ï╪º. ┘ä╪º ┘è┘à┘â┘å ┘ä╪ú┘è ╪«┘ê╪º╪▒╪▓┘à┘è╪⌐ ╪ú┘å ╪¬┘å╪╕╪▒ ┘ü┘è ╪╣┘è┘å┘è ╪╖╪º┘ä╪¿ ┘ê╪¬┘é┘ê┘ä "╪ú╪ñ┘à┘å ╪¿┘â" ┘ê╪¬╪╣┘å┘è ╪░┘ä┘â.</p> <p>╪║┘è╪▒ ╪ú┘å ╪º┘ä╪│╪ñ╪º┘ä╪î ╪Ñ╪░╪º ╪ú┘Å╪╣┘è╪»╪¬ ╪╡┘è╪º╪║╪¬┘ç╪î ┘è╪╡╪¿╪¡ ╪╣╪º╪»┘ä┘ï╪º.</p> <p>┘ä┘å ┘è┘Å╪│╪¬╪¿╪»┘ä ╪º┘ä┘à╪╣┘ä┘à┘ê┘å. ┘ä┘â┘å ╪╣┘à┘ä ╪º┘ä╪¬╪»╪▒┘è╪│ ╪│┘è╪¬╪║┘è╪▒╪î ┘ê┘ç┘ê ┘è╪¬╪║┘è╪▒ ╪¿╪º┘ä┘ü╪╣┘ä.</p> <p>╪º┘ä┘à╪╣┘ä┘æ┘à ╪º┘ä╪░┘è ┘è╪¼╪¿ ╪ú┘å ┘è┘é╪╢┘è ╪│╪º╪╣╪º╪¬ ┘ü┘è ╪º┘ä╪¬╪╡╪¡┘è╪¡ ┘è┘à┘â┘å┘ç ╪¿╪»┘ä┘ï╪º ┘à┘å ╪░┘ä┘â ┘é╪╢╪º╪í ╪¬┘ä┘â ╪º┘ä╪│╪º╪╣╪º╪¬ ┘ü┘è ┘à┘é╪º╪¿┘ä╪⌐ ╪º┘ä╪╖┘ä╪º╪¿. ╪º┘ä┘à╪╣┘ä┘æ┘à ╪º┘ä╪░┘è ┘ä╪º ┘è╪│╪¬╪╖┘è╪╣ ╪Ñ╪╣╪╖╪º╪í ┘â┘ä ╪╖╪º┘ä╪¿ ╪º┘ç╪¬┘à╪º┘à┘ï╪º ┘ü╪▒╪»┘è┘ï╪º ┘è┘à┘â┘å┘ç╪î ╪¿╪»╪╣┘à ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è╪î ╪ú┘å ┘è┘é╪¬╪▒╪¿ ╪ú┘â╪½╪▒ ┘à┘å ┘ü╪╣┘ä ╪░┘ä┘â. ╪º┘ä┘à╪╣┘ä┘æ┘à ╪º┘ä╪░┘è ┘ç┘ê ╪º┘ä┘à╪╡╪»╪▒ ╪º┘ä┘ê╪¡┘è╪» ┘ä┘ä┘à╪╣╪▒┘ü╪⌐ ┘ü┘è ╪║╪▒┘ü╪⌐ ┘à┘å ╪º┘ä┘à╪¬╪╣┘ä┘à┘è┘å ┘è╪¬╪¡┘ê┘ä ╪Ñ┘ä┘ë ╪┤┘è╪í ╪ó╪«╪▒ ╪╣┘å╪»┘à╪º ╪¬╪¬┘ê┘ü╪▒ ╪º┘ä┘à╪╣╪▒┘ü╪⌐ ┘ü┘è ┘â┘ä ┘à┘â╪º┘å: ┘à┘ä╪º╪¡┘ï╪º╪î ┘ê┘à╪▒╪┤╪»┘ï╪º╪î ┘ê┘à╪»╪▒╪¿┘ï╪º.</p> <p>┘ç╪░╪º ╪º┘ä╪¬╪¡┘ê┘ä ┘ä┘è╪│ ┘à╪▒┘è╪¡┘ï╪º ╪¬┘à╪º┘à┘ï╪º. ┘è╪╖┘ä╪¿ ┘à┘å ╪º┘ä┘à╪╣┘ä┘à┘è┘å ╪º┘ä╪¬╪«┘ä┘è ╪╣┘å ┘é╪»╪▒ ┘à┘å ╪º┘ä╪│┘è╪╖╪▒╪⌐ ╪╣┘ä┘ë ┘ü╪╡┘ê┘ä┘ç┘à ╪º┘ä╪»╪▒╪º╪│┘è╪⌐╪î ┘ê╪º┘ä╪½┘é╪⌐ ╪¿╪ú┘å╪╕┘à╪⌐ ┘ä┘à ┘è╪¿┘å┘ê┘ç╪º ┘ê┘é╪» ┘ä╪º ┘è┘ü┘ç┘à┘ê┘å┘ç╪º ┘â┘ä┘è┘ï╪º. ┘ê┘è╪¬╪╖┘ä╪¿ ┘à┘ç╪º╪▒╪º╪¬ ╪¼╪»┘è╪»╪⌐ ΓÇö ╪º┘ä┘é╪»╪▒╪⌐ ╪╣┘ä┘ë ╪¬┘é┘è┘è┘à ╪¼┘ê╪»╪⌐ ╪ú╪»╪º╪⌐╪î ┘ê╪º┘ä╪¬╪╣╪▒┘ü ╪╣┘ä┘ë ╪¬╪¡┘è┘æ╪▓╪º╪¬┘ç╪º╪î ┘ê╪º╪╣╪¬╪▒╪º╪╢ ╪ú╪«╪╖╪º╪ª┘ç╪º. ┘ç╪░┘ç ┘ä┘è╪│╪¬ ┘à╪╖╪º┘ä╪¿ ╪¬╪º┘ü┘ç╪⌐.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪º┘ä┘à╪╣┘ä┘à: ┘é╪¿┘ä ╪º┘ä┘à╪╣┘ä┘à: ┘à╪╣ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ---------------------- -------------------------- ╪º┘ä┘à╪╡╪»╪▒ ╪º┘ä┘ê╪¡┘è╪» ┘ä┘ä┘à╪╣╪▒┘ü╪⌐ ┘à┘ä╪º╪¡ / ┘à╪»╪▒╪¿ ┘è╪╡╪¡╪¡ ┘â┘ä ┘à┘é╪º┘ä ┘è╪╡╪¡╪¡ ╪╣╪º┘ä┘è ╪º┘ä┘à╪«╪º╪╖╪▒ ┘ü┘é╪╖ ╪»╪▒╪│ ┘ê╪º╪¡╪» ┘ä┘ä╪¼┘à┘è╪╣ ┘è┘ê╪¼┘æ┘ç ╪º┘ä╪╣┘à┘ä ╪º┘ä╪┤╪«╪╡┘è ╪º┘ä┘ê╪¡┘è╪» ╪º┘ä╪░┘è ┘è┘à┘å╪¡ ╪º┘å╪¬╪¿╪º┘ç┘ï╪º ╪º┘å╪¬╪¿╪º┘ç ┘à╪▒┘â┘æ╪▓ ╪╣┘ä┘ë ╪º┘ä╪¡╪º╪¼╪⌐ ╪º┘ä╪¡┘é┘è┘é┘è╪⌐ </code></pre> </div> <p>╪│┘è╪¬┘é╪¿┘ä ╪¿╪╣╪╢ ╪º┘ä┘à╪╣┘ä┘à┘è┘å ╪º┘ä╪¬╪║┘è┘è╪▒. ┘ê╪│┘è┘é╪º┘ê┘à┘ç ╪ó╪«╪▒┘ê┘å╪î ┘ê┘à┘é╪º┘ê┘à╪¬┘ç┘à ┘ä┘è╪│╪¬ ┘à╪¼╪▒╪» ╪╣┘å╪º╪». ╪Ñ┘å┘ç╪º ╪¡╪░╪▒ ┘à┘ü┘ç┘ê┘à ╪¬╪¼╪º┘ç ┘é┘ê╪⌐ ┘ä┘à ┘è╪«╪¬╪º╪▒┘ê┘ç╪º ┘ê┘ä╪º ┘è╪│╪¬╪╖┘è╪╣┘ê┘å ╪º┘ä╪│┘è╪╖╪▒╪⌐ ╪╣┘ä┘è┘ç╪º ┘â┘ä┘è┘ï╪º. ╪º┘ä╪¬┘å┘ü┘è╪░ ╪º┘ä╪¡┘â┘è┘à ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ┘è┘é╪▒ ╪¿┘ç╪░╪º. ┘ä╪º ┘è┘ü╪▒╪╢ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘à┘å ╪º┘ä╪ú╪╣┘ä┘ë ┘à╪¬┘ê┘é╪╣┘ï╪º ╪º┘ä╪º┘à╪¬┘å╪º┘å. ┘è╪╣┘à┘ä ┘à╪╣ ╪º┘ä┘à╪╣┘ä┘à┘è┘å╪î ┘ê┘è┘Å╪»╪▒┘æ╪¿┘ç┘à╪î ┘ê┘è╪│╪¬┘à╪╣ ╪Ñ┘ä┘è┘ç┘à╪î ┘ê┘è╪╣╪º┘à┘ä┘ç┘à ┘â╪┤╪▒┘â╪º╪í ┘ä╪º ╪╣┘é╪¿╪º╪¬.</p> <p>┘ç┘å╪º┘â╪î ╪ú┘è╪╢┘ï╪º╪î ┘å┘é╪╖╪⌐ ╪ú╪╣┘à┘é ╪¡┘ê┘ä ╪º┘ä╪»┘ê╪▒ ╪º┘ä╪¿╪┤╪▒┘è.</p> <p>╪ú┘é┘ê┘ë ┘é┘ê╪⌐ ┘ü┘è ╪¡┘è╪º╪⌐ ╪┤╪«╪╡ ╪┤╪º╪¿╪î ╪¿╪╣╪» ╪º┘ä╪ú╪│╪▒╪⌐╪î ┘ç┘è ┘ü┘è ╪º┘ä╪║╪º┘ä╪¿ ╪º┘ä┘à╪╣┘ä┘æ┘à. ┘è┘à┘â┘å ┘ä┘ä┘à╪╣┘ä┘æ┘à ╪ú┘å ┘è╪║┘è╪▒ ┘à╪│╪º╪▒ ╪¡┘è╪º╪⌐ ╪¿╪╖╪▒┘é ┘è╪╡╪╣╪¿ ┘é┘è╪º╪│┘ç╪º ┘ê┘è╪│╪¬╪¡┘è┘ä ╪¬╪¡┘ê┘è┘ä┘ç╪º ╪Ñ┘ä┘ë ╪«┘ê╪º╪▒╪▓┘à┘è╪⌐. ┘ä╪º ┘è┘é┘ä┘ä ┘à╪¼┘è╪í ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘à┘å ┘ç╪░┘ç ╪º┘ä┘é┘ê╪⌐╪¢ ╪¿┘ä ╪Ñ┘å┘ç ┘è┘ü╪│╪¡┘ç╪º ┘ü┘è ╪º┘ä┘ê╪╢┘ê╪¡. ┘ê┘à╪╣ ╪¬┘ê┘ä┘è ╪º┘ä╪ó┘ä╪º╪¬ ╪º┘ä┘à┘ç╪º┘à ╪º┘ä┘à┘è┘â╪º┘å┘è┘â┘è╪⌐ ┘ä┘ä╪¬╪»╪▒┘è╪│╪î ┘è╪╡╪¿╪¡ ╪º┘ä╪╣┘à┘ä ╪º┘ä╪¿╪┤╪▒┘è ╪¡┘é┘ï╪º ΓÇö ╪º┘ä╪¬╪┤╪¼┘è╪╣╪î ┘ê╪º┘ä╪º╪╣╪¬┘é╪º╪»╪î ┘ê╪º┘ä╪╣┘ä╪º┘é╪⌐ ΓÇö ╪¿┘ê╪╢┘ê╪¡ ╪ú┘â╪¿╪▒ ┘à╪▒┘â╪▓ ╪º┘ä┘à┘ç┘å╪⌐╪î ┘ä╪º ╪ú┘é┘ä.</p> <p>╪º┘ä┘à╪╣┘ä┘æ┘à╪î ╪¿╪╣╪¿╪º╪▒╪⌐ ╪ú╪«╪▒┘ë╪î ┘ä┘è╪│ ╪╢╪¡┘è╪⌐ ╪╣╪╡╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è. ╪º┘ä┘à╪╣┘ä┘æ┘à ┘ç┘ê ╪ú┘â╪½╪▒ ┘à┘ê╪º╪▒╪»┘ç ╪¼┘ê┘ç╪▒┘è╪⌐.</p> <h2> 7. ┘é┘è╪º╪│ ┘à╪º ┘è┘å╪¼╪¡: ╪º┘ä╪ú╪»┘ä╪⌐ ┘ê╪º┘ä┘å╪¬╪º╪ª╪¼ </h2> <p>╪ú┘å ┘å╪╣╪¬┘é╪» ╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è╪│╪º╪╣╪» ╪º┘ä╪╖┘ä╪º╪¿ ╪╣┘ä┘ë ╪º┘ä╪¬╪╣┘ä┘à ╪┤┘è╪í. ┘ê╪ú┘å ┘å╪½╪¿╪¬ ╪░┘ä┘â ╪┤┘è╪í ╪ó╪«╪▒.</p> <p>┘é╪º╪╣╪»╪⌐ ╪º┘ä╪ú╪»┘ä╪⌐ ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¡╪»┘è╪½╪⌐ ┘ê┘ç┘è ╪║┘è╪▒ ┘à╪¬╪│╪º┘ê┘è╪⌐. ╪¿╪╣╪╢ ╪º┘ä┘à╪¼╪º┘ä╪º╪¬ ╪»┘Å╪▒╪│╪¬ ╪╣┘ä┘ë ┘å╪╖╪º┘é ┘ê╪º╪│╪╣╪¢ ┘è╪╣╪¬┘à╪» ╪¿╪╣╪╢┘ç╪º ╪º┘ä╪ó╪«╪▒ ╪╣┘ä┘ë ╪¡┘â╪º┘è╪º╪¬ ┘à┘ê╪╣┘É╪»╪⌐ ┘ê╪º╪»╪╣╪º╪í╪º╪¬ ╪º┘ä╪¿╪º╪╣╪⌐. ╪º┘ä╪¬┘é┘è┘è┘à ╪º┘ä╪¼╪º╪» ┘è╪¬╪╖┘ä╪¿ ╪ú┘å ┘å┘å╪╕╪▒ ╪Ñ┘ä┘ë ┘à╪º ┘è╪╕┘ç╪▒┘ç ╪º┘ä╪¿╪¡╪½ ┘ü╪╣┘ä┘ï╪º.</p> <p>┘ü┘è ╪º┘ä╪▒┘è╪º╪╢┘è╪º╪¬ ┘ê╪º┘ä┘à┘ê╪º╪» ╪º┘ä╪Ñ╪¼╪▒╪º╪ª┘è╪⌐╪î ╪¬┘â┘ê┘å ╪º┘ä╪ú╪»┘ä╪⌐ ╪ú┘â╪½╪▒ ╪¬╪┤╪¼┘è╪╣┘ï╪º. ╪ú╪¿┘ä╪║╪¬ ╪º┘ä╪¬╪¡┘ä┘è┘ä╪º╪¬ ╪º┘ä╪¬┘ä┘ê┘è╪⌐ ┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬╪»╪▒┘è╪│ ╪º┘ä╪░┘â┘è╪⌐ ╪╣┘å ╪¬╪ú╪½┘è╪▒╪º╪¬ ┘à╪¬┘ê╪│╪╖┘è╪⌐ ╪Ñ┘è╪¼╪º╪¿┘è╪⌐ ╪╣┘ä┘ë ┘å╪¬╪º╪ª╪¼ ╪º┘ä╪¬╪╣┘ä┘à╪î ┘à╪╣ ╪¿╪╣╪╢ ╪º┘ä╪»╪▒╪º╪│╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¼╪» ┘à┘â╪º╪│╪¿ ┘é╪º╪¿┘ä╪⌐ ┘ä┘ä┘à┘é╪º╪▒┘å╪⌐ ╪¿╪¬┘ä┘â ╪º┘ä╪«╪º╪╡╪⌐ ╪¿╪º┘ä╪¬╪»╪▒┘è╪│ ╪º┘ä╪¿╪┤╪▒┘è ╪º┘ä┘ü╪▒╪»┘è. ╪º╪¬╪│╪º┘é ┘ç╪░┘ç ╪º┘ä┘å╪¬╪º╪ª╪¼ ╪╣╪¿╪▒ ╪»╪▒╪º╪│╪º╪¬ ┘à╪¬╪╣╪»╪»╪⌐ ╪│╪¿╪¿ ╪¡┘é┘è┘é┘è ┘ä┘ä╪¬┘ü╪º╪ñ┘ä.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ┘é┘ê╪⌐ ╪º┘ä╪ú╪»┘ä╪⌐ ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪▒┘è╪º╪╢┘è╪º╪¬ / ╪Ñ╪¼╪▒╪º╪ª┘è ΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûê ┘é┘ê┘è╪⌐ ┘é╪▒╪º╪í╪⌐ / ┘ä╪║╪º╪¬ ΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûêΓûê ┘à╪╣╪¬╪»┘ä╪⌐ ┘â╪¬╪º╪¿╪⌐ / ┘ü┘ç┘à ΓûêΓûêΓûêΓûêΓûêΓûê ┘à╪«╪¬┘ä╪╖╪⌐ ╪Ñ╪¿╪»╪º╪╣ / ┘à╪¬┘é╪»┘à ΓûêΓûêΓûê ╪╢╪╣┘è┘ü╪⌐ </code></pre> </div> <p>┘ü┘è ╪º┘ä┘é╪▒╪º╪í╪⌐ ┘ê╪¬╪╣┘ä┘à ╪º┘ä┘ä╪║╪º╪¬╪î ╪º┘ä╪╡┘ê╪▒╪⌐ ╪ú┘â╪½╪▒ ╪º╪«╪¬┘ä╪º╪╖┘ï╪º. ╪º┘ä╪ú╪»┘ê╪º╪¬ ╪º┘ä╪¬┘è ╪¬┘é╪»┘à ┘à┘à╪º╪▒╪│╪⌐ ╪º┘ä┘à┘ü╪▒╪»╪º╪¬ ┘ê╪¬┘à╪º╪▒┘è┘å ╪º┘ä┘é╪▒╪º╪í╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ╪¬╪╕┘ç╪▒ ┘ü┘ê╪º╪ª╪». ╪º┘ä╪ú╪»┘ê╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¡╪º┘ê┘ä ╪¬┘é┘è┘è┘à ╪ú┘ê ╪¬╪¡╪│┘è┘å ╪º┘ä┘â╪¬╪º╪¿╪⌐ ┘ê╪º┘ä┘ü┘ç┘à ┘à┘å ╪º┘ä╪▒╪¬╪¿╪⌐ ╪º┘ä╪ú╪╣┘ä┘ë ╪¬╪╕┘ç╪▒ ┘å╪¬╪º╪ª╪¼ ╪ú┘é┘ä ╪º╪¬╪│╪º┘é┘ï╪º╪î ┘à┘à╪º ┘è╪╣┘â╪│ ╪╡╪╣┘ê╪¿╪⌐ ╪º┘ä┘à┘ç┘à╪⌐ ╪º┘ä╪ú╪│╪º╪│┘è╪⌐.</p> <p>┘ü┘è ╪º┘ä┘à╪¼╪º┘ä╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¬╪╖┘ä╪¿ ┘ü┘ç┘à┘ï╪º ┘à┘ü╪º┘ç┘è┘à┘è┘ï╪º ╪╣┘à┘è┘é┘ï╪º ╪ú┘ê ╪Ñ╪¿╪»╪º╪╣┘ï╪º ΓÇö ╪º┘ä╪╣┘ä┘ê┘à ╪º┘ä╪Ñ┘å╪│╪º┘å┘è╪⌐ ┘ê╪º┘ä┘ü┘å┘ê┘å ┘ê╪¡┘ä ╪º┘ä┘à╪┤┘â┘ä╪º╪¬ ╪º┘ä┘à╪¬┘é╪»┘à ΓÇö ╪¬┘â┘ê┘å ╪º┘ä╪ú╪»┘ä╪⌐ ╪ú╪▒┘é┘æ. ┘ç┘å╪º ╪¬╪╡╪¿╪¡ ╪¡╪»┘ê╪» ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪º┘ä╪¡╪º┘ä┘è╪î ┘ê╪╣╪¼╪▓┘ç ╪╣┘å ╪º┘ä┘ü┘ç┘à ╪º┘ä╪¡┘é┘è┘é┘è╪î ╪ú┘â╪½╪▒ ╪¿╪▒┘ê╪▓┘ï╪º. ╪»╪▒╪º╪│╪º╪¬ ┘é┘ä┘è┘ä╪⌐ ╪¬╪┤┘è╪▒ ╪Ñ┘ä┘ë ╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è┘à┘â┘å┘ç ╪º╪│╪¬╪¿╪»╪º┘ä ╪º┘ä╪¬╪╣┘ä┘è┘à ╪º┘ä╪¿╪┤╪▒┘è ╪º┘ä┘à╪º┘ç╪▒ ┘ü┘è ┘ç╪░┘ç ╪º┘ä┘à╪¼╪º┘ä╪º╪¬╪î ┘ê┘â╪½┘è╪▒ ┘à┘å┘ç╪º ╪¬╪┤┘è╪▒ ╪Ñ┘ä┘ë ╪ú┘å┘ç ┘ä╪º ┘è╪│╪¬╪╖┘è╪╣ ╪¿╪╣╪» ╪Ñ╪╢╪º┘ü╪⌐ ╪º┘ä┘â╪½┘è╪▒ ╪Ñ┘ä┘ë ┘à╪º ┘è┘ü╪╣┘ä┘ç ┘à╪╣┘ä┘æ┘à ╪¼┘è╪» ╪¿╪º┘ä┘ü╪╣┘ä.</p> <p>╪¿╪º┘ä┘à╪½┘ä ╪ú┘ç┘à┘è╪⌐ ╪│╪ñ╪º┘ä <em>┘à┘å</em> ╪¬╪│╪º╪╣╪» ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º.</p> <p>┘ç┘å╪º┘â ┘é┘ä┘é ┘à╪│╪¬┘à╪▒╪î ╪¬╪»╪╣┘à┘ç ╪¿╪╣╪╢ ╪º┘ä╪ú╪¿╪¡╪º╪½╪î ┘à┘å ╪ú┘å ╪º┘ä╪ú╪»┘ê╪º╪¬ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ╪¬╪│╪º╪╣╪» ╪º┘ä╪╖┘ä╪º╪¿ ╪º┘ä╪░┘è┘å ┘è╪¡╪¬╪º╪¼┘ê┘å ╪ú┘é┘ä ┘à╪│╪º╪╣╪»╪⌐ ┘ê╪¬┘ü╪┤┘ä ┘ü┘è ╪º┘ä┘ê╪╡┘ê┘ä ╪Ñ┘ä┘ë ┘à┘å ┘è╪¡╪¬╪º╪¼┘ê┘å┘ç╪º ╪ú┘â╪½╪▒. ╪º┘ä╪╖┘ä╪º╪¿ ╪░┘ê┘ê ╪º┘ä╪¬┘å╪╕┘è┘à ╪º┘ä╪░╪º╪¬┘è ╪º┘ä┘à╪¬╪╖┘ê╪▒╪î ┘ê╪º┘ä┘é╪▒╪º╪í╪⌐ ╪º┘ä╪¼┘è╪»╪⌐╪î ┘ê╪º┘ä╪¿┘è┘ê╪¬ ╪º┘ä╪»╪º╪╣┘à╪⌐ ┘é╪» ┘è╪│╪¬┘ü┘è╪»┘ê┘å ╪ú┘â╪½╪▒ ┘à┘å ╪º┘ä╪¬╪╣┘ä┘à ╪º┘ä┘à╪»╪╣┘ê┘à ╪¿╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ä╪ú┘å┘ç┘à ┘è╪│╪¬╪╖┘è╪╣┘ê┘å ╪º┘ä╪¬┘å┘é┘ä ┘ü┘è┘ç ╪¿╪º╪│╪¬┘é┘ä╪º┘ä┘è╪⌐. ╪ú┘à╪º ╪º┘ä╪╖┘ä╪º╪¿ ╪º┘ä╪░┘è┘å ┘è┘ü╪¬┘é╪▒┘ê┘å ╪Ñ┘ä┘ë ┘ç╪░┘ç ╪º┘ä┘à╪▓╪º┘è╪º ┘ü┘é╪» ┘è╪╡╪º╪▒╪╣┘ê┘å ┘ä╪º╪│╪¬╪«╪»╪º┘à ╪º┘ä╪ú╪»┘ê╪º╪¬ ╪¿┘ü╪╣╪º┘ä┘è╪⌐╪î ┘ê┘è╪¬╪▒╪º╪¼╪╣┘ê┘å ╪ú┘â╪½╪▒. ╪¬┘à┘è┘ä ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪Ñ┘ä┘ë ╪¬╪╢╪«┘è┘à ╪ú┘ê╪¼┘ç ╪╣╪»┘à ╪º┘ä┘à╪│╪º┘ê╪º╪⌐ ╪º┘ä┘é╪º╪ª┘à╪⌐ ╪¿╪»┘ä┘ï╪º ┘à┘å ┘à╪¡┘ê┘ç╪º╪î ┘ê┘è┘Å╪╕┘ç╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪╣┘ä╪º┘à╪º╪¬ ┘à┘é┘ä┘é╪⌐ ╪╣┘ä┘ë ╪º╪¬╪¿╪º╪╣ ╪º┘ä┘å┘à╪╖ ┘å┘ü╪│┘ç.</p> <p>┘ç┘å╪º┘â ╪ú┘è╪╢┘ï╪º ╪¬╪¡╪»┘è ╪º┘ä╪ú╪»┘ä╪⌐ ┘å┘ü╪│┘ç╪º.</p> <p>╪¼╪▓╪í ┘â╪¿┘è╪▒ ┘à┘å ╪º┘ä╪¿╪¡╪½ ┘ü┘è ┘ç╪░╪º ╪º┘ä┘à╪¼╪º┘ä ╪¬┘à┘ê┘ä┘ç ╪º┘ä╪┤╪▒┘â╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¿┘è╪╣ ╪º┘ä╪ú╪»┘ê╪º╪¬. ┘ê╪¼╪▓╪í ┘â╪¿┘è╪▒ ┘è┘Å╪¼╪▒┘ë ┘ü┘è ╪╕╪▒┘ê┘ü ┘é╪╡┘è╪▒╪⌐ ┘à╪╡╪╖┘å╪╣╪⌐ ╪¿╪»┘ä┘ï╪º ┘à┘å ╪º┘ä┘ü╪╡┘ê┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪⌐ ╪º┘ä╪¡┘é┘è┘é┘è╪⌐ ╪╣┘ä┘ë ┘à╪»┘ë ┘ü╪¬╪▒╪º╪¬ ╪╖┘ê┘è┘ä╪⌐. ╪º┘å╪¡┘è╪º╪▓ ╪º┘ä┘å╪┤╪▒ ΓÇö ┘à┘è┘ä ╪º┘ä┘à╪¼┘ä╪º╪¬ ╪Ñ┘ä┘ë ┘å╪┤╪▒ ╪º┘ä┘å╪¬╪º╪ª╪¼ ╪º┘ä╪Ñ┘è╪¼╪º╪¿┘è╪⌐ ΓÇö ┘è╪┤┘ê┘æ┘ç ╪º┘ä╪╡┘ê╪▒╪⌐ ╪ú┘â╪½╪▒. ┘è╪¼╪¿ ╪ú┘å ┘è╪╣╪º┘à┘ä ╪º┘ä╪¬┘é┘è┘è┘à ╪º┘ä╪╡╪º╪»┘é ╪º┘ä┘å╪¬╪º╪ª╪¼ ╪º┘ä┘ê╪º╪╣╪»╪⌐ ╪¿╪»╪▒╪¼╪⌐ ┘à┘å ╪º┘ä╪┤┘â ┘ê┘è╪╖╪º┘ä╪¿ ╪¿╪º┘ä╪¬┘â╪▒╪º╪▒ ┘ü┘è ╪│┘è╪º┘é╪º╪¬ ╪ú╪╡┘è┘ä╪⌐.</p> <p>╪º┘ä╪º╪│╪¬┘å╪¬╪º╪¼ ╪º┘ä┘â┘ä┘è ┘ä┘è╪│ ┘à┘å╪¬╪┤┘è┘ï╪º ┘ê┘ä╪º ┘à╪│╪¬┘ç┘è┘å┘ï╪º.</p> <p>┘è┘à┘â┘å ┘ä┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪¬╪¡╪│┘è┘å ╪º┘ä╪¬╪╣┘ä┘à ┘ü┘è ╪º┘ä╪╕╪▒┘ê┘ü ╪º┘ä┘à┘å╪º╪│╪¿╪⌐. ╪¬╪┤┘à┘ä ╪¬┘ä┘â ╪º┘ä╪╕╪▒┘ê┘ü ╪ú╪╡┘ê┘ä ╪¬╪»╪▒┘è╪│ ╪│┘ä┘è┘à╪⌐╪î ┘ê╪ú╪»┘ê╪º╪¬ ╪¼┘è╪»╪⌐ ╪º┘ä╪¬╪╡┘à┘è┘à╪î ┘ê┘à╪╣┘ä┘à┘è┘å ┘à╪»╪▒┘æ╪¿┘è┘å╪î ┘ê╪º┘ç╪¬┘à╪º┘à┘ï╪º ╪¿╪º┘ä╪Ñ┘å╪╡╪º┘ü. ╪╣┘å╪»┘à╪º ╪¬┘â┘ê┘å ┘ç╪░┘ç ╪º┘ä╪╣┘å╪º╪╡╪▒ ╪¡╪º╪╢╪▒╪⌐╪î ╪¬┘â┘ê┘å ╪º┘ä┘à┘â╪º╪│╪¿ ╪¡┘é┘è┘é┘è╪⌐. ┘ê╪╣┘å╪»┘à╪º ╪¬┘â┘ê┘å ╪║╪º╪ª╪¿╪⌐╪î ┘ä╪º ╪¬┘ü┘è╪» ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘â╪½┘è╪▒┘ï╪º ┘ê┘é╪» ╪¬╪╢╪▒. ╪º┘ä┘à╪¬╪║┘è╪▒ ╪º┘ä╪░┘è ┘è╪¡╪»╪» ╪º┘ä┘å╪¼╪º╪¡ ┘ä┘è╪│ ╪¬╪╣┘é┘è╪» ╪º┘ä╪«┘ê╪º╪▒╪▓┘à┘è╪⌐. ╪Ñ┘å┘ç ╪¼┘ê╪»╪⌐ ╪º┘ä╪¿┘è╪ª╪⌐ ╪º┘ä╪¬┘è ╪¬╪╣┘à┘ä ┘ü┘è┘ç╪º ╪º┘ä╪«┘ê╪º╪▒╪▓┘à┘è╪⌐.</p> <h2> 8. ╪º┘ä╪ú╪«┘ä╪º┘é ┘ê╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ┘ê╪º┘ä╪¬╪¡┘è┘æ╪▓ </h2> <p>┘ä┘å ╪¬┘â╪¬┘à┘ä ╪ú┘è ┘à┘å╪º┘é╪┤╪⌐ ╪╣┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪»┘ê┘å ┘à┘ê╪º╪¼┘ç╪⌐ ╪º┘ä╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪ú╪«┘ä╪º┘é┘è╪⌐ ╪º┘ä╪¬┘è ╪¬┘ä┘é┘è ╪¿╪╕┘ä╪º┘ä┘ç╪º ╪╣┘ä┘ë ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º. ┘ç╪░┘ç ┘ä┘è╪│╪¬ ╪ú┘ü┘â╪º╪▒┘ï╪º ┘ä╪º╪¡┘é╪⌐. ╪Ñ┘å┘ç╪º ┘à╪▒┘â╪▓┘è╪⌐.</p> <p>╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ┘ç┘è ╪º┘ä╪┤╪║┘ä ╪º┘ä╪┤╪º╪║┘ä ╪º┘ä╪ú┘ê┘ä ┘ê╪º┘ä╪ú┘â╪½╪▒ ╪Ñ┘ä╪¡╪º╪¡┘ï╪º.</p> <p>╪░┘â╪º╪í ╪º╪╡╪╖┘å╪º╪╣┘è ╪¬╪╣┘ä┘è┘à┘è ┘à╪¿┘å┘è ╪╣┘ä┘ë ╪º┘ä╪¿┘è╪º┘å╪º╪¬. ┘ä╪¬╪«╪╡┘è╪╡ ╪º┘ä╪¬╪╣┘ä┘à╪î ┘è╪¼┘à╪╣ ╪º┘ä┘å╪╕╪º┘à ┘à╪╣┘ä┘ê┘à╪º╪¬ ╪╣┘å ┘à╪º ┘è╪╣╪▒┘ü┘ç ╪º┘ä╪╖╪º┘ä╪¿╪î ┘ê┘â┘è┘ü ┘è╪¬╪╣┘ä┘à╪î ┘ê╪¿╪ú┘è ╪│╪▒╪╣╪⌐ ┘è╪¬┘é╪»┘à╪î ┘ê╪ú┘è┘å ┘è┘ê╪º╪¼┘ç ╪º┘ä╪╡╪╣┘ê╪¿╪º╪¬. ╪ú╪¡┘è╪º┘å┘ï╪º ┘è╪¼┘à╪╣ ╪ú┘â╪½╪▒ ╪¿┘â╪½┘è╪▒: ╪╢╪║╪╖╪º╪¬ ╪º┘ä┘à┘ü╪º╪¬┘è╪¡╪î ┘ê╪ú┘å┘à╪º╪╖ ╪º┘ä┘é╪▒╪º╪í╪⌐╪î ┘ê╪º┘ä╪º╪│╪¬╪¼╪º╪¿╪º╪¬ ╪º┘ä╪╣╪º╪╖┘ü┘è╪⌐ ╪º┘ä┘à╪│╪¬╪»┘ä ╪╣┘ä┘è┘ç╪º ┘à┘å ╪¬╪╣╪º╪¿┘è╪▒ ╪º┘ä┘ê╪¼┘ç ╪ú┘ê ╪│╪▒╪╣╪º╪¬ ╪º┘ä┘â╪¬╪º╪¿╪⌐. ┘ç╪░┘ç ╪¿┘è╪º┘å╪º╪¬ ╪¡┘à┘è┘à╪⌐╪î ╪¬┘Å╪¼┘à╪╣ ┘à┘å ╪º┘ä╪ú╪╖┘ü╪º┘ä╪î ╪║╪º┘ä╪¿┘ï╪º ╪»┘ê┘å ┘à┘ê╪º┘ü┘é╪⌐ ╪░╪º╪¬ ┘à╪╣┘å┘ë.</p> <p>╪¡╪¼┘à ╪º┘ä╪¼┘à╪╣ ┘à╪░┘ç┘ä╪î ┘ê╪º┘ä╪ú╪║╪▒╪º╪╢ ┘ä┘è╪│╪¬ ┘ê╪º╪╢╪¡╪⌐ ╪»╪º╪ª┘à┘ï╪º. ╪º┘ä╪¿┘è╪º┘å╪º╪¬ ╪º┘ä┘à╪¼┘à╪╣╪⌐ ┘ä╪¬╪¡╪│┘è┘å ╪º┘ä╪¬╪╣┘ä┘à ┘è┘à┘â┘å ╪º╪│╪¬╪«╪»╪º┘à┘ç╪º ┘ä╪║╪º┘è╪º╪¬ ╪ú╪«╪▒┘ë ΓÇö ╪¬╪▒╪¬┘è╪¿ ╪º┘ä┘à╪»╪º╪▒╪│╪î ┘ê╪¬╪│┘ê┘è┘é ╪º┘ä┘à┘å╪¬╪¼╪º╪¬ ┘ä┘ä╪╖┘ä╪º╪¿╪î ┘ê╪¿┘å╪º╪í ┘à┘ä┘ü╪º╪¬ ╪¬╪╣╪▒┘è┘ü┘è╪⌐ ╪¬╪¬╪¿╪╣ ╪º┘ä╪┤╪«╪╡ ╪╖┘ê┘è┘ä┘ï╪º ╪¿╪╣╪» ┘à╪║╪º╪»╪▒╪¬┘ç ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è. ╪º┘ä╪ú╪╖┘ü╪º┘ä ╪º┘ä╪░┘è┘å ╪¬┘Å╪¿┘å┘ë ┘à┘ä┘ü╪º╪¬┘ç┘à ╪º┘ä┘è┘ê┘à ┘ä┘è╪│┘ê╪º ┘ü┘è ┘ê╪╢╪╣ ┘è╪│┘à╪¡ ┘ä┘ç┘à ╪¿╪º┘ä┘à┘ê╪º┘ü┘é╪⌐ ╪╣┘ä┘ë ┘â┘è┘ü┘è╪⌐ ╪º╪│╪¬╪«╪»╪º┘à ╪¿┘è╪º┘å╪º╪¬┘ç┘à ╪¿╪╣╪» ╪╣╪┤╪▒┘è┘å ╪╣╪º┘à┘ï╪º. ╪╣╪»┘à ╪º┘ä╪¬┘à╪º╪½┘ä ┘ç╪░╪º ╪¿┘è┘å ╪º┘ä┘é┘ê╪⌐ ┘ê╪º┘ä┘å╪¬┘è╪¼╪⌐ ┘è╪│╪¬╪¡┘é ╪º┘ç╪¬┘à╪º┘à┘ï╪º ╪ú┘â╪¿╪▒ ╪¿┘â╪½┘è╪▒ ┘à┘à╪º ╪¬┘ä┘é╪º┘ç.</p> <p>╪½┘à ┘ç┘å╪º┘â ╪º┘ä╪¬╪¡┘è┘æ╪▓.</p> <p>╪¬╪¬╪╣┘ä┘à ╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘à┘å ╪º┘ä╪¿┘è╪º┘å╪º╪¬╪î ┘ê╪º┘ä╪¿┘è╪º┘å╪º╪¬ ╪¬╪╣┘â╪│ ╪º┘ä╪╣╪º┘ä┘à╪î ╪¿┘à╪º ┘ü┘è ╪░┘ä┘â ╪ú┘ê╪¼┘ç ╪╕┘ä┘à┘ç. ┘å╪╕╪º┘à ┘à╪»╪▒┘æ╪¿ ╪╣┘ä┘ë ┘à┘é╪º┘ä╪º╪¬ ┘à┘å ┘ü╪ª╪⌐ ╪»┘è┘à┘ê╪║╪▒╪º┘ü┘è╪⌐ ┘é╪» ┘è┘é┘ä┘æ┘ä ┘à┘å┘ç╪¼┘è┘ï╪º ┘à┘å ┘é┘è┘à╪⌐ ┘â╪¬╪º╪¿╪⌐ ┘ü╪ª╪⌐ ╪ú╪«╪▒┘ë. ┘å╪╕╪º┘à ╪¬┘â┘è┘ü┘è ┘à╪¡╪│┘Ä┘æ┘å ╪╣┘ä┘ë ╪ú╪»╪º╪í ╪╖┘ä╪º╪¿ ╪░┘ê┘è ┘à┘ê╪º╪▒╪» ╪¼┘è╪»╪⌐ ┘é╪» ┘è╪│╪º╪í ╪¬┘é┘è┘è┘à ╪º┘ä╪╖┘ä╪º╪¿ ┘à┘å ╪«┘ä┘ü┘è╪º╪¬ ┘à╪«╪¬┘ä┘ü╪⌐ ┘ê┘è╪«╪»┘à┘ç┘à ╪»┘ê┘å ╪º┘ä┘à╪│╪¬┘ê┘ë. ╪º┘ä╪¬╪¡┘è┘æ╪▓ ┘ü┘è ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ä┘è╪│ ╪«╪╖╪ú┘ï ┘è╪╕┘ç╪▒ ╪ú╪¡┘è╪º┘å┘ï╪º╪¢ ╪Ñ┘å┘ç ╪│┘à╪⌐ ┘ç┘è┘â┘ä┘è╪⌐ ╪¬╪╕┘ç╪▒ ┘â┘ä┘à╪º ┘â╪º┘å╪¬ ╪º┘ä╪¿┘è╪º┘å╪º╪¬ ╪º┘ä╪ú╪│╪º╪│┘è╪⌐ ╪║┘è╪▒ ┘à╪¬╪│╪º┘ê┘è╪⌐╪î ┘ê╪╣╪¿╪í ╪º┘â╪¬╪┤╪º┘ü┘ç ┘ê╪¬╪╡╪¡┘è╪¡┘ç ┘ä╪º ┘è┘é╪╣ ╪╣┘ä┘ë ╪º┘ä╪¿┘è╪º┘å╪º╪¬ ╪¿┘ä ╪╣┘ä┘ë ╪º┘ä╪¿╪┤╪▒ ╪º┘ä╪░┘è┘å ┘è╪╡┘à┘à┘ê┘å ╪º┘ä┘å╪╕╪º┘à ┘ê┘è┘å╪┤╪▒┘ê┘å┘ç.</p> <p>┘ç┘å╪º┘â ╪ú┘è╪╢┘ï╪º ╪¿┘Å╪╣╪» ╪ú╪«┘ä╪º┘é┘è ╪ú╪»┘é: ╪«╪╖╪▒ ╪º┘ä┘à╪▒╪º┘é╪¿╪⌐.</p> <p>╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è ╪º┘ä╪░┘è ╪¬┘Å╪▒╪º┘é┘Ä╪¿ ┘ü┘è┘ç ┘â┘ä ╪¡╪▒┘â╪⌐╪î ┘ê┘è┘Å╪│╪¼┘Ä┘æ┘ä ┘â┘ä ╪«╪╖╪ú╪î ┘ê╪¬┘Å╪¡╪│╪¿ ┘â┘ä ╪¬┘ê┘é┘ü╪⌐╪î ┘ç┘ê ┘ü╪╡┘ä ╪¬╪║┘è┘æ╪▒╪¬ ╪╖╪¿┘è╪╣╪¬┘ç╪î ╪│┘ê╪º╪í ╪ú╪▒╪º╪» ╪░┘ä┘â ╪ú╪¡╪» ╪ú┘à ┘ä╪º. ╪º┘ä╪╖┘ä╪º╪¿ ╪¬╪¡╪¬ ╪º┘ä┘à┘ä╪º╪¡╪╕╪⌐ ╪º┘ä┘à╪│╪¬┘à╪▒╪⌐ ┘è╪¬╪╣┘ä┘à┘ê┘å ╪¿╪┤┘â┘ä ┘à╪«╪¬┘ä┘ü. ┘è┘é┘ä┘ä┘ê┘å ┘à┘å ╪º┘ä┘à╪«╪º╪╖╪▒╪⌐. ┘è┘à╪¬╪½┘ä┘ê┘å. ╪º┘ä╪╡┘ü╪º╪¬ ╪º┘ä╪¬┘è ┘è┘ü╪¬╪▒╪╢ ╪ú┘å ┘è╪▓╪▒╪╣┘ç╪º ╪º┘ä╪¬╪╣┘ä┘è┘à ΓÇö ╪º┘ä┘ü╪╢┘ê┘ä╪î ┘ê╪º┘ä╪¬╪¼╪▒┘è╪¿╪î ┘ê╪º┘ä╪┤╪¼╪º╪╣╪⌐ ╪╣┘ä┘ë ╪º┘ä╪«╪╖╪ú ΓÇö ┘è┘à┘â┘å ┘é┘à╪╣┘ç╪º ╪¿╪╡┘à╪¬ ╪¿╪¿┘è╪ª╪⌐ ╪¬╪▒╪º┘é╪¿ ╪╣┘å ┘é╪▒╪¿ ╪┤╪»┘è╪».</p> <p>┘ä╪º ┘è╪╣┘å┘è ╪ú┘è ┘à┘å ┘ç╪░╪º ╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è╪¼╪¿ ╪ú┘å ┘è╪▒┘ü╪╢. ┘ä┘â┘å┘ç ┘è╪╣┘å┘è ╪ú┘å ╪¬╪¿┘å┘è┘ç ┘è╪¬╪╖┘ä╪¿ ╪│┘é╪º┘ä╪⌐ ╪ú╪«┘ä╪º┘é┘è╪⌐ ┘à╪¬╪╣┘à╪»╪⌐: ┘é┘ê╪º╪╣╪» ┘ê╪º╪╢╪¡╪⌐ ╪╣┘å ╪º┘ä╪¿┘è╪º┘å╪º╪¬╪î ┘ê╪¡┘à╪º┘è╪º╪¬ ┘é┘ê┘è╪⌐ ┘ä┘ä┘à┘ê╪º┘ü┘é╪⌐╪î ┘ê╪Ñ╪┤╪▒╪º┘ü ┘à╪│╪¬┘é┘ä╪î ┘ê╪┤┘ü╪º┘ü┘è╪⌐ ╪¡┘é┘è┘é┘è╪⌐ ╪¡┘ê┘ä ┘à╪º ╪¬┘ü╪╣┘ä┘ç ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪¿╪º┘ä┘à╪╣┘ä┘ê┘à╪º╪¬ ╪º┘ä╪¬┘è ╪¬╪¼┘à╪╣┘ç╪º. ╪º┘ä╪ú╪«┘ä╪º┘é ┘ä┘è╪│╪¬ ┘à┘è╪▓╪⌐ ╪¬┘Å╪╢╪º┘ü ┘ä╪º╪¡┘é┘ï╪º. ╪Ñ┘å┘ç╪º ┘é┘è╪» ┘è╪¼╪¿ ╪ú┘å ┘è┘Å╪»┘à╪¼ ┘à┘å ╪º┘ä╪¿╪»╪º┘è╪⌐.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪º┘ä╪¬┘ê╪º╪▓┘å ╪º┘ä╪░┘è ┘è╪¼╪¿ ╪ú┘å ┘å╪¡┘é┘é┘ç | ________/ \________ / \ ┘à┘â╪│╪¿ ╪¬╪╣┘ä┘è┘à┘è ╪¬┘â┘ä┘ü╪⌐ ╪¿╪┤╪▒┘è╪⌐ ╪¬╪╣┘ä┘æ┘à ┘à╪«╪╡╪╡ ┘ü┘é╪»╪º┘å ╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ╪¬╪║╪░┘è╪⌐ ╪▒╪º╪¼╪╣╪⌐ ┘ü┘ê╪▒┘è╪⌐ ╪¬╪¡┘è┘æ╪▓ ┘ê┘à┘ä┘ü╪º╪¬ ╪¬╪╣╪▒┘è┘ü┘è╪⌐ ┘â┘ä ╪╖╪º┘ä╪¿ ┘è┘Å╪«╪»┘Ä┘à ╪º┘ä╪¬╪╣┘ä┘à ╪¬╪¡╪¬ ╪º┘ä┘à╪▒╪º┘é╪¿╪⌐ ___________________ ___________________ ^ ^ |_____ ╪Ñ╪¿┘é╪º╪í ╪º┘ä╪╣┘è┘å ╪╣┘ä┘ë ┘â┘ä┘è┘ç┘à╪º _____| </code></pre> </div> <h2> 9. ╪º┘ä╪│┘è╪º╪│╪⌐ ┘ê╪º┘ä╪¬┘å┘ü┘è╪░ </h2> <p>╪º┘ä╪│┘è╪º╪│╪⌐ ╪º┘ä╪¼┘è╪»╪⌐ ┘ç┘è ╪º┘ä┘ü╪▒┘é ╪¿┘è┘å ╪ú╪»╪º╪⌐ ╪¬╪«╪»┘à ┘ê╪ú╪»╪º╪⌐ ╪¬╪│┘è╪╖╪▒.</p> <p>┘ä╪º ╪¬╪¬╪¡╪»╪» ┘à╪│╪º╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ╪¿╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘ê╪¡╪»┘ç╪º. ╪¬╪┤┘â┘ä┘ç ┘é╪▒╪º╪▒╪º╪¬ ╪¬┘Å╪¬╪«╪░ ┘ü┘è ╪º┘ä┘à╪¼╪º┘ä╪│ ╪º┘ä╪¬╪┤╪▒┘è╪╣┘è╪⌐ ┘ê╪º┘ä┘ê╪▓╪º╪▒╪º╪¬ ┘ê┘à╪¼╪º┘ä╪│ ╪º┘ä┘à╪»╪º╪▒╪│ ┘ê┘à┘â╪º╪¬╪¿ ╪º┘ä┘à┘å╪º╪╖┘é. ╪¬╪¡╪»╪» ┘ç╪░┘ç ╪º┘ä┘é╪▒╪º╪▒╪º╪¬ ┘à╪º ┘ç┘ê ┘à╪│┘à┘ê╪¡ ╪¿┘ç╪î ┘ê┘à╪º ┘è┘Å┘à┘ê┘æ┘ä╪î ┘ê┘à╪º ┘è┘Å╪»╪▒┘Ä┘æ╪│╪î ┘ê┘à┘å ┘è┘Å╪¡┘à┘ë.</p> <p>╪º┘ä╪ú┘ê┘ä┘ê┘è╪⌐ ╪º┘ä╪│┘è╪º╪│┘è╪⌐ ╪º┘ä╪ú┘ê┘ä┘ë ┘ç┘è ╪¡┘à╪º┘è╪⌐ ╪º┘ä╪¿┘è╪º┘å╪º╪¬. ┘è╪│╪¬╪¡┘é ╪º┘ä╪╖┘ä╪º╪¿ ┘å┘ü╪│ ╪╢┘à╪º┘å╪º╪¬ ╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ┘ü┘è ╪º┘ä┘ü╪╢╪º╪í╪º╪¬ ╪º┘ä╪▒┘é┘à┘è╪⌐ ╪º┘ä╪¬┘è ┘è╪¬┘à╪¬╪╣┘ê┘å ╪¿┘ç╪º ╪¿╪º┘ä┘ü╪╣┘ä ┘ü┘è ╪º┘ä┘ü╪╢╪º╪í╪º╪¬ ╪º┘ä┘à╪º╪»┘è╪⌐. ┘è╪¼╪¿ ╪ú┘å ╪¬╪¡┘â┘à ┘é┘ê╪º╪╣╪» ┘é┘ê┘è╪⌐ ┘à╪º ╪º┘ä╪¿┘è╪º┘å╪º╪¬ ╪º┘ä╪¬┘è ┘è┘à┘â┘å ╪¼┘à╪╣┘ç╪º╪î ┘ê┘à┘å ┘è┘à┘â┘å┘ç ╪º┘ä┘ê╪╡┘ê┘ä ╪Ñ┘ä┘è┘ç╪º╪î ┘ê┘à╪»╪⌐ ╪º┘ä╪º╪¡╪¬┘ü╪º╪╕ ╪¿┘ç╪º╪î ┘ê┘à╪º ╪Ñ╪░╪º ┘â╪º┘å ┘è┘à┘â┘å ╪º╪│╪¬╪«╪»╪º┘à┘ç╪º ┘ä╪ú┘è ╪║╪▒╪╢ ┘è╪¬╪¼╪º┘ê╪▓ ╪¬╪╣┘ä┘è┘à ╪º┘ä╪╖╪º┘ä╪¿ ┘å┘ü╪│┘ç. ┘è╪¼╪¿ ╪ú┘å ╪¬┘â┘ê┘å ┘ç╪░┘ç ╪º┘ä┘é┘ê╪º╪╣╪» ┘é╪º╪¿┘ä╪⌐ ┘ä┘ä╪Ñ┘å┘ü╪º╪░╪î ┘ê╪¬╪¬╪▒╪¬╪¿ ╪╣┘ä┘è┘ç╪º ╪╣┘ê╪º┘é╪¿ ╪¡┘é┘è┘é┘è╪⌐ ╪╣┘å╪» ╪º┘ä╪º┘å╪¬┘ç╪º┘â.</p> <p>╪º┘ä╪ú┘ê┘ä┘ê┘è╪⌐ ╪º┘ä╪½╪º┘å┘è╪⌐ ┘ç┘è ╪º┘ä╪Ñ┘å╪╡╪º┘ü. ╪Ñ╪░╪º ╪¬┘Å╪▒┘â ┘ä┘ä╪│┘ê┘é╪î ┘ü╪│╪¬╪¬╪»┘ü┘é ╪ú┘ü╪╢┘ä ╪ú╪»┘ê╪º╪¬ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪Ñ┘ä┘ë ╪º┘ä┘à╪»╪º╪▒╪│ ╪º┘ä╪¬┘è ╪¬╪│╪¬╪╖┘è╪╣ ╪»┘ü╪╣ ╪½┘à┘å┘ç╪º╪î ┘à┘à╪º ┘è┘ê╪│┘æ╪╣ ╪º┘ä┘ü╪¼┘ê╪⌐ ╪¿┘è┘å ╪º┘ä┘à┘å╪º╪╖┘é ╪º┘ä╪║┘å┘è╪⌐ ┘ê╪º┘ä┘ü┘é┘è╪▒╪⌐╪î ╪¿┘è┘å ╪º┘ä╪¿┘ä╪»╪º┘å ╪░╪º╪¬ ╪º┘ä┘à┘ê╪º╪▒╪» ┘ê╪º┘ä╪¿┘ä╪»╪º┘å ╪º┘ä╪¬┘è ┘ä╪º ╪¬┘à┘ä┘â┘ç╪º. ┘ä┘ä╪│┘è╪º╪│╪⌐ ╪º┘ä╪╣╪º┘à╪⌐ ╪»┘ê╪▒ ┘ü┘è ╪╢┘à╪º┘å ┘ê╪╡┘ê┘ä ┘ü┘ê╪º╪ª╪» ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪Ñ┘ä┘ë ╪º┘ä╪╖┘ä╪º╪¿ ╪º┘ä╪ú┘â╪½╪▒ ╪º╪│╪¬┘ü╪º╪»╪⌐. ┘é╪» ┘è╪¬╪╖┘ä╪¿ ┘ç╪░╪º ╪º╪│╪¬╪½┘à╪º╪▒┘ï╪º ╪╣╪º┘à┘ï╪º╪î ┘ê╪ú╪»┘ê╪º╪¬ ┘à┘ü╪¬┘ê╪¡╪⌐ ╪º┘ä┘à╪╡╪»╪▒╪î ┘ê╪┤╪▒╪º┘â╪º╪¬ ╪¬┘ü╪╢┘ä ╪º┘ä┘ê╪╡┘ê┘ä ╪╣┘ä┘ë ╪º┘ä╪▒╪¿╪¡.</p> <p>╪º┘ä╪ú┘ê┘ä┘ê┘è╪⌐ ╪º┘ä╪½╪º┘ä╪½╪⌐ ┘ç┘è ╪»╪╣┘à ╪º┘ä┘à╪╣┘ä┘à┘è┘å. ┘è┘å╪¼╪¡ ╪º┘ä╪¬╪¿┘å┘è ╪ú┘ê ┘è┘ü╪┤┘ä ┘ü┘è ╪º┘ä┘ü╪╡┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪î ┘ê╪º┘ä┘ü╪╡┘ê┘ä ╪º┘ä╪»╪▒╪º╪│┘è╪⌐ ┘è╪»┘è╪▒┘ç╪º ╪º┘ä┘à╪╣┘ä┘à┘ê┘å. ┘è╪¼╪¿ ╪ú┘å ╪¬╪¬╪╢┘à┘å ╪º┘ä╪│┘è╪º╪│╪º╪¬ ╪¬┘à┘ê┘è┘ä┘ï╪º ┘ä┘ä╪¬╪»╪▒┘è╪¿╪î ┘ê┘ê┘é╪¬┘ï╪º ┘ä┘ä╪¬╪¼╪▒┘è╪¿╪î ┘ê╪º╪¡╪¬╪▒╪º┘à┘ï╪º ┘ä┘ä╪¡┘â┘à ╪º┘ä┘à┘ç┘å┘è. ╪º┘ä┘à╪╣┘ä┘æ┘à ╪º┘ä╪░┘è ┘è┘Å╪╣╪╖┘ë ╪ú╪»╪º╪⌐ ╪½┘à ┘è┘Å╪¬╪▒┘â ┘ê╪¡╪»┘ç ┘ç┘ê ┘à╪╣┘ä┘æ┘à ╪│┘è╪▒┘ü╪╢┘ç╪º. ╪º┘ä┘à╪╣┘ä┘æ┘à ╪º┘ä╪░┘è ┘è┘Å╪»╪╣┘à ┘ê┘è┘Å╪»╪▒┘Ä┘æ╪¿ ┘ê┘è┘ê╪½┘é ╪¿┘ç ┘ç┘ê ┘à╪╣┘ä┘æ┘à ┘é╪» ┘è╪¡┘ê┘æ┘ä┘ç╪º.</p> <p>╪º┘ä╪ú┘ê┘ä┘ê┘è╪⌐ ╪º┘ä╪▒╪º╪¿╪╣╪⌐ ┘ç┘è ╪º┘ä╪¬┘é┘è┘è┘à ╪º┘ä┘à╪│╪¬┘é┘ä. ┘ä╪º ┘è┘å╪¿╪║┘è ┘ä┘ä╪¡┘â┘ê┘à╪⌐ ┘ê╪º┘ä┘à╪»╪º╪▒╪│ ╪º┘ä╪º╪╣╪¬┘à╪º╪» ╪╣┘ä┘ë ╪º┘ä╪¿╪º╪╣╪⌐ ┘ä╪Ñ╪½╪¿╪º╪¬ ╪ú┘å ┘à┘å╪¬╪¼╪º╪¬┘ç┘à ╪¬╪╣┘à┘ä. ┘ç┘å╪º┘â ┘à┘â╪º┘å ┘ä┘ä╪¬┘é┘è┘è┘à ╪º┘ä┘à╪│╪¬┘é┘ä ╪º┘ä╪╡╪º╪▒┘à ┘ê╪º┘ä╪┤┘ü╪º┘ü ╪º┘ä╪░┘è ┘è╪╖╪▒╪¡ ╪ú╪│╪ª┘ä╪⌐ ╪╡╪╣╪¿╪⌐ ╪╣┘å ╪º┘ä┘ü╪╣╪º┘ä┘è╪⌐ ┘ê╪º┘ä╪ú┘à╪º┘å ┘ê╪º┘ä╪Ñ┘å╪╡╪º┘ü ┘é╪¿┘ä ┘å╪┤╪▒ ┘ç╪░┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ┘ä┘à┘ä╪º┘è┘è┘å ╪º┘ä╪ú╪╖┘ü╪º┘ä.</p> <p>╪ú╪«┘è╪▒┘ï╪º╪î ┘è╪¼╪¿ ╪ú┘å ╪¬┘â┘ê┘å ╪º┘ä╪│┘è╪º╪│╪⌐ ┘à╪▒┘å╪⌐. ╪¬╪¬╪╖┘ê╪▒ ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪ú╪│╪▒╪╣ ┘à┘à╪º ┘è┘à┘â┘å ┘ä┘ä┘é┘ê╪º┘å┘è┘å ╪ú┘å ╪¬┘ê╪º┘â╪¿. ┘è╪¡╪¬╪º╪¼ ╪º┘ä┘à┘å╪╕┘à┘ê┘å ╪Ñ┘ä┘ë ╪ó┘ä┘è╪º╪¬ ╪¬╪│╪¬╪╖┘è╪╣ ╪º┘ä╪º╪│╪¬╪¼╪º╪¿╪⌐ ┘ä┘ä╪¬╪║┘è┘è╪▒ ╪»┘ê┘å ╪º┘ä╪¬╪╢╪¡┘è╪⌐ ╪¿╪º┘ä╪¡┘à╪º┘è╪º╪¬ ╪º┘ä╪¬┘è ┘è╪¬╪╖┘ä╪¿┘ç╪º ╪º┘ä╪º╪│╪¬┘é╪▒╪º╪▒. ┘ç╪░╪º ╪¬┘ê╪º╪▓┘å ╪╡╪╣╪¿╪î ┘ê╪º┘ä╪«╪╖╪ú ┘ü┘è ╪ú┘è ┘à┘å ╪º┘ä╪º╪¬╪¼╪º┘ç┘è┘å ┘à┘â┘ä┘ü ΓÇö ╪¿╪╖┘è╪í ╪¼╪»┘ï╪º╪î ┘ü┘è╪¡╪»╪½ ╪º┘ä╪╢╪▒╪▒╪¢ ┘ê╪│╪▒┘è╪╣ ╪¼╪»┘ï╪º╪î ┘ü┘è╪«╪¬┘å┘é ╪º┘ä╪º╪¿╪¬┘â╪º╪▒.</p> <p>╪º┘ä┘ç╪»┘ü ╪º┘ä┘ê╪º┘é╪╣┘è ┘ä┘è╪│ ╪¡┘ä ┘â┘ä ┘ç╪░┘ç ╪º┘ä╪¬┘ê╪¬╪▒╪º╪¬ ╪¿╪┤┘â┘ä ┘à╪½╪º┘ä┘è╪î ╪¿┘ä ╪Ñ╪»╪º╪▒╪¬┘ç╪º ╪¿╪¬╪╡┘à┘è┘à ┘ê╪┤┘ü╪º┘ü┘è╪⌐╪î ┘à╪╣ ╪¿┘é╪º╪í ┘à╪╡╪º┘ä╪¡ ╪º┘ä╪╖┘ä╪º╪¿ ┘ü┘è ╪º┘ä┘à╪▒┘â╪▓.</p> <h2> 10. ┘à╪│╪¬┘é╪¿┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à </h2> <p>╪º┘ä╪¬┘å╪¿╪ñ ╪¿╪º┘ä┘à╪│╪¬┘é╪¿┘ä ┘à┘ç┘à╪⌐ ╪º┘ä╪¡┘à┘é┘ë. ┘ê┘à╪╣ ╪░┘ä┘â╪î ┘è┘à┘â┘å┘å╪º ╪¬┘à┘è┘è╪▓ ╪º╪¬╪¼╪º┘ç╪º╪¬ ╪¬╪│╪¬╪¡┘é ╪ú┘å ╪¬┘Å╪ñ╪«╪░ ╪╣┘ä┘ë ┘à╪¡┘à┘ä ╪º┘ä╪¼╪».</p> <p>╪º┘ä╪º╪¬╪¼╪º┘ç ╪º┘ä╪ú┘ê┘ä ┘ç┘ê ╪º┘ä┘é╪»╪▒╪⌐. ╪│╪¬╪│╪¬┘à╪▒ ╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪ú┘å ╪¬╪╡╪¿╪¡ ╪ú┘é┘ê┘ë ┘ê╪ú┘â╪½╪▒ ╪╖┘ä╪º┘é╪⌐ ┘ê╪ú┘â╪½╪▒ ╪¬┘å┘ê╪╣┘ï╪º. ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘è ╪¬╪¿╪»┘ê ┘à╪½┘è╪▒╪⌐ ┘ä┘ä╪Ñ╪╣╪¼╪º╪¿ ╪º┘ä┘è┘ê┘à ╪│╪¬╪¿╪»┘ê ╪¿╪»╪º╪ª┘è╪⌐ ╪¿╪╣╪» ╪╣┘é╪». ┘ç╪░╪º ╪┤╪¿┘ç ┘à╪ñ┘â╪».</p> <p>╪º┘ä╪º╪¬╪¼╪º┘ç ╪º┘ä╪½╪º┘å┘è ┘ç┘ê ╪º┘ä╪¬┘â╪º┘à┘ä. ╪│┘è╪¬┘ê┘é┘ü ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪╣┘å ┘â┘ê┘å┘ç ┘à┘è╪▓╪⌐ ┘à┘à┘è╪▓╪⌐ ┘ê┘è╪╡╪¿╪¡ ╪¼╪▓╪í┘ï╪º ╪║┘è╪▒ ┘à╪▒╪ª┘è ┘à┘å ╪º┘ä╪¿┘è╪ª╪⌐ ╪º┘ä╪¬╪╣┘ä┘è┘à┘è╪⌐. ╪│┘è╪│┘â┘å ┘ü┘è ╪º┘ä┘à╪¡╪¬┘ê┘ë ╪º┘ä╪░┘è ┘è┘é╪▒╪ñ┘ç ╪º┘ä╪╖┘ä╪º╪¿╪î ┘ê╪º┘ä╪¬┘é┘è┘è┘à╪º╪¬ ╪º┘ä╪¬┘è ┘è╪«╪╢╪╣┘ê┘å ┘ä┘ç╪º╪î ┘ê╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ╪º┘ä╪¬┘è ┘è╪¬┘ä┘é┘ê┘å┘ç╪º╪î ┘ê╪¡╪¬┘ë ┘ü┘è ╪╖╪▒┘è┘é╪⌐ ╪¼╪»┘ê┘ä╪⌐ ╪º┘ä┘ü╪╡┘ê┘ä ┘ê╪Ñ╪»╪º╪▒╪¬┘ç╪º. ┘ä┘å ┘è┘â┘ê┘å ╪º┘ä╪│╪ñ╪º┘ä ╪¿╪╣╪» ╪º┘ä╪ó┘å ┘à╪º ╪Ñ╪░╪º ┘â╪º┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪¡╪º╪╢╪▒┘ï╪º╪î ╪¿┘ä ┘à╪»┘ë ╪¼┘ê╪»╪⌐ ╪º╪│╪¬╪«╪»╪º┘à┘ç.</p> <p>╪º┘ä╪º╪¬╪¼╪º┘ç ╪º┘ä╪½╪º┘ä╪½ ┘ç┘ê ╪¬╪«╪╡┘è╪╡ ╪ú╪╣┘à┘é. ┘é╪» ╪¬╪┤┘â┘ä ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä┘à╪│╪¬┘é╪¿┘ä┘è╪⌐ ┘å┘à┘ê╪░╪¼┘ï╪º ┘ä┘è╪│ ┘ü┘é╪╖ ┘ä┘à╪º ┘è╪╣╪▒┘ü┘ç ╪º┘ä╪╖╪º┘ä╪¿╪î ╪¿┘ä ┘â┘è┘ü ┘è╪¬╪╣┘ä┘à ╪ú┘ü╪╢┘ä ΓÇö ╪ú╪│┘ä┘ê╪¿┘ç ╪º┘ä┘à╪╣╪▒┘ü┘è╪î ┘ê┘à┘ä┘ü┘ç ╪º┘ä╪¬╪¡┘ü┘è╪▓┘è╪î ┘ê╪ú┘å┘à╪º╪╖ ╪º┘å╪¬╪¿╪º┘ç┘ç. ┘ç╪░╪º ┘é╪» ┘è┘é╪»┘à ╪»╪╣┘à┘ï╪º ╪º╪│╪¬╪½┘å╪º╪ª┘è┘ï╪º ┘ä┘ä┘à╪¬╪╣┘ä┘à┘è┘å ╪º┘ä╪ú┘ü╪▒╪º╪». ┘ê┘é╪» ┘è╪¼┘ä╪¿ ┘à╪╣┘ç ╪ú┘è╪╢┘ï╪º ┘à╪«╪º╪╖╪▒ ╪¼╪»┘è╪»╪⌐ ┘ê╪«╪╖┘è╪▒╪⌐ ╪╣┘ä┘ë ╪º┘ä╪«╪╡┘ê╪╡┘è╪⌐ ┘ê╪º┘ä╪º╪│╪¬┘é┘ä╪º┘ä┘è╪⌐╪î ╪Ñ╪░ ╪¬╪╡╪¿╪¡ ╪º┘ä┘à┘ä╪º┘à╪¡ ╪º┘ä╪»╪º╪«┘ä┘è╪⌐ ┘ä┘ä╪╣┘é┘ä ╪¿┘è╪º┘å╪º╪¬.</p> <p>┘ç┘å╪º┘â ╪ú┘è╪╢┘ï╪º ╪Ñ┘à┘â╪º┘å┘è╪⌐ ╪Ñ╪╣╪º╪»╪⌐ ╪¬╪╣╪▒┘è┘ü ╪º┘ä┘ç╪»┘ü ╪º┘ä╪¬╪╣┘ä┘è┘à┘è ┘å┘ü╪│┘ç.</p> <p>╪Ñ╪░╪º ┘â╪º┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è╪│╪¬╪╖┘è╪╣ ╪ú╪»╪º╪í ╪º┘ä┘â╪½┘è╪▒ ┘à┘å ╪º┘ä┘à┘ç╪º┘à ╪º┘ä┘à╪╣╪▒┘ü┘è╪⌐ ╪º┘ä╪¬┘è ┘å╪»╪▒┘æ╪│┘ç╪º ╪¡╪º┘ä┘è┘ï╪º ΓÇö ╪º┘ä╪¡╪│╪º╪¿╪î ┘ê╪º┘ä┘é┘ê╪º╪╣╪»╪î ┘ê╪º┘ä╪º╪│╪¬╪▒╪¼╪º╪╣╪î ┘ê╪¡╪¬┘ë ╪¿╪╣╪╢ ╪ú╪┤┘â╪º┘ä ╪º┘ä╪¬╪¡┘ä┘è┘ä ΓÇö ┘ü╪¬┘å╪┤╪ú ╪¡┘è┘å┘ç╪º ╪º┘ä╪ú╪│╪ª┘ä╪⌐: ┘à╪º╪░╪º ┘è╪¼╪¿ ╪ú┘å ╪¬╪»╪▒┘æ╪│ ╪º┘ä┘à╪»╪º╪▒╪│╪ƒ ┘é╪» ┘è╪¬╪¡┘ê┘ä ╪º┘ä╪¼┘ê╪º╪¿ ┘å╪¡┘ê ┘à╪º ┘ç┘ê ╪¿╪┤╪▒┘è ╪¿╪╡┘ê╪▒╪⌐ ┘à┘à┘è╪▓╪⌐: ╪º┘ä╪¬┘ü┘â┘è╪▒ ╪º┘ä┘å┘é╪»┘è╪î ┘ê╪º┘ä╪Ñ╪¿╪»╪º╪╣╪î ┘ê╪º┘ä╪ú╪«┘ä╪º┘é╪î ┘ê╪º┘ä╪¬╪╣╪º┘ê┘å╪î ┘ê╪º┘ä┘à╪▒┘ê┘å╪⌐╪î ┘ê╪º┘ä┘é╪»╪▒╪⌐ ╪╣┘ä┘ë ╪╖╪▒╪¡ ╪ú╪│╪ª┘ä╪⌐ ╪¼┘è╪»╪⌐ ╪¿╪»┘ä┘ï╪º ┘à┘å ┘à╪¼╪▒╪» ╪º┘ä╪Ñ╪¼╪º╪¿╪⌐ ╪╣┘å┘ç╪º. ┘é╪» ┘ä╪º ┘è╪»┘à╪▒ ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ç╪░┘ç ╪º┘ä┘à┘ç╪º╪▒╪º╪¬. ┘ê┘é╪» ┘è╪▒┘ü╪╣┘ç╪º╪î ╪¿╪¼╪╣┘ä ╪º┘ä╪Ñ╪»╪▒╪º┘â ╪º┘ä╪▒┘ê╪¬┘è┘å┘è ╪▒╪«┘è╪╡┘ï╪º╪î ╪Ñ┘ä┘ë ┘à╪▒┘â╪▓ ╪º┘ä╪¬╪╣┘ä┘è┘à.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ╪Ñ╪»╪▒╪º┘â ╪▒┘ê╪¬┘è┘å┘è ┘à╪º ╪¬╪»╪▒┘æ╪│┘ç ╪º┘ä┘à╪»╪º╪▒╪│ _____________________ ______________________ ╪¡╪│╪º╪¿ XXXX ╪¬┘ü┘â┘è╪▒ ┘å┘é╪»┘è ┘é┘ê╪º╪╣╪» XXXX ╪Ñ╪¿╪»╪º╪╣ ╪º╪│╪¬╪▒╪¼╪º╪╣ XXXX ╪ú╪«┘ä╪º┘é ┘ê╪¬╪╣╪º┘ê┘å ╪¬╪¡┘ä┘è┘ä ╪ú╪│╪º╪│┘è XXXX ┘à╪▒┘ê┘å╪⌐ ┘ê╪ú╪│╪ª┘ä╪⌐ ╪¼┘è╪»╪⌐ </code></pre> </div> <p>┘ê┘à╪╣ ╪░┘ä┘â. ╪º┘ä┘à╪│╪¬┘é╪¿┘ä ╪║┘è╪▒ ┘à┘â╪¬┘ê╪¿. ┘è┘Å╪╡┘å╪╣ ╪¿╪º┘ä╪«┘è╪º╪▒╪º╪¬.</p> <p>╪º┘ä┘é╪▒╪º╪▒╪º╪¬ ╪º┘ä╪¬┘è ┘è╪¬╪«╪░┘ç╪º ╪º┘ä╪¼┘è┘ä ╪º┘ä╪¡╪º┘ä┘è ┘à┘å ╪º┘ä┘à╪╣┘ä┘à┘è┘å ┘ê╪º┘ä╪¬┘é┘å┘è┘è┘å ┘ê╪╡╪º┘å╪╣┘è ╪º┘ä╪│┘è╪º╪│╪º╪¬ ╪│╪¬╪¡╪»╪» ┘à╪º ╪Ñ╪░╪º ┘â╪º┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è╪╡╪¿╪¡ ┘é┘ê╪⌐ ┘ä╪¬╪╣┘à┘è┘é ╪º┘ä┘ü╪▒╪╡╪⌐ ╪ú┘à ┘ä╪¬╪╡┘ä┘æ╪¿ ╪╣╪»┘à ╪º┘ä┘à╪│╪º┘ê╪º╪⌐. ┘â┘ä╪º ╪º┘ä┘à╪│╪¬┘é╪¿┘ä┘è┘å ┘à╪¬╪º╪¡. ┘â┘ä┘è┘ç┘à╪º ╪¬╪»╪╣┘à┘ç ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ┘å┘ü╪│┘ç╪º. ┘à╪º ┘è┘ü╪╡┘ä┘ç┘à╪º ┘ç┘ê ╪º┘ä╪¡┘â┘à╪⌐ ΓÇö ╪¡┘â┘à╪⌐ ╪º┘ä╪¿┘å╪º╪í ╪¿╪¡╪░╪▒╪î ┘ê╪º┘ä╪¬┘å╪╕┘è┘à ╪¿╪╡┘ê╪▒╪⌐ ╪¡┘â┘è┘à╪⌐╪î ┘ê╪º┘ä╪¬╪▒┘é╪¿ ┘ä┘ä╪╢╪▒╪▒╪î ┘ê╪Ñ╪¿┘é╪º╪í ╪º┘ä╪Ñ┘å╪│╪º┘å╪î ┘ä╪º ╪º┘ä╪ó┘ä╪⌐╪î ┘ü┘è ┘à╪▒┘â╪▓ ╪º┘ä╪¬╪╣┘ä┘à.</p> <h2> 11. ╪º┘ä╪«╪º╪¬┘à╪⌐ </h2> <p>┘ê╪╡┘ä ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪Ñ┘ä┘ë ╪º┘ä╪¬╪╣┘ä┘è┘à. ╪Ñ┘å┘ç ┘ä┘è╪│ ┘é╪º╪»┘à┘ï╪º╪¢ ╪Ñ┘å┘ç ┘ç┘å╪º.</p> <p>╪¼╪º╪»┘ä╪¬ ┘ç╪░┘ç ╪º┘ä╪▒╪│╪º┘ä╪⌐ ╪¿╪ú┘å ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪¡┘é┘è┘é┘è╪⌐╪î ┘ê╪ú┘å ┘ê╪╣╪»┘ç╪º ╪ú╪╡┘è┘ä╪î ┘ê╪ú┘å ┘à╪«╪º╪╖╪▒┘ç╪º ╪¼╪│┘è┘à╪⌐. ┘ê╪¼╪º╪»┘ä╪¬ ╪¿╪ú┘å ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘è┘à┘â┘å┘ç ╪¬╪¡╪│┘è┘å ╪º┘ä╪¬╪╣┘ä┘à╪î ┘ä┘â┘å ┘ü┘é╪╖ ┘ü┘è ╪º┘ä╪╕╪▒┘ê┘ü ╪º┘ä╪╡╪¡┘è╪¡╪⌐. ┘ê╪¼╪º╪»┘ä╪¬ ╪¿╪ú┘å ╪º┘ä┘à╪╣┘ä┘à┘è┘å ┘ä┘å ┘è┘Å╪│╪¬╪¿╪»┘ä┘ê╪º╪î ┘ä┘â┘å ╪╣┘à┘ä┘ç┘à ╪│┘è╪¬╪║┘è╪▒. ┘ê╪¼╪º╪»┘ä╪¬ ╪¿╪ú┘å ╪ú╪╣┘à┘é ╪º┘ä╪ú╪│╪ª┘ä╪⌐ ╪º┘ä╪¬┘è ┘è╪½┘è╪▒┘ç╪º ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ┘ä┘è╪│╪¬ ╪¬┘é┘å┘è╪⌐ ╪¿┘ä ╪Ñ┘å╪│╪º┘å┘è╪⌐.</p> <p>┘å╪¡┘å╪î ╪¿┘à╪╣┘å┘ë ┘à╪º╪î ┘ê╪º┘é┘ü┘ê┘å ╪╣┘å╪» ┘à┘ü╪¬╪▒┘é ╪╖╪▒┘é.</p> <p>┘à╪│╪º╪▒ ┘è┘é┘ê╪» ╪Ñ┘ä┘ë ┘à╪│╪¬┘é╪¿┘ä ╪¬┘à╪» ┘ü┘è┘ç ╪º┘ä╪ú┘å╪╕┘à╪⌐ ╪º┘ä╪¬┘â┘è┘ü┘è╪⌐ ┘â┘ä ╪╣┘é┘ä ┘é╪º╪»╪▒╪î ┘ê╪¬┘â┘ê┘å ┘ü┘è┘ç ╪º┘ä╪¬╪║╪░┘è╪⌐ ╪º┘ä╪▒╪º╪¼╪╣╪⌐ ┘ü┘ê╪▒┘è╪⌐╪î ┘ê┘ä╪º ┘è╪│┘é╪╖ ┘ü┘è┘ç ╪╖╪º┘ä╪¿ ╪¿╪╡┘à╪¬ ╪╣╪¿╪▒ ╪º┘ä╪┤┘é┘ê┘é ┘ä╪ú┘å ╪ó┘ä╪⌐ ┘ä╪º╪¡╪╕╪¬ ╪ú┘ê┘ä┘ï╪º. ┘ç╪░┘ç ╪▒╪ñ┘è╪⌐ ╪¼┘à┘è┘ä╪⌐.</p> <p>╪º┘ä┘à╪│╪º╪▒ ╪º┘ä╪ó╪«╪▒ ┘è┘é┘ê╪» ╪Ñ┘ä┘ë ┘à╪│╪¬┘é╪¿┘ä ╪¬┘â┘ê┘å ┘ü┘è┘ç ╪ú┘ü╪╢┘ä ╪º┘ä╪ú╪»┘ê╪º╪¬ ┘ä┘ä┘é┘ä╪⌐ ┘ü┘é╪╖╪î ┘ê┘è┘Å╪¿┘å┘ë ┘ü┘è┘ç ┘à┘ä┘ü ╪¬╪╣╪▒┘è┘ü ┘ä┘ä╪ú╪╖┘ü╪º┘ä ┘ê┘è┘Å┘ü╪▒╪▓┘ê┘å ╪¿┘ê╪º╪│╪╖╪⌐ ╪«┘ê╪º╪▒╪▓┘à┘è╪º╪¬ ╪║╪º┘à╪╢╪⌐╪î ┘ê┘è┘Å╪¡╪│┘Ä┘æ┘å ┘ü┘è┘ç ╪º┘ä╪¬╪╣┘ä┘à ╪¡╪¬┘ë ┘è┘ü┘é╪» ╪▒┘ê╪¡┘ç. ┘ç╪░┘ç ╪▒╪ñ┘è╪⌐ ╪¬╪¡╪░┘è╪▒┘è╪⌐.</p> <p>╪º┘ä╪╖╪▒┘è┘é ┘ä╪º ╪¬╪«╪¬╪º╪▒┘ç ╪º┘ä╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º. ┘å╪¡┘å ┘à┘å ┘å╪«╪¬╪º╪▒┘ç.</p> <p>┘ê╪╣╪» ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ┘ü┘è ╪º┘ä╪¬╪╣┘ä┘è┘à ┘ç┘ê ╪ú┘å┘ç ╪▒╪¿┘à╪º ┘è┘â╪│╪▒ ╪ú╪«┘è╪▒┘ï╪º ╪º┘ä╪»┘ê╪▒╪⌐ ╪º┘ä╪¬╪º╪▒┘è╪«┘è╪⌐ ┘ä╪«┘è╪¿╪⌐ ╪º┘ä╪ú┘à┘ä ╪º┘ä╪¬┘è ╪¬╪¿╪╣╪½ ┘â┘ä ╪¬┘â┘å┘ê┘ä┘ê╪¼┘è╪º ╪¬╪╣┘ä┘è┘à┘è╪⌐ ┘é╪¿┘ä┘ç ΓÇö ╪ú┘å┘ç ╪▒╪¿┘à╪º ┘è┘ü┘è ╪¿╪¡┘ä┘à ╪¬╪╣┘ä┘à ┘è┘å╪¡┘å┘è ╪¡┘ê┘ä ╪º┘ä┘à╪¬╪╣┘ä┘à. ╪º┘ä╪«╪╖╪▒ ╪ú┘å ┘è╪╡╪¿╪¡ ┘à╪»╪«┘ä┘ï╪º ╪ó╪«╪▒ ┘ü┘è ╪¬┘ä┘â ╪º┘ä╪»┘ê╪▒╪⌐╪î ╪┤┘è╪ª┘ï╪º ┘ä╪º┘à╪╣┘ï╪º ╪ó╪«╪▒ ┘ê╪╣╪» ┘â╪½┘è╪▒┘ï╪º ┘ê┘é╪»┘æ┘à ┘é┘ä┘è┘ä┘ï╪º╪î ╪ú┘ê ╪ú╪│┘ê╪ú╪î ╪ú┘å ┘è┘ü┘è ╪¿┘ê╪╣╪»┘ç ╪¿╪┤┘â┘ä ╪º┘å╪¬┘é╪º╪ª┘è ┘ä┘é┘ä╪⌐ ┘à╪¡╪╕┘ê╪╕╪⌐.</p> <p>┘ä╪º ╪┤┘è╪í ┘à┘å ╪º┘ä┘å╪¬┘è╪¼╪¬┘è┘å ╪¡╪¬┘à┘è. ┘â┘ä╪º┘ç┘à╪º ┘à┘à┘â┘å. ┘à╪│╪¬┘é╪¿┘ä ╪º┘ä╪¬╪╣┘ä┘à ΓÇö ╪│┘ê╪º╪í ╪¼╪╣┘ä┘ç ╪º┘ä╪░┘â╪º╪í ╪º┘ä╪º╪╡╪╖┘å╪º╪╣┘è ╪ú┘ê╪│╪╣ ╪ú┘ê ╪ú╪╢┘è┘é╪î ╪ú╪║┘å┘ë ╪ú┘ê ╪ú╪¼┘ê┘ü╪î ╪ú╪╣╪»┘ä ╪ú┘ê ╪ú┘é┘ä ΓÇö ┘è┘é╪╣ ┘ü┘è ╪º┘ä╪«┘è╪º╪▒╪º╪¬ ╪º┘ä╪¬┘è ┘å╪¬╪«╪░┘ç╪º ╪º┘ä╪ó┘å.</p> <p>╪│┘è┘â┘ê┘å ┘à┘å ╪º┘ä╪«┘è╪º┘å╪⌐ ╪º┘ä╪║╪▒┘è╪¿╪⌐ ┘ä╪Ñ┘å╪│╪º┘å┘è╪¬┘å╪º ╪ú┘å ┘å╪¬╪«╪░ ┘ç╪░┘ç ╪º┘ä╪«┘è╪º╪▒╪º╪¬ ╪¿┘ä╪º ┘à╪¿╪º┘ä╪º╪⌐. ┘ê╪│╪¬┘â┘ê┘å ╪«┘è╪º┘å╪⌐ ┘à┘à╪º╪½┘ä╪⌐ ╪ú┘å ┘å╪▒┘ü╪╢ ╪º╪¬╪«╪º╪░┘ç╪º ╪Ñ╪╖┘ä╪º┘é┘ï╪º.</p> <h2> 12. ╪º┘ä┘à╪▒╪º╪¼╪╣ </h2> <p>Baker, R. S. (2016). Stupid tutoring systems, intelligent humans. <em>International Journal of Artificial Intelligence in Education</em>, 26(2), 600ΓÇô614.</p> <p>Cuban, L. (1986). <em>Teachers and Machines: The Classroom Use of Technology Since 1920</em>. Teachers College Press.</p> <p>Kulik, J. A., &amp; Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. <em>Review of Educational Research</em>, 86(1), 42ΓÇô78.</p> <p>Luckin, R., Holmes, W., Griffiths, M., &amp; Forcier, L. B. (2016). <em>Intelligence Unleashed: An Argument for AI in Education</em>. Pearson.</p> <p>Ma, W., Adesope, O. O., Nesbit, J. C., &amp; Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. <em>Journal of Educational Psychology</em>, 106(4), 901ΓÇô918.</p> <p>Selwyn, N. (2019). <em>Should Robots Replace Teachers? AI and the Future of Education</em>. Polity Press.</p> <p>Skinner, B. F. (1958). Teaching machines. <em>Science</em>, 128(3330), 969ΓÇô977.</p> <p>Williamson, B. (2017). <em>Big Data in Education: The Digital Future of Learning, Policy and Practice</em>. Sage.</p> <p>Zawacki-Richter, O., Mar├¡n, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. <em>International Journal of Educational Technology in Higher Education</em>, 16(1), 39.</p> ogc16 Thu, 27 Aug 2026 12:33:11 +0000 https://dev.to/ogc16/comprender-la-clase-como-la-inteligencia-artificial-reescribe-nuestra-forma-de-aprender-1b0 https://dev.to/ogc16/comprender-la-clase-como-la-inteligencia-artificial-reescribe-nuestra-forma-de-aprender-1b0 <h2> Tesis </h2> <h2> Resumen </h2> <p>La inteligencia artificial ya no es una promesa lejana. Está en la sala de clases, ahora. Esta tesis examina cómo las herramientas de IA están transformando la educación — desde los sistemas de tutoría personalizada que se adaptan a cada aprendiz, hasta la calificación inteligente que da retroalimentación en segundos, pasando por las silenciosas cuestiones éticas que acompañan a cada innovación.</p> <p>El argumento aquí no es que la IA reemplazará a los docentes. No lo hará. En cambio, esta tesis plantea una afirmación más matizada: la IA, usada con criterio, puede liberar a los docentes para hacer lo que las máquinas no pueden hacer — acompañar, inspirar y conectar. Sin embargo, esa promesa depende por completo de cómo elijamos construir y desplegar estos sistemas.</p> <p>Los capítulos que siguen trazan un camino: desde las raíces históricas de la tecnología educativa, pasando por el panorama actual de las herramientas de IA, hasta los efectos medibles en el aprendizaje, y finalmente hasta las cuestiones éticas y de política que ningún estudiante, docente o responsable de políticas puede permitirse ignorar.</p> <h2> Índice </h2> <ol> <li>Introducción</li> <li>Breve historia de la tecnología educativa</li> <li>Cómo funciona la IA en el aula</li> <li>Aprendizaje personalizado y sistemas adaptativos</li> <li>Evaluación y retroalimentación automatizadas</li> <li>El rol del docente en la era de la IA</li> <li>Medir lo que funciona: evidencia y resultados</li> <li>Ética, privacidad y sesgo</li> <li>Política e implementación</li> <li>El futuro de la IA en la educación</li> <li>Conclusión</li> <li>Referencias</li> </ol> <h2> 1. Introducción </h2> <p>Vivimos en una época de herramientas extraordinarias.</p> <p>La misma tecnología que impulsa los autos autónomos y el diagnóstico médico le entrega hoy a un estudiante un tutor que nunca se cansa, nunca juzga y recuerda cada respuesta que ha dado. Es una perspectiva asombrosa. Y ya está aquí.</p> <p>Esta tesis trata sobre lo que ocurre cuando esa realidad encuentra el aula.</p> <p>La educación siempre ha sido lenta para cambiar. Los pizarrones dieron paso a las pizarras blancas. Las pizarras blancas, a los proyectores. Los proyectores, a las tabletas. Pero debajo de cada capa de nuevo hardware, la estructura fundamental ha permanecido obstinadamente familiar: un docente, muchos estudiantes, un currículo fijo y un reloj que tic-tac lo mismo para todos. La inteligencia artificial amenaza con romper esa estructura. Ofrece la posibilidad de un aprendizaje que se dobla alrededor del aprendiz, en lugar de forzar al aprendiz a doblarse alrededor del horario.</p> <p>La pregunta de investigación central que guía este trabajo es simple de enunciar y difícil de responder: ¿puede la inteligencia artificial hacer la educación más efectiva y más equitativa, y si es así, bajo qué condiciones?</p> <p>Para responderla, primero debemos entender qué hacen realmente estos sistemas. Debemos separar el progreso genuino del marketing y la exageración. Debemos enfrentar verdades incómodas sobre los datos, la privacidad y el riesgo de que las herramientas bien intencionadas profundicen las mismas desigualdades que afirman eliminar.</p> <p>Esta tesis procede en tres movimientos.</p> <p>Primero, construimos contexto. Miramos de dónde viene la tecnología educativa y por qué las revoluciones anteriores no cumplieron sus promesas. Segundo, examinamos el presente. Exploramos las formas concretas en que la IA ya opera dentro del aula, y sopesamos la evidencia de sus efectos. Tercero, miramos hacia adelante. Enfrentamos la ética, las políticas y las decisiones que determinarán si esta revolución sirve a cada estudiante o solo a unos pocos privilegiados.</p> <p>Lo que está en juego es alto. La educación es el gran motor de la oportunidad, la fuerza que eleva a las familias a través de las generaciones. Si la IA la fortalece, ganamos algo incalculable. Si la IA la estrecha, perdemos algo que quizás nunca recuperemos. Esta tesis es un intento de comprender qué futuro estamos construyendo.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> FUTURO UNO FUTURO DOS (oportunidad) (desigualdad) | | | toda mente se estira | las mejores herramientas para pocos | nadie cae en silencio | clasificación opaca de los niños | la retroalimentación es | el aprendizaje pierde su alma | inmediata | \____________ ____________/ \/ / \ / TÚ \ / decides \ /___________\ </code></pre> </div> <h2> 2. Breve historia de la tecnología educativa </h2> <p>Para entender hacia dónde vamos, primero debemos entender de dónde venimos.</p> <p>La historia de la tecnología en la educación es una historia de ciclos. Una y otra vez, un nuevo invento llega envuelto en grandes promesas de transformación. Una y otra vez, se asienta en un papel secundario — útil, pero rara vez revolucionario.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> GRAN PROMESA | v ESPERANZA Y EUFORIA | v LA REALIDAD SE IMPONE | v PAPEL SECUNDARIO &lt;---- la máquina no gobierna el aula </code></pre> </div> <p>Consideremos la radio. Cuando las transmisiones de radio llegaron a los hogares estadounidenses en la década de 1920, los entusiastas predijeron que todos los niños del país pronto serían enseñados por un puñado de brillantes conferencistas, cuyas voces llegarían hasta las cocinas de las granjas y los departamentos de las ciudades. No ocurrió. La radio se convirtió en un complemento, no en un reemplazo.</p> <p>Consideremos la televisión. En las décadas de 1950 y 1960, la televisión educativa prometió llevar a los mejores maestros del mundo a cada sala de estar. También se desvaneció en un segundo plano, una oferta de nicho más que un fundamento de un nuevo sistema.</p> <p>Luego vino la computadora, y con ella una nueva ola de optimismo.</p> <p>La instrucción programada, un término acuñado por el psicólogo B. F. Skinner en la década de 1950, ofreció una visión tentadora: contenido dividido en pequeños pasos, cada estudiante avanzando a su propio ritmo, con retroalimentación inmediata en cada paso. Las "máquinas de enseñar" de Skinner eran mecánicas, torpes y limitadas. Pero el pensamiento detrás de ellas — que el aprendizaje podía personalizarse mediante una secuenciación cuidadosa y un refuerzo constante — sembró una semilla que crecería durante décadas.</p> <p>La llegada de la computadora personal en la década de 1980 llevó las computadoras a las escuelas a gran escala. La instrucción asistida por computadora apareció en laboratorios y aulas de todo el mundo desarrollado. Sin embargo, los resultados fueron, según la mayoría de las medidas, modestos. Muchas máquinas acumularon polvo. Muchos programas quedaron sin uso.</p> <p>Los académicos han ofrecido formas de explicar este patrón de exageración y decepción.</p> <p>Larry Cuban, historiador de la educación en Stanford, es quizás la voz más prominente. Su trabajo documenta una verdad recurrente: las escuelas son notablemente resistentes al cambio fundamental, absorbiendo nuevas tecnologías sin ser transformadas por ellas. El análisis de Cuban sugiere que la innovación tecnológica rara vez da forma a la escolarización; en cambio, la escolarización da forma a la tecnología, doblándola para que encaje en las rutinas, los horarios y las estructuras de poder existentes.</p> <p>En este paisaje de expectativas moderadas llegó la inteligencia artificial.</p> <p>La IA es diferente de sus predecesoras en un aspecto crucial. Las tecnologías anteriores eran esencialmente sistemas de distribución — transportaban contenido de una fuente a un aprendiz. La IA, en cambio, puede responder. Puede adaptarse. Puede conversar. Puede generar. Estas no son diferencias triviales. Un libro de texto no puede ver que estás luchando y cambiar su explicación. Un sistema inteligente, sí. Esa única capacidad es la razón por la que muchos observadores creen que la IA romperá el ciclo histórico que la radio, la televisión y la computadora personal no pudieron.</p> <p>Si tienen razón es la pregunta que esta tesis existe para examinar.</p> <h2> 3. Cómo funciona la IA en el aula </h2> <p>Es tentador pensar en la IA de un aula como una cosa única y monolítica. No lo es.</p> <p>Un aula de hoy tiene más probabilidades de contener muchos sistemas de IA diferentes, cada uno haciendo un trabajo distinto. Algunos son visibles. Otros no.</p> <p>Consideremos el sistema de gestión del aprendizaje que lleva el registro de asistencia y calificaciones. Debajo de su superficie, los algoritmos predicen en silencio qué estudiantes están en riesgo de abandonar, marcando sus nombres para una intervención antes de que el problema se vuelva obvio. Esto es IA, aunque pocos lo llamarían así.</p> <p>Consideremos el asistente de escritura que subraya los errores gramaticales de un estudiante y sugiere mejoras. Usa redes neuronales entrenadas con millones de ejemplos de escritura humana. Es IA, integrada de manera tan perfecta que parece ordinaria.</p> <p>Consideremos la plataforma de tutoría que presenta un problema, espera una respuesta y luego — basándose en esa única respuesta — decide qué mostrar a continuación. Esta es la IA en su forma educativa más directa: un sistema tomando decisiones pedagógicas en tiempo real.</p> <p>Hay un punto técnico importante que vale la pena enunciar con claridad.</p> <p>La ola actual de IA en la educación se basa en gran parte en el aprendizaje automático y, en particular, en el aprendizaje profundo. Estos sistemas no siguen reglas escritas a mano. Aprenden patrones a partir de vastas cantidades de datos. Un modelo de lenguaje que genera retroalimentación para los ensayos de los estudiantes no ha aprendido gramática como un conjunto de principios; ha absorbido la gramática leyendo miles de millones de oraciones e interiorizando sus regularidades estadísticas.</p> <p>Esto tiene dos consecuencias profundas.</p> <p>Primero, estos sistemas son notablemente capaces. Pueden producir prosa fluida, resolver problemas matemáticos, responder preguntas en una asombrosa variedad de temas y entablar diálogos prolongados que a menudo se sienten naturales.</p> <p>Segundo, estos sistemas son extrañamente desconocidos. Debido a que aprenden en lugar de ser programados, su comportamiento no siempre es predecible o interpretable. Un sistema puede producir una explicación brillante en un momento y un error confiado al siguiente. No tiene verdadera comprensión en el sentido humano; tiene patrones, y los patrones son poderosos pero no son lo mismo que la comprensión.</p> <p>En el aula, estas fortalezas y debilidades se traducen en realidades concretas.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> LA IA EN EL AULA ________________________________________________________ | | | Sistema de gestión Asistente de escritura | | predice el abandono corrige la gramática | | | | | | Plataforma de tutoría Motor de evaluación | | adapta cada problema califica ensayos | | | | TODOS USAN: DATOS --&gt; PATRONES --&gt; PREDICCIÓN | |________________________________________________________| </code></pre> </div> <p>Las fortalezas permiten la práctica personalizada, la retroalimentación instantánea y un grado de atención individual que ningún docente humano podría brindar simultáneamente a treinta estudiantes. Las debilidades crean riesgos: la posibilidad de reforzar errores, la opacidad de la toma de decisiones y el peligro siempre presente de que un sistema se equivoque con toda la confianza injustificada de una máquina que ha visto la respuesta correcta una vez y ahora cree que siempre es así.</p> <p>Un relato honesto de la IA en la educación debe sostener ambas verdades a la vez. La tecnología es genuinamente prometedora. Y es genuinamente imperfecta.</p> <h2> 4. Aprendizaje personalizado y sistemas adaptativos </h2> <p>Pocas frases en la tecnología educativa se invocan con más reverencia y menos precisión que "aprendizaje personalizado".</p> <p>La idea central, sin embargo, es simple y atractiva. Cada estudiante aprende de manera diferente. Cada estudiante llega con conocimientos previos diferentes, fortalezas diferentes, debilidades diferentes y ritmos de progreso diferentes. Un sistema que pueda percibir estas diferencias y responder a ellas individualmente tiene el potencial de servir a cada estudiante mucho mejor que un currículo único.</p> <p>Los sistemas de aprendizaje adaptativo encarnan esta idea en software.</p> <p>Estos sistemas suelen funcionar en un ciclo. Presentan un problema; observan la respuesta del estudiante; actualizan un modelo interno del conocimiento de ese estudiante; y eligen el siguiente problema según ese modelo. Si un estudiante tiene dificultades con las fracciones, el sistema le da más práctica con fracciones antes de avanzar. Si el estudiante las domina, el sistema avanza. El bucle de retroalimentación es el corazón del diseño.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> PRESENTA OBSERVA ACTUALIZA un problema la respuesta el modelo del estudiante \ / | \ / | \________/______ | | | ELIGE EL SIGUIENTE PROBLEMA &lt;-------+ </code></pre> </div> <p>Los resultados pueden ser sorprendentes en contextos específicos.</p> <p>En matemáticas, los sistemas adaptativos han mostrado algunas de las evidencias más sólidas. Los estudios de sistemas utilizados para la práctica aritmética y la instrucción algebraica han reportado ganancias significativas en el rendimiento de los estudiantes, particularmente para los que comienzan rezagados respecto de sus pares. El mecanismo tiene sentido intuitivo: un estudiante que de otro modo se quedaría atrás y abandonaría — o que se aburriría sentado mientras sus compañeros alcanzan — recibe en cambio un trabajo calibrado precisamente a su nivel. Se preserva la experiencia de flujo, del trabajo que no es ni demasiado fácil ni demasiado difícil.</p> <p>Sin embargo, la evidencia no es uniformemente positiva, y las razones son instructivas.</p> <p>Primero, los sistemas adaptativos solo valen lo que vale su modelo subyacente del conocimiento. Un estudiante puede resolver un problema correctamente por razones equivocadas, o incorrectamente por las correctas. El sistema solo ve resultados, no el razonamiento detrás de ellos. Cuando el objetivo es una habilidad procedimental, esto puede importar poco. Cuando el objetivo es la comprensión conceptual, importa enormemente.</p> <p>Segundo, la personalización del ritmo no es lo mismo que la personalización del significado. Un sistema puede ajustar la dificultad de los problemas, pero no puede captar fácilmente por qué un estudiante ha perdido la motivación, ni reconocer un momento de lucha personal, ni ofrecer el aliento que solo un humano atento puede dar. Estas dimensiones del aprendizaje importan, y las máquinas siguen siendo en gran medida ciegas a ellas.</p> <p>Tercero, existe el riesgo de que la personalización estreche en lugar de ampliar la ambición. Un sistema que optimiza el dominio de un currículo fijo puede mantener a un estudiante con dificultades practicando eternamente las mismas habilidades procedimentales, sin exponerlo nunca a los aspectos más ricos y creativos de la materia que podrían encender su interés. La personalización, mal hecha, puede volverse una jaula.</p> <p>La literatura sobre el aprendizaje adaptativo, por tanto, llega a una conclusión equilibrada. La tecnología es real y, en las condiciones adecuadas, produce beneficios medibles. Pero es una herramienta entre otras, no un reemplazo de una pedagogía reflexiva. Su valor depende, como siempre, de la habilidad con que se use.</p> <h2> 5. Evaluación y retroalimentación automatizadas </h2> <p>Calificar está entre las tareas que más tiempo consumen y menos se aprecian en la enseñanza.</p> <p>Un docente con cien estudiantes puede pasar una tarde revisando cien conjuntos de trabajos escritos. Las exigencias de la calificación compiten directamente con las exigencias de la planificación, de la preparación y de las interacciones humanas que dan sentido a la enseñanza. No es de extrañar que la automatización haya llegado aquí con fuerza.</p> <p>La evaluación automatizada viene en varias formas.</p> <p>La más simple implica preguntas de opción múltiple y de respuesta corta, donde las computadoras han calificado con precisión durante décadas. Estos sistemas son confiables, rápidos e irreprochables en muchos contextos, aunque prueban una franja estrecha del aprendizaje y recompensan el reconocimiento sobre el razonamiento.</p> <p>Más ambiciosos son los sistemas diseñados para evaluar escritos extensos.</p> <p>Estos sistemas usan el procesamiento del lenguaje natural para evaluar ensayos en dimensiones como la claridad, la organización, la gramática y la adherencia a una consigna. En pruebas estandarizadas, se ha demostrado que estos sistemas producen puntajes que se correlacionan razonablemente con los evaluadores humanos. Para la retroalimentación formativa de bajo riesgo — ayudar a un estudiante a ver que un párrafo va fuera de tema, que una tesis está enterrada, que faltan transiciones — las herramientas pueden ser genuinamente útiles, ofreciendo orientación a una escala que ningún docente podría igualar.</p> <p>Pero los riesgos son sustanciales y merecen un escrutinio.</p> <p>Un sistema de escritura entrenado con ensayos pasados aprende cómo eran los ensayos pasados. Puede recompensar la estructura fórmulaica, castigar la individualidad estilística y ser burlado por estudiantes que aprenden a escribir según el patrón que el algoritmo favorece. Más preocupante, estos sistemas no pueden leer de verdad. Detectan señales estadísticas, no significado. Un ensayo bellamente argumentado que se aparta de la convención puede ser calificado con dureza; un ensayo vacío que refleja la forma esperada puede recibir una calificación generosa.</p> <p>También está la cuestión de qué hace la automatización a la retroalimentación misma.</p> <p>La retroalimentación es más poderosa cuando es inmediata, específica y alentadora. La IA puede brindar las dos primeras y, en cierta medida, la tercera. Pero la retroalimentación también es una relación. Un estudiante que recibe una calificación y algunos comentarios de una máquina ha recibido información; un estudiante que discute su trabajo con un docente ha recibido algo más. Lo primero es eficiente. Lo segundo es transformador. Un sistema educativo saludable necesita ambos y debe resistir la tentación de dejar que la eficiencia desplace la transformación.</p> <p>El camino más sensato parece ser híbrido.</p> <p>Deja que las máquinas manejen lo rutinario, lo repetible, las partes de la evaluación que desgastan a los docentes. Reserva para los humanos el juicio que requiere comprensión, el aliento que requiere empatía y las conversaciones que requieren presencia. Esta división del trabajo no es un compromiso; es un principio de diseño.</p> <h2> 6. El rol del docente en la era de la IA </h2> <p>Toda conversación sobre la IA en la educación termina llegando a la misma pregunta ansiosa: ¿las máquinas reemplazarán a los docentes?</p> <p>La respuesta corta es no. La respuesta más larga es más interesante.</p> <p>Enseñar no es simplemente la transmisión de información. Si lo fuera, entonces sí, el argumento del reemplazo sería fuerte, porque las máquinas son excelentes transmitiendo información. Pero enseñar también es el cultivo de relaciones, la modelación de la curiosidad, el reconocimiento de un estudiante que lucha en silencio, la celebración de un estudiante que finalmente comprende. Estos actos requieren una presencia humana. Ningún algoritmo puede mirar a un estudiante a los ojos y decir "creo en ti" y sentirlo de verdad.</p> <p>Sin embargo, la pregunta, reformulada, se vuelve justa.</p> <p>Los docentes no serán reemplazados. Pero el trabajo de enseñar cambiará, y ya está cambiando.</p> <p>El docente que debe pasar horas calificando podría, en cambio, pasar esas horas conversando con los estudiantes. El docente que no puede dar a cada estudiante una atención individual podría, con el apoyo de la IA, acercarse más a hacerlo. El docente que es la única fuente de conocimiento en una sala de aprendices se convierte en otra cosa cuando el conocimiento está disponible en todas partes: un navegador, un mentor, un entrenador.</p> <p>Este cambio no es del todo cómodo. Pide a los docentes ceder una medida de control sobre sus propias aulas, confiar en sistemas que no construyeron y que quizás no entiendan del todo. Exige nuevas habilidades — la capacidad de evaluar la calidad de una herramienta, de reconocer sus sesgos, de interceptar sus errores. Estas no son exigencias triviales.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> DOCENTE: ANTES DOCENTE: CON IA --------------------- ---------------------- fuente única de saber navegador / entrenador califica cada ensayo califica solo lo de alto riesgo misma lección para todos dirige el trabajo personalizado único que da atención atención centrada en la necesidad real </code></pre> </div> <p>Algunos docentes abrazarán el cambio. Otros resistirán, y su resistencia no es mera terquedad. Es una desconfianza comprensible hacia una fuerza que no eligieron y que no pueden controlar del todo. La implementación sabia de la IA en la educación reconoce esto. No impone la tecnología desde arriba esperando gratitud. Trabaja con los docentes, los forma, los escucha y los trata como socios en lugar de obstáculos.</p> <p>Hay, además, un punto más profundo sobre el rol humano.</p> <p>La fuerza más poderosa en la vida de una persona joven, después de la familia, es a menudo un docente. Un docente puede cambiar la trayectoria de una vida de maneras difíciles de medir e imposibles de algoritmizar. La llegada de la IA no disminuye este poder; si acaso, lo pone en un relieve más nítido. A medida que las máquinas asumen las tareas mecánicas de la enseñanza, el trabajo genuinamente humano — el aliento, la creencia, la relación — se vuelve más claramente el centro de la profesión, no menos.</p> <p>El docente, en otras palabras, no es la víctima de la era de la IA. El docente es su recurso más esencial.</p> <h2> 7. Medir lo que funciona: evidencia y resultados </h2> <p>Una cosa es creer que la IA ayuda a los estudiantes a aprender. Otra es demostrarlo.</p> <p>La base de evidencia para la IA en la educación es joven y es desigual. Algunas áreas se han estudiado extensamente; otras descansan en anécdotas prometedoras y afirmaciones de los vendedores. La evaluación seria exige que miremos lo que la investigación muestra realmente.</p> <p>En matemáticas y materias procedimentales, la evidencia es la más alentadora. Los meta-análisis de sistemas de tutoría inteligente han reportado efectos promedio positivos en los resultados de aprendizaje, con algunos estudios encontrando ganancias comparables a las de la tutoría humana individual. La consistencia de estos resultados en múltiples estudios es una razón genuina para el optimismo.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> FUERZA DE LA EVIDENCIA PARA LA IA EN LA EDUCACIÓN Matemática / procedimental ████████████████ fuerte Lectura / lenguas ███████████ moderada Escritura / comprensión ██████ mixta Creatividad / avanzado ███ débil </code></pre> </div> <p>En lectura y aprendizaje de idiomas, el panorama es más mixto. Las herramientas que brindan práctica de vocabulario y ejercicios de lectura adaptativos muestran beneficios. Las que intentan evaluar o mejorar la escritura y la comprensión de orden superior muestran resultados menos consistentes, reflejando la dificultad de la tarea subyacente.</p> <p>En los campos que exigen comprensión conceptual profunda o creatividad — las humanidades, las artes, la resolución avanzada de problemas — la evidencia es más escasa. Aquí, las limitaciones de la IA actual, su incapacidad para comprender de verdad, se vuelven más prominentes. Pocos estudios sugieren que la IA pueda reemplazar la instrucción humana calificada en estos dominios, y muchos sugieren que aún no puede agregar mucho más allá de lo que un buen docente ya hace.</p> <p>Igualmente importante es la pregunta de <em>a quién</em> ayuda la tecnología.</p> <p>Hay una preocupación persistente, respaldada por algunas investigaciones, de que las herramientas adaptativas ayudan a los estudiantes que menos necesitan ayuda y no logran llegar a quienes más la necesitan. Los estudiantes con autorregulación bien desarrollada, buena alfabetización y hogares solidarios pueden beneficiarse más del aprendizaje impulsado por la IA porque pueden navegarlo de manera independiente. Los estudiantes que carecen de estas ventajas pueden tener dificultades para usar las herramientas de manera eficaz, quedándose aún más atrás. La tecnología tiene una tendencia a amplificar las desigualdades existentes en lugar de eliminarlas, y la IA muestra señales preocupantes de seguir el mismo patrón.</p> <p>También está el desafío de la evidencia misma.</p> <p>Gran parte de la investigación en esta área está financiada por las empresas que venden las herramientas. Gran parte se realiza en condiciones breves y artificiales en lugar de en aulas reales durante largos períodos. El sesgo de publicación — la tendencia de las revistas a publicar resultados positivos — distorsiona aún más el panorama. Una evaluación honesta debe tratar los hallazgos prometedores con un grado de escepticismo y exigir una replicación en entornos auténticos.</p> <p>La conclusión general no es ni eufórica ni desdeñosa.</p> <p>La IA puede mejorar el aprendizaje en las condiciones adecuadas. Esas condiciones incluyen una pedagogía sólida, herramientas bien diseñadas, docentes capacitados y atención a la equidad. Cuando estos elementos están presentes, las ganancias son reales. Cuando están ausentes, la tecnología hace poco bien y puede hacer daño. La variable que determina el éxito no es la sofisticación del algoritmo. Es la calidad del entorno en el que opera el algoritmo.</p> <h2> 8. Ética, privacidad y sesgo </h2> <p>Ninguna discusión sobre la IA en la educación estaría completa sin enfrentar las cuestiones éticas que ensombrecen la tecnología. Estas no son pensamientos posteriores. Son centrales.</p> <p>La privacidad es la primera y más inmediata preocupación.</p> <p>La IA educativa está construida sobre datos. Para personalizar el aprendizaje, un sistema recopila información sobre lo que un estudiante sabe, cómo aprende, qué tan rápido progresa, dónde tiene dificultades. A veces recopila mucho más: teclas presionadas, patrones de lectura, respuestas emocionales inferidas de las expresiones faciales o las velocidades de escritura. Estos son datos íntimos, recopilados de niños, a menudo sin un consentimiento significativo.</p> <p>La escala de la recopilación es asombrosa y los propósitos no siempre son claros. Los datos recopilados para mejorar el aprendizaje pueden usarse para otros fines — clasificar escuelas, hacer marketing dirigido a estudiantes, construir perfiles que siguen a una persona mucho después de dejar el aula. Los niños perfilados hoy no están en posición de consentir cómo se usarán sus datos en veinte años. Esta asimetría de poder y consecuencia merece mucha más atención de la que ha recibido.</p> <p>Luego está el sesgo.</p> <p>Los sistemas de IA aprenden de los datos, y los datos reflejan el mundo, incluidas sus injusticias. Un sistema entrenado con ensayos de un grupo demográfico puede subvalorar sistemáticamente la escritura de otro. Un sistema adaptativo optimizado con el rendimiento de estudiantes con recursos puede juzgar mal y atender deficientemente a estudiantes de contextos diferentes. El sesgo en la IA no es un error que aparece ocasionalmente; es una característica estructural que aparece cada vez que los datos subyacentes son desiguales, y la carga de detectarlo y corregirlo no recae en los datos sino en los humanos que diseñan y despliegan el sistema.</p> <p>También hay una dimensión ética más sutil: el riesgo de la vigilancia.</p> <p>Un aula en la que cada acción es monitoreada, cada error registrado, cada pausa cronometrada, es un aula que ha cambiado de carácter, lo haya querido alguien o no. Los estudiantes bajo observación constante aprenden de manera diferente. Corren menos riesgos. Se conforman. Las mismas cualidades que la educación pretende cultivar — la curiosidad, la experimentación, el coraje de equivocarse — pueden ser suprimidas silenciosamente por un entorno que observa demasiado de cerca.</p> <p>Nada de esto significa que la IA deba rechazarse. Pero sí significa que su adopción exige un andamiaje ético deliberado: reglas claras sobre los datos, fuertes protecciones del consentimiento, supervisión independiente y transparencia genuina sobre lo que estos sistemas hacen con la información que recopilan. La ética no es una función que se añada después. Es una restricción que debe integrarse desde el principio.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> EL EQUILIBRIO QUE DEBEMOS LOGRAR | ________/ \________ / \ GANANCIA EDUCATIVA COSTO HUMANO aprendizaje personal. pérdida de privacidad retroalimentación sesgo y elaboración de perfiles inmediata cada estudiante servido aprender bajo vigilancia ___________________ ___________________ ^ ^ |______ mantener la mirada en ambos ______| </code></pre> </div> <h2> 9. Política e implementación </h2> <p>Una buena política es la diferencia entre una herramienta que sirve y una que domina.</p> <p>La trayectoria de la IA en la educación no está determinada solo por la tecnología. Está moldeada por decisiones tomadas en los legislativos, los ministerios, los consejos escolares y las oficinas de distrito. Estas decisiones determinan qué se permite, qué se financia, qué se enseña y quién está protegido.</p> <p>La primera prioridad política es la protección de los datos. Los estudiantes merecen las mismas garantías de privacidad en los espacios digitales que ya tienen en los físicos. Reglas firmes deberían regir qué datos pueden recopilarse, quién puede acceder a ellos, cuánto tiempo pueden conservarse y si pueden usarse para cualquier propósito más allá de la propia educación del estudiante. Estas reglas deberían ser exigibles y acarrear consecuencias reales en caso de violación.</p> <p>La segunda prioridad es la equidad. Dejada al mercado, las mejores herramientas de IA fluirán hacia las escuelas que puedan pagarlas, ampliando la brecha entre distritos ricos y pobres, entre países con recursos y países sin ellos. La política pública tiene un rol para asegurar que los beneficios de la IA lleguen a los estudiantes que más tienen que ganar. Esto puede exigir inversión pública, herramientas de código abierto y asociaciones que prioricen el acceso sobre el beneficio.</p> <p>La tercera prioridad es el apoyo a los docentes. La adopción triunfa o fracasa en el aula, y las aulas son dirigidas por docentes. Las políticas deben incluir financiamiento para la capacitación, tiempo para la experimentación y respeto por el criterio profesional. Un docente a quien se le da una herramienta y luego se le abandona es un docente que la rechazará. Un docente que es apoyado, capacitado y en quien se confía es un docente que puede transformarla.</p> <p>La cuarta prioridad es la evaluación independiente. El gobierno y las escuelas no deberían depender de los vendedores para probar que sus propios productos funcionan. Hay un lugar para la evaluación independiente, rigurosa y transparente que haga preguntas difíciles sobre efectividad, seguridad y equidad antes de que estos sistemas se desplieguen a millones de niños.</p> <p>Finalmente, la política debe ser ágil. La tecnología evoluciona más rápido de lo que las leyes pueden seguir. Los reguladores necesitan mecanismos que puedan responder al cambio sin sacrificar las protecciones que la estabilidad exige. Este es un equilibrio difícil, y equivocarse en cualquiera de los dos sentidos es costoso — demasiado lento, y se hace daño; demasiado rápido, y se sofoca la innovación.</p> <p>La meta realista no es resolver perfectamente todas estas tensiones, sino gestionarlas deliberadamente, de manera transparente y con los intereses de los estudiantes en el centro.</p> <h2> 10. El futuro de la IA en la educación </h2> <p>Predecir el futuro es una tarea de tontos. Aun así, podemos discernir tendencias que vale la pena tomar en serio.</p> <p>La primera tendencia es la capacidad. Los sistemas de IA seguirán volviéndose más poderosos, más fluidos y más versátiles. Los sistemas que parecen impresionantes hoy parecerán primitivos en una década. Esto es casi seguro.</p> <p>La segunda tendencia es la integración. La IA dejará de ser una característica distintiva y se convertirá en una parte invisible del entorno educativo. Vivirá en el contenido que los estudiantes leen, en las evaluaciones que rinden, en la retroalimentación que reciben, incluso en cómo se programan y gestionan las aulas. La pregunta ya no será si la IA está presente, sino qué tan bien se usa.</p> <p>La tercera tendencia es una personalización más profunda. Los sistemas futuros podrían modelar no solo lo que un estudiante sabe, sino cómo aprende mejor — su estilo cognitivo, su perfil motivacional, sus patrones de atención. Esto podría brindar un apoyo extraordinario a los aprendices individuales. También podría traer consigo riesgos nuevos y serios para la privacidad y la autonomía, a medida que los contornos internos de una mente se convierten en datos.</p> <p>También está la posibilidad de una redefinición del objetivo educativo mismo.</p> <p>Si la IA puede realizar muchas de las tareas cognitivas que actualmente enseñamos — cálculo, gramática, recuperación, incluso algunas formas de análisis — entonces surge la pregunta: ¿qué deberían enseñar las escuelas? La respuesta podría desplazarse hacia lo distintivamente humano: el pensamiento crítico, la creatividad, la ética, la colaboración, la resiliencia, la capacidad de hacer buenas preguntas en lugar de simplemente responderlas. La IA podría no destruir estas habilidades. Podría, al volver barata la cognición rutinaria, elevarlas al centro de la educación.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> COGNICIÓN RUTINARIA QUÉ ENSEÑAN LAS ESCUELAS _____________________ ______________________ cálculo XXXX pensamiento crítico gramática XXXX creatividad recuperación XXXX ética y colaboración análisis básico XXXX resiliencia y buenas preguntas </code></pre> </div> <p>Y sin embargo. El futuro no está escrito. Se hace con elecciones.</p> <p>Las decisiones que tome la generación actual de educadores, tecnólogos y responsables de políticas determinarán si la IA se convierte en una fuerza para profundizar la oportunidad o para endurecer la desigualdad. Ambos futuros están disponibles. Ambos están sostenidos por la misma tecnología. Lo que los separa es el juicio — el juicio de construir con cuidado, de regular con sabiduría, de vigilar el daño y de mantener al ser humano, no a la máquina, en el centro del aprendizaje.</p> <h2> 11. Conclusión </h2> <p>La inteligencia artificial ha llegado a la educación. No está por venir; ya está aquí.</p> <p>Esta tesis ha sostenido que la tecnología es real, que su promesa es genuina y que sus riesgos son serios. Ha sostenido que la IA puede mejorar el aprendizaje, pero solo en las condiciones adecuadas. Ha sostenido que los docentes no serán reemplazados, pero que su trabajo cambiará. Ha sostenido que las preguntas más profundas que plantea la IA en la educación no son técnicas sino humanas.</p> <p>Estamos, en cierto sentido, en una bifurcación del camino.</p> <p>Un camino lleva a un futuro en el que los sistemas adaptativos estiran cada mente capaz, en el que la retroalimentación es inmediata, en el que ningún estudiante cae silenciosamente por las grietas porque una máquina lo notó primero. Esta es una visión hermosa.</p> <p>El otro camino lleva a un futuro en el que las mejores herramientas pertenecen solo a unos pocos, en el que los niños son perfilados y clasificados por algoritmos opacos, en el que el aprendizaje se optimiza hasta perder su alma. Esta es una visión de advertencia.</p> <p>El camino no lo elige la tecnología. Lo elegimos nosotros.</p> <p>La promesa de la IA en la educación es que quizás, finalmente, rompa el ciclo histórico de decepción que ha seguido a cada tecnología educativa anterior — que quizás cumpla el sueño de un aprendizaje que se dobla alrededor del aprendiz. El riesgo es que se convierta en otra entrada de ese ciclo, otro objeto brillante que prometió mucho y entregó poco, o peor, que entregue su promesa selectivamente a unos pocos afortunados.</p> <p>Ninguno de los dos resultados es inevitable. Ambos son posibles. El futuro del aprendizaje — si la IA lo hace más amplio o más estrecho, más rico o más hueco, más justo o menos — reside en las elecciones que hacemos ahora.</p> <p>Sería una extraña traición a nuestra humanidad hacer esas elecciones con descuido. Sería una traición igual de grande negarse a hacerlas del todo.</p> <h2> 12. Referencias </h2> <p>Baker, R. S. (2016). Stupid tutoring systems, intelligent humans. <em>International Journal of Artificial Intelligence in Education</em>, 26(2), 600–614.</p> <p>Cuban, L. (1986). <em>Teachers and Machines: The Classroom Use of Technology Since 1920</em>. Teachers College Press.</p> <p>Kulik, J. A., &amp; Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. <em>Review of Educational Research</em>, 86(1), 42–78.</p> <p>Luckin, R., Holmes, W., Griffiths, M., &amp; Forcier, L. B. (2016). <em>Intelligence Unleashed: An Argument for AI in Education</em>. Pearson.</p> <p>Ma, W., Adesope, O. O., Nesbit, J. C., &amp; Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. <em>Journal of Educational Psychology</em>, 106(4), 901–918.</p> <p>Selwyn, N. (2019). <em>Should Robots Replace Teachers? AI and the Future of Education</em>. Polity Press.</p> <p>Skinner, B. F. (1958). Teaching machines. <em>Science</em>, 128(3330), 969–977.</p> <p>Williamson, B. (2017). <em>Big Data in Education: The Digital Future of Learning, Policy and Practice</em>. Sage.</p> <p>Zawacki-Richter, O., Marín, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. <em>International Journal of Educational Technology in Higher Education</em>, 16(1), 39.</p> ai education learning ogc16 Thu, 27 Aug 2026 12:29:16 +0000 https://dev.to/ogc16/-comprendre-la-classe-comment-lintelligence-artificielle-reecrit-notre-facon-dapprendre-1940 https://dev.to/ogc16/-comprendre-la-classe-comment-lintelligence-artificielle-reecrit-notre-facon-dapprendre-1940 <h2> Mémoire </h2> <h2> Résumé </h2> <p>L'intelligence artificielle n'est plus une promesse lointaine. Elle est dans la salle de classe, maintenant. Ce mémoire examine comment les outils d'IA transforment l'éducation — des systèmes de tutorat personnalisés qui s'adaptent à chaque apprenant, à la notation intelligente qui donne un retour en quelques secondes, jusqu'aux discrètes questions éthiques qui accompagnent chaque innovation.</p> <p>L'argument ici n'est pas que l'IA remplacera les enseignants. Elle ne le fera pas. Ce mémoire avance plutôt une thèse plus nuancée : l'IA, utilisée avec discernement, peut libérer les enseignants pour faire ce que les machines ne peuvent pas faire — encadrer, inspirer et créer du lien. Pourtant, cette promesse dépend entièrement de la façon dont nous choisissons de construire et de déployer ces systèmes.</p> <p>Les chapitres qui suivent tracent un chemin : des racines historiques de la technologie éducative, en passant par le paysage actuel des outils d'IA, jusqu'aux effets mesurables sur l'apprentissage, et enfin jusqu'aux questions éthiques et politiques qu'aucun élève, enseignant ou décideur ne peut se permettre d'ignorer.</p> <h2> Table des matières </h2> <ol> <li>Introduction</li> <li>Brève histoire de la technologie éducative</li> <li>Comment l'IA fonctionne en classe</li> <li>Apprentissage personnalisé et systèmes adaptatifs</li> <li>Évaluation et retour automatisés</li> <li>Le rôle de l'enseignant à l'ère de l'IA</li> <li>Mesurer ce qui fonctionne : preuves et résultats</li> <li>Éthique, vie privée et biais</li> <li>Politique et mise en œuvre</li> <li>L'avenir de l'IA dans l'éducation</li> <li>Conclusion</li> <li>Références</li> </ol> <h2> 1. Introduction </h2> <p>Nous vivons à une époque d'outils extraordinaires.</p> <p>La même technologie qui alimente les voitures autonomes et le diagnostic médical remet aujourd'hui à un élève un tuteur qui ne se lasse jamais, ne juge jamais et se souvient de chaque réponse qu'il a jamais donnée. C'est une perspective étonnante. Et elle est déjà là.</p> <p>Ce mémoire porte sur ce qui se produit quand cette réalité rencontre la salle de classe.</p> <p>L'éducation a toujours été lente à changer. Les tableaux noirs ont cédé la place aux tableaux blancs. Les tableaux blancs aux projecteurs. Les projecteurs aux tablettes. Mais sous chaque couche de nouveau matériel, la structure fondamentale est restée obstinément familière : un enseignant, de nombreux élèves, un programme fixe, et une horloge qui tictaque de la même façon pour tout le monde. L'intelligence artificielle menace de briser cette structure. Elle offre la possibilité d'un apprentissage qui se plie autour de l'apprenant, plutôt que de forcer l'apprenant à se plier autour du calendrier.</p> <p>La question de recherche centrale qui guide ce travail est simple à énoncer et difficile à répondre : l'intelligence artificielle peut-elle rendre l'éducation plus efficace et plus équitable, et si oui, à quelles conditions ?</p> <p>Pour y répondre, il faut d'abord comprendre ce que ces systèmes font réellement. Il faut distinguer le progrès authentique du battage marketing. Il faut affronter des vérités inconfortables sur les données, la vie privée et le risque que des outils bien intentionnés approfondissent les inégalités mêmes qu'ils prétendent effacer.</p> <p>Ce mémoire procède en trois mouvements.</p> <p>D'abord, nous bâtissons le contexte. Nous regardons d'où vient la technologie éducative et pourquoi les révolutions précédentes ont manqué leurs promesses. Ensuite, nous examinons le présent. Nous explorons les façons concrètes dont l'IA opère déjà dans les salles de classe, et nous pesons les preuves de ses effets. Enfin, nous regardons vers l'avant. Nous affrontons l'éthique, les politiques et les choix qui détermineront si cette révolution sert chaque élève ou seulement quelques privilégiés.</p> <p>L'enjeu est élevé. L'éducation est le grand moteur de l'opportunité, la force qui élève les familles à travers les générations. Si l'IA la renforce, nous gagnons quelque chose d'inestimable. Si l'IA la rétrécit, nous perdons quelque chose que nous ne retrouverons peut-être jamais. Ce mémoire est une tentative de comprendre quel avenir nous construisons.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> FUTUR UN FUTUR DEUX (opportunité) (inégalité) | | | tout esprit tendu | les meilleurs outils pour peu | personne ne tombe | tri opaque des enfants | retour immédiat | l'apprentissage perd son âme | | \____________ ___________/ \/ / \ / VOUS \ / décidez \ /___________\ </code></pre> </div> <h2> 2. Brève histoire de la technologie éducative </h2> <p>Pour comprendre où nous allons, il faut d'abord comprendre d'où nous venons.</p> <p>L'histoire de la technologie dans l'éducation est une histoire de cycles. À maintes reprises, une nouvelle invention arrive enveloppée de grandes promesses de transformation. À maintes reprises, elle se replie dans un rôle secondaire — utile, mais rarement révolutionnaire.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> GRANDE PROMESSE | v ESPOIR &amp; EUPHORIE | v LA RÉALITÉ S'INSTALLE | v RÔLE SECONDAIRE &lt;---- la machine ne gouverne pas la classe </code></pre> </div> <p>Considérez la radio. Quand les émissions de radio atteignirent les foyers américains dans les années 1920, des enthousiastes prédirent que chaque enfant du pays serait bientôt enseigné par une poignée de brillants conférenciers, leurs voix portées jusqu'aux cuisines des fermes et aux appartements des villes. Cela ne se produisit pas. La radio devint un complément, non un remplacement.</p> <p>Considérez la télévision. Dans les années 1950 et 1960, la télévision éducative promit d'amener les meilleurs enseignants du monde dans chaque salon. Elle aussi s'effaça à l'arrière-plan, une offre de niche plutôt qu'un fondement de nouveau système.</p> <p>Puis vint l'ordinateur, et avec lui une nouvelle vague d'optimisme.</p> <p>L'enseignement programmé, terme inventé par le psychologue B. F. Skinner dans les années 1950, offrit une vision tentante : un contenu divisé en petites étapes, chaque élève avançant à son propre rythme, avec un retour immédiat à chaque tournant. Les « machines à enseigner » de Skinner étaient mécaniques, maladroites et limitées. Mais la pensée qui les sous-tendait — que l'apprentissage pouvait être personnalisé par un séquençage minutieux et un renforcement constant — planta une graine qui allait grandir pendant des décennies.</p> <p>L'arrivée de l'ordinateur personnel dans les années 1980 amena les ordinateurs dans les écoles à grande échelle. L'enseignement assisté par ordinateur apparut dans les laboratoires et les salles de classe de tout le monde développé. Pourtant, les résultats furent, selon la plupart des mesures, modestes. De nombreuses machines prirent la poussière. De nombreux logiciels restèrent inutilisés.</p> <p>Les chercheurs ont proposé des façons d'expliquer ce schéma de battage et de déception.</p> <p>Larry Cuban, historien de l'éducation à Stanford, est peut-être la voix la plus éminente. Son travail documente une vérité récurrente : les écoles sont remarquablement résistantes au changement fondamental, absorbant les nouvelles technologies sans en être transformées. L'analyse de Cuban suggère que l'innovation technologique façonne rarement la scolarité ; plutôt, la scolarité façonne la technologie, la pliant aux routines, aux emplois du temps et aux structures de pouvoir existants.</p> <p>C'est dans ce paysage d'attentes modérées que l'intelligence artificielle est arrivée.</p> <p>L'IA diffère de ses prédécesseurs sur un point crucial. Les technologies antérieures étaient essentiellement des systèmes de distribution — elles transportaient du contenu d'une source vers un apprenant. L'IA, en revanche, peut répondre. Elle peut s'adapter. Elle peut converser. Elle peut générer. Ces différences ne sont pas triviales. Un manuel ne peut pas voir que vous avez du mal et changer son explication. Un système intelligent, si. Cette seule capacité explique pourquoi de nombreux observateurs croient que l'IA brisera le cycle historique que la radio, la télévision et l'ordinateur personnel n'ont pas pu briser.</p> <p>S'ils ont raison est la question que ce mémoire existe pour examiner.</p> <h2> 3. Comment l'IA fonctionne en classe </h2> <p>Il est tentant de penser à l'IA d'une salle de classe comme à une chose unique et monolithique. Ce n'est pas le cas.</p> <p>Une salle de classe d'aujourd'hui contient plus probablement de nombreux systèmes d'IA différents, chacun faisant un travail différent. Certains sont visibles. Certains ne le sont pas.</p> <p>Considérez le système de gestion de l'apprentissage qui suit l'assiduité et les notes. Sous sa surface, des algorithmes prédisent silencieusement quels élèves risquent de décrocher, signalant leurs noms pour une intervention avant que le problème ne devienne évident. C'est de l'IA, bien que peu de gens l'appelleraient ainsi.</p> <p>Considérez l'assistant de rédaction qui souligne les erreurs de grammaire d'un élève et suggère des améliorations. Il utilise des réseaux de neurones entraînés sur des millions d'exemples d'écriture humaine. C'est de l'IA, intégrée si parfaitement qu'elle semble ordinaire.</p> <p>Considérez la plateforme de tutorat qui présente un problème, attend une réponse, puis — sur la base de cette seule réponse — décide quoi montrer ensuite. C'est l'IA dans sa forme éducative la plus directe : un système prenant des décisions pédagogiques en temps réel.</p> <p>Il y a un point technique important à énoncer clairement.</p> <p>La vague actuelle d'IA dans l'éducation repose en grande partie sur l'apprentissage automatique, et en particulier sur l'apprentissage profond. Ces systèmes ne suivent pas des règles écrites à la main. Ils apprennent des schémas à partir de vastes quantités de données. Un modèle de langage qui génère un retour pour les essais d'élèves n'a pas appris la grammaire comme un ensemble de principes ; il a absorbé la grammaire en lisant des milliards de phrases et en intériorisant leurs régularités statistiques.</p> <p>Cela a deux conséquences profondes.</p> <p>Premièrement, ces systèmes sont remarquablement compétents. Ils peuvent produire une prose courante, résoudre des problèmes mathématiques, répondre à des questions dans une gamme étonnante de sujets et s'engager dans un dialogue prolongé qui semble souvent naturel.</p> <p>Deuxièmement, ces systèmes sont étrangement inconnus. Parce qu'ils apprennent plutôt que d'être programmés, leur comportement n'est pas toujours prévisible ou interprétable. Un système peut produire une explication brillante un instant et une erreur confiante le suivant. Il n'a pas de vraie compréhension au sens humain ; il a des schémas, et les schémas sont puissants mais pas les mêmes que la compréhension.</p> <p>En classe, ces forces et faiblesses se traduisent en réalités concrètes.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> L'IA DANS LA SALLE DE CLASSE ________________________________________________________ | | | Système de gestion Assistant de rédaction | | prédit le décrochage corrige la grammaire | | | | | | Plateforme de tutorat Moteur d'évaluation | | adapte chaque problème note les essais | | | | TOUS UTILISENT : DONNÉES --&gt; SCHÉMAS --&gt; PRÉDICTION | |________________________________________________________| </code></pre> </div> <p>Les forces permettent la pratique personnalisée, le retour instantané et un degré d'attention individuelle qu'aucun enseignant humain ne pourrait offrir simultanément à trente élèves. Les faiblesses créent des risques : la possibilité de renforcer les erreurs, l'opacité de la prise de décision et le danger omniprésent qu'un système se trompe avec toute la confiance injustifiée d'une machine qui a vu la bonne réponse une fois et croit maintenant qu'elle est toujours juste.</p> <p>Un récit honnête de l'IA dans l'éducation doit tenir ces deux vérités à la fois. La technologie est réellement prometteuse. Et elle est réellement imparfaite.</p> <h2> 4. Apprentissage personnalisé et systèmes adaptatifs </h2> <p>Peu de phrases dans la technologie éducative sont invoquées avec plus de révérence et moins de précision que « l'apprentissage personnalisé ».</p> <p>L'idée centrale, cependant, est simple et attrayante. Chaque élève apprend différemment. Chaque élève arrive avec des connaissances préalables différentes, des forces différentes, des faiblesses différentes et des taux de progrès différents. Un système qui peut percevoir ces différences et y répondre individuellement a le potentiel de servir chaque élève bien mieux qu'un programme unique.</p> <p>Les systèmes d'apprentissage adaptatif incarnent cette idée dans un logiciel.</p> <p>Ces systèmes fonctionnent généralement en cycle. Ils présentent un problème ; ils observent la réponse de l'élève ; ils mettent à jour un modèle interne des connaissances de cet élève ; et ils choisissent le problème suivant en fonction de ce modèle. Si un élève a du mal avec les fractions, le système lui donne plus de pratique avec les fractions avant de continuer. Si un élève les maîtrise, le système avance. La boucle de rétroaction est le cœur de la conception.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> PRÉSENTE OBSERVE MET À JOUR un problème la réponse le modèle de l'élève \ / | \ / | \________/______ | | | CHOISIT LE PROCHAIN PROBLÈME &lt;-------+ </code></pre> </div> <p>Les résultats peuvent être frappants dans des contextes spécifiques.</p> <p>En mathématiques, les systèmes adaptatifs ont montré certaines des preuves les plus fortes. Des études de systèmes utilisés pour la pratique arithmétique et l'instruction algébrique ont rapporté des gains significatifs dans la réussite des élèves, en particulier pour les apprenants qui commencent en retard par rapport à leurs pairs. Le mécanisme a un sens intuitif : un élève qui autrement décrocherait et abandonnerait — ou qui s'ennuierait assis pendant que les camarades rattrapent — reçoit à la place un travail calibré précisément à son niveau. L'expérience du flux, d'un travail ni trop facile ni trop difficile, est préservée.</p> <p>Pourtant, les preuves ne sont pas uniformément positives, et les raisons sont instructives.</p> <p>Premièrement, les systèmes adaptatifs ne valent que ce que vaut leur modèle sous-jacent de la connaissance. Un élève peut résoudre un problème correctement pour de mauvaises raisons, ou incorrectement pour de bonnes. Le système ne voit que les résultats, pas le raisonnement derrière eux. Quand le but est une compétence procédurale, cela importe peu. Quand le but est une compréhension conceptuelle, cela importe énormément.</p> <p>Deuxièmement, la personnalisation du rythme n'est pas la même que la personnalisation du sens. Un système peut ajuster la difficulté des problèmes, mais il ne peut pas facilement saisir pourquoi un élève a perdu sa motivation, ni reconnaître un moment de difficulté personnelle, ni offrir l'encouragement que seul un humain attentionné peut fournir. Ces dimensions de l'apprentissage comptent, et les machines restent largement aveugles à elles.</p> <p>Troisièmement, il y a un risque que la personnalisation rétrécisse plutôt qu'elle n'élargisse l'ambition. Un système qui optimise la maîtrise d'un programme fixe pourrait garder un élève en difficulté à pratiquer éternellement les mêmes compétences procédurales, sans jamais l'exposer aux aspects plus riches et plus créatifs du sujet qui pourraient réellement enflammer son intérêt. La personnalisation, mal faite, peut devenir une cage.</p> <p>La littérature sur l'apprentissage adaptatif aboutit donc à une conclusion équilibrée. La technologie est réelle et, dans les bonnes conditions, produit des bénéfices mesurables. Mais c'est un outil parmi d'autres, pas un remplacement d'une pédagogie réfléchie. Sa valeur dépend, comme toujours, de l'habileté avec laquelle elle est utilisée.</p> <h2> 5. Évaluation et retour automatisés </h2> <p>La correction est parmi les tâches les plus chronophages et les moins appréciées de l'enseignement.</p> <p>Un enseignant avec cent élèves peut passer une soirée à parcourir cent ensembles de travaux écrits. Les exigences de la correction concurrencent directement les exigences de la planification, de la préparation et des interactions humaines qui donnent son sens à l'enseignement. Il n'est pas étonnant que l'automatisation soit arrivée ici avec force.</p> <p>L'évaluation automatisée se présente sous plusieurs formes.</p> <p>La plus simple implique des questions à choix multiples et à réponse courte, où les ordinateurs notent avec précision depuis des décennies. Ces systèmes sont fiables, rapides et inoffensifs dans de nombreux contextes, bien qu'ils testent une tranche étroite de l'apprentissage et récompensent la reconnaissance plutôt que le raisonnement.</p> <p>Plus ambitieux sont les systèmes conçus pour évaluer les écrits longs.</p> <p>Ces systèmes utilisent le traitement du langage naturel pour évaluer les essais sur des dimensions telles que la clarté, l'organisation, la grammaire et l'adhésion à un sujet. Aux tests standardisés, de tels systèmes ont montré qu'ils produisent des scores qui corrèlent raisonnablement avec les évaluateurs humains. Pour un retour formatif à faible enjeu — aider un élève à voir qu'un paragraphe est hors sujet, qu'une thèse est enfouie, que des transitions manquent — les outils peuvent être réellement utiles, offrant des conseils à une échelle qu'aucun enseignant ne peut égaler.</p> <p>Mais les risques sont substantiels et méritent un examen attentif.</p> <p>Un système de rédaction entraîné sur des essais passés apprend à quoi ressemblaient les essais passés. Il peut récompenser une structure stéréotypée, punir l'individualité stylistique et être contourné par des élèves qui apprennent à écrire selon le schéma que l'algorithme favorise. Plus troublant, ces systèmes ne peuvent pas vraiment lire. Ils détectent des signaux statistiques, pas du sens. Un essai magnifiquement argumenté qui s'écarte de la convention peut être noté durement ; un essai vide qui reflète la forme attendue peut recevoir une note généreuse.</p> <p>Il y a aussi la question de ce que l'automatisation fait au retour lui-même.</p> <p>Le retour est le plus puissant quand il est immédiat, spécifique et encourageant. L'IA peut fournir les deux premiers, et dans une certaine mesure le troisième. Mais le retour est aussi une relation. Un élève qui reçoit une note et quelques commentaires d'une machine a reçu de l'information ; un élève qui discute de son travail avec un enseignant a reçu davantage. Le premier est efficace. Le second est transformateur. Un système éducatif sain a besoin des deux, et doit résister à la tentation de laisser l'efficacité évincer la transformation.</p> <p>La voie la plus sage semble être hybride.</p> <p>Laissez les machines gérer le routinier, le répétable, les parties de l'évaluation qui épuisent les enseignants. Réservez aux humains le jugement qui exige la compréhension, l'encouragement qui exige l'empathie et les conversations qui exigent la présence. Cette division du travail n'est pas un compromis ; c'est un principe de conception.</p> <h2> 6. Le rôle de l'enseignant à l'ère de l'IA </h2> <p>Chaque conversation sur l'IA dans l'éducation finit par arriver à la même question anxieuse : les machines remplaceront-elles les enseignants ?</p> <p>La réponse courte est non. La réponse plus longue est plus intéressante.</p> <p>Enseigner n'est pas simplement la transmission d'informations. Si c'était le cas, alors oui, l'argument du remplacement serait fort, car les machines excellent à transmettre des informations. Mais enseigner, c'est aussi la culture de relations, la modélisation de la curiosité, la reconnaissance d'un élève qui lutte silencieusement, la célébration d'un élève qui comprend enfin. Ces actes exigent une présence humaine. Aucun algorithme ne peut regarder un élève dans les yeux et dire « je crois en toi » en le pensant.</p> <p>Pourtant, la question, reformulée, devient juste.</p> <p>Les enseignants ne seront pas remplacés. Mais le travail d'enseigner changera, et il change déjà.</p> <p>L'enseignant qui doit passer des heures à corriger pourrait à la place passer ces heures en tête-à-tête avec les élèves. L'enseignant qui ne peut pas donner à chaque élève une attention individuelle pourrait, avec le soutien de l'IA, s'en rapprocher. L'enseignant qui est la seule source de savoir dans une salle d'apprenants devient autre chose quand le savoir est disponible partout : un navigateur, un mentor, un coach.</p> <p>Ce changement n'est pas entièrement confortable. Il demande aux enseignants d'abandonner une mesure de contrôle sur leurs propres classes, de faire confiance à des systèmes qu'ils n'ont pas construits et qu'ils ne comprennent peut-être pas entièrement. Il exige de nouvelles compétences — la capacité d'évaluer la qualité d'un outil, de reconnaître ses biais, d'intercepter ses erreurs. Ce ne sont pas des exigences triviales.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ENSEIGNANT : AVANT ENSEIGNANT : AVEC L'IA -------------------- ---------------------- source unique de savoir navigateur / coach corrige chaque essai corrige haut-enjeu seulement même leçon pour tous dirige le travail personnalisé seul à donner l'attention attention centrée sur le vrai besoin </code></pre> </div> <p>Certains enseignants embrasseront le changement. D'autres y résisteront, et leur résistance n'est pas un simple entêtement. C'est une méfiance compréhensible envers une force qu'ils n'ont pas choisie et qu'ils ne peuvent pas entièrement contrôler. La mise en œuvre sage de l'IA dans l'éducation le reconnaît. Elle n'impose pas la technologie d'en haut en attendant de la gratitude. Elle travaille avec les enseignants, les forme, les écoute et les traite comme des partenaires plutôt que des obstacles.</p> <p>Il y a aussi un point plus profond sur le rôle humain.</p> <p>La force la plus puissante dans la vie d'un jeune, après la famille, est souvent un enseignant. Un enseignant peut changer la trajectoire d'une vie de façons difficiles à mesurer et impossibles à algorithmiser. L'arrivée de l'IA ne diminue pas ce pouvoir ; si quelque chose, elle le met en relief plus net. Alors que les machines prennent en charge les tâches mécaniques de l'enseignement, le travail vraiment humain — l'encouragement, la croyance, la relation — devient plus clairement le centre de la profession, pas moins.</p> <p>L'enseignant, autrement dit, n'est pas la victime de l'ère de l'IA. L'enseignant est sa ressource la plus essentielle.</p> <h2> 7. Mesurer ce qui fonctionne : preuves et résultats </h2> <p>C'est une chose de croire que l'IA aide les élèves à apprendre. C'en est une autre de le démontrer.</p> <p>La base de preuves pour l'IA dans l'éducation est jeune, et elle est inégale. Certains domaines ont été largement étudiés ; d'autres reposent sur des anecdotes prometteuses et des affirmations de vendeurs. Une évaluation sérieuse exige que nous regardions ce que la recherche montre réellement.</p> <p>En mathématiques et dans les matières procédurales, les preuves sont les plus encourageantes. Des méta-analyses de systèmes de tutorat intelligent ont rapporté des effets moyens positifs sur les résultats d'apprentissage, certaines études trouvant des gains comparables à ceux du tutorat humain individuel. La cohérence de ces résultats à travers plusieurs études est une raison authentique d'optimisme.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> FORCE DES PREUVES POUR L'IA DANS L'ÉDUCATION Math / procédural ████████████████ forte Lecture / langues ███████████ modérée Écriture / compréhension ██████ mitigée Créativité / avancé ███ mince </code></pre> </div> <p>En lecture et en apprentissage des langues, le tableau est plus mitigé. Les outils qui fournissent la pratique du vocabulaire et des exercices de lecture adaptatifs montrent des bénéfices. Les outils qui tentent d'évaluer ou d'améliorer l'écriture et la compréhension d'ordre supérieur montrent des résultats moins cohérents, reflétant la difficulté de la tâche sous-jacente.</p> <p>Dans les domaines exigeant une compréhension conceptuelle profonde ou la créativité — les humanités, les arts, la résolution de problèmes avancée — les preuves sont les plus minces. Ici, les limites de l'IA actuelle, son incapacité à vraiment comprendre, deviennent les plus saillantes. Peu d'études suggèrent que l'IA peut remplacer une instruction humaine qualifiée dans ces domaines, et beaucoup suggèrent qu'elle ne peut pas encore ajouter grand-chose au-delà de ce qu'un bon enseignant fait déjà.</p> <p>Tout aussi importante est la question de <em>qui</em> la technologie aide.</p> <p>Il y a une inquiétude persistante, soutenue par certaines recherches, que les outils adaptatifs aident les élèves qui ont le moins besoin d'aide et échouent à atteindre ceux qui en ont le plus besoin. Les élèves ayant une autorégulation bien développée, une bonne littératie et des foyers solidaires peuvent bénéficier le plus de l'apprentissage piloté par l'IA parce qu'ils peuvent le naviguer de façon indépendante. Les élèves qui manquent de ces avantages peuvent lutter pour utiliser les outils efficacement, tombant encore plus loin derrière. La technologie a une tendance à amplifier les inégalités existantes plutôt que de les effacer, et l'IA montre des signes inquiétants de suivre le même schéma.</p> <p>Il y a aussi le défi des preuves elles-mêmes.</p> <p>Une grande partie de la recherche dans ce domaine est financée par les entreprises qui vendent les outils. Une grande partie est menée dans des conditions courtes et artificielles plutôt que dans de vraies classes sur de longues périodes. Le biais de publication — la tendance des revues à publier des résultats positifs — déforme davantage le tableau. Une évaluation honnête doit traiter les découvertes prometteuses avec un degré de scepticisme et exiger une réplication dans des contextes authentiques.</p> <p>La conclusion d'ensemble n'est ni extatique ni dédaigneuse.</p> <p>L'IA peut améliorer l'apprentissage dans les bonnes conditions. Ces conditions incluent une pédagogie saine, des outils bien conçus, des enseignants formés et une attention à l'équité. Quand ces éléments sont présents, les gains sont réels. Quand ils sont absents, la technologie fait peu de bien et peut faire du mal. La variable qui détermine le succès n'est pas la sophistication de l'algorithme. C'est la qualité de l'environnement dans lequel l'algorithme opère.</p> <h2> 8. Éthique, vie privée et biais </h2> <p>Aucune discussion sur l'IA dans l'éducation ne serait complète sans affronter les questions éthiques qui ombragent la technologie. Ce ne sont pas des pensées après coup. Elles sont centrales.</p> <p>La vie privée est la première et la plus immédiate des préoccupations.</p> <p>L'IA éducative est construite sur des données. Pour personnaliser l'apprentissage, un système collecte des informations sur ce qu'un élève sait, comment il apprend, à quelle vitesse il progresse, où il a du mal. Parfois, il collecte bien plus : frappes clavier, schémas de lecture, réponses émotionnelles déduites des expressions faciales ou des vitesses de frappe. Ce sont des données intimes, recueillies auprès d'enfants, souvent sans consentement significatif.</p> <p>L'échelle de la collecte est stupéfiante, et les fins ne sont pas toujours claires. Les données collectées pour améliorer l'apprentissage peuvent être utilisées à d'autres fins — classer les écoles, faire du marketing auprès des élèves, construire des profils qui suivent une personne longtemps après qu'elle a quitté la classe. Les enfants profilés aujourd'hui ne sont pas en position de consentir à la façon dont leurs données seront utilisées dans vingt ans. Cette asymétrie de pouvoir et de conséquence mérite bien plus d'attention qu'elle n'en a reçue.</p> <p>Puis il y a le biais.</p> <p>Les systèmes d'IA apprennent à partir de données, et les données reflètent le monde, y compris ses injustices. Un système entraîné sur des essais d'un groupe démographique peut systématiquement sous-évaluer l'écriture d'un autre. Un système adaptatif optimisé sur la performance d'élèves bien dotés en ressources peut mal juger et sous-server les élèves de milieux différents. Le biais dans l'IA n'est pas un bug qui apparaît occasionnellement ; c'est une caractéristique structurelle qui apparaît chaque fois que les données sous-jacentes sont inégales, et la charge de le détecter et de le corriger incombe non pas aux données mais aux humains qui conçoivent et déploient le système.</p> <p>Il y a aussi une dimension éthique plus subtile : le risque de surveillance.</p> <p>Une classe dans laquelle chaque action est surveillée, chaque erreur enregistrée, chaque pause chronométrée, est une classe qui a changé de caractère, que quelqu'un l'ait voulu ou non. Les élèves sous observation constante apprennent différemment. Ils prennent moins de risques. Ils se conforment. Les qualités mêmes que l'éducation est censée cultiver — la curiosité, l'expérimentation, le courage de se tromper — peuvent être silencieusement réprimées par un environnement qui regarde de trop près.</p> <p>Rien de tout cela ne signifie que l'IA doit être rejetée. Mais cela signifie que son adoption exige un échafaudage éthique délibéré : des règles claires sur les données, de fortes protections du consentement, une surveillance indépendante et une véritable transparence sur ce que ces systèmes font des informations qu'ils recueillent. L'éthique n'est pas une fonctionnalité à ajouter plus tard. C'est une contrainte qui doit être intégrée dès le début.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> L'ÉQUILIBRE À TROUVER | ________/ \________ / \ GAIN ÉDUCATIF COÛT HUMAIN apprentissage person. perte de vie privée retour immédiat biais &amp; profilage chaque élève servi apprendre sous surveillance ___________________ ___________________ ^ ^ |______ garder l'œil sur les deux ______| </code></pre> </div> <h2> 9. Politique et mise en œuvre </h2> <p>Une bonne politique fait la différence entre un outil qui sert et un outil qui domine.</p> <p>La trajectoire de l'IA dans l'éducation n'est pas déterminée par la technologie seule. Elle est façonnée par des décisions prises dans les législatures, les ministères, les conseils scolaires et les bureaux de district. Ces décisions déterminent ce qui est permis, ce qui est financé, ce qui est enseigné et qui est protégé.</p> <p>La première priorité politique est la protection des données. Les élèves méritent les mêmes garanties de vie privée dans les espaces numériques qu'ils ont déjà dans les espaces physiques. Des règles fortes devraient régir quelles données peuvent être collectées, qui peut y accéder, combien de temps elles peuvent être conservées et si elles peuvent être utilisées à toute fin au-delà de l'éducation de l'élève lui-même. Ces règles devraient être applicables et entraîner de réelles conséquences en cas de violation.</p> <p>La deuxième priorité est l'équité. Laissés au marché, les meilleurs outils d'IA iront vers les écoles qui peuvent se les permettre, élargissant l'écart entre les districts riches et pauvres, entre les pays avec des ressources et ceux sans. La politique publique a un rôle pour garantir que les bénéfices de l'IA atteignent les élèves qui ont le plus à gagner. Cela peut exiger un investissement public, des outils open source et des partenariats qui priorisent l'accès sur le profit.</p> <p>La troisième priorité est le soutien aux enseignants. L'adoption réussit ou échoue dans la classe, et les classes sont dirigées par des enseignants. Les politiques doivent inclure un financement pour la formation, du temps pour l'expérimentation et le respect du jugement professionnel. Un enseignant à qui l'on donne un outil et qu'on abandonne est un enseignant qui le rejettera. Un enseignant qui est soutenu, formé et en qui on a confiance est un enseignant qui peut le transformer.</p> <p>La quatrième priorité est l'évaluation indépendante. Le gouvernement et les écoles ne devraient pas compter sur les vendeurs pour prouver que leurs propres produits fonctionnent. Il y a une place pour une évaluation indépendante, rigoureuse et transparente qui pose des questions difficiles sur l'efficacité, la sécurité et l'équité avant que ces systèmes soient déployés à des millions d'enfants.</p> <p>Enfin, la politique doit être agile. La technologie évolue plus vite que les lois ne peuvent suivre. Les régulateurs ont besoin de mécanismes qui peuvent répondre au changement sans sacrifier les protections que la stabilité exige. C'est un équilibre difficile, et se tromper dans l'un ou l'autre sens coûte cher — trop lent, et le mal est fait ; trop rapide, et l'innovation est étouffée.</p> <p>Le but réaliste n'est pas de résoudre parfaitement toutes ces tensions, mais de les gérer délibérément, de façon transparente et avec les intérêts des élèves au centre.</p> <h2> 10. L'avenir de l'IA dans l'éducation </h2> <p>Prédire l'avenir est une tâche de fou. Pourtant, nous pouvons discerner des tendances dignes d'être prises au sérieux.</p> <p>La première tendance est la capacité. Les systèmes d'IA continueront de devenir plus puissants, plus courants et plus polyvalents. Les systèmes qui semblent impressionnants aujourd'hui paraîtront primitifs dans une décennie. C'est presque certain.</p> <p>La deuxième tendance est l'intégration. L'IA cessera d'être une caractéristique distincte et deviendra une partie invisible de l'environnement éducatif. Elle vivra dans le contenu que les élèves lisent, les évaluations qu'ils passent, le retour qu'ils reçoivent, même dans la façon dont les classes sont planifiées et gérées. La question ne sera plus de savoir si l'IA est présente, mais à quel point elle est bien utilisée.</p> <p>La troisième tendance est une personnalisation plus profonde. Les systèmes futurs pourraient modéliser non seulement ce qu'un élève sait, mais comment il apprend le mieux — son style cognitif, son profil motivationnel, ses schémas d'attention. Cela pourrait offrir un soutien extraordinaire aux apprenants individuels. Cela pourrait aussi apporter de nouveaux et graves risques pour la vie privée et l'autonomie, alors que les contours intérieurs d'un esprit deviennent des données.</p> <p>Il y a aussi la possibilité d'une redéfinition de l'objectif éducatif lui-même.</p> <p>Si l'IA peut accomplir beaucoup des tâches cognitives que nous enseignons actuellement — le calcul, la grammaire, le rappel, même certaines formes d'analyse — alors la question se pose : qu'est-ce que les écoles devraient enseigner ? La réponse pourrait se déplacer vers ce qui est distinctement humain : la pensée critique, la créativité, l'éthique, la collaboration, la résilience, la capacité de poser de bonnes questions plutôt que de simplement y répondre. L'IA pourrait ne pas détruire ces compétences. Elle pourrait, en rendant la cognition de routine bon marché, les élever au centre de l'éducation.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> COGNITION DE ROUTINE CE QUE LES ÉCOLES ENSEIGNENT _____________________ ______________________ calcul XXXX pensée critique grammaire XXXX créativité rappel XXXX éthique &amp; collaboration analyse basique XXXX résilience &amp; bonnes questions </code></pre> </div> <p>Et pourtant. L'avenir n'est pas écrit. Il est fait de choix.</p> <p>Les choix faits par la génération actuelle d'éducateurs, de technologues et de décideurs détermineront si l'IA devient une force pour approfondir l'opportunité ou pour durcir l'inégalité. Les deux avenirs sont disponibles. Les deux sont soutenus par la même technologie. Ce qui les sépare, c'est le jugement — le jugement de construire avec soin, de réglementer sagement, de surveiller le mal et de garder l'être humain, pas la machine, au centre de l'apprentissage.</p> <h2> 11. Conclusion </h2> <p>L'intelligence artificielle est arrivée dans l'éducation. Elle n'arrive pas ; elle est là.</p> <p>Ce mémoire a soutenu que la technologie est réelle, que sa promesse est authentique et que ses risques sont sérieux. Il a soutenu que l'IA peut améliorer l'apprentissage, mais seulement dans les bonnes conditions. Il a soutenu que les enseignants ne seront pas remplacés, mais que leur travail changera. Il a soutenu que les questions les plus profondes soulevées par l'IA dans l'éducation ne sont pas techniques mais humaines.</p> <p>Nous sommes, en un sens, à un embranchement de la route.</p> <p>Un chemin mène à un avenir où des systèmes adaptatifs tendent chaque esprit capable, où le retour est immédiat, où aucun élève ne tombe silencieusement à travers les fissures parce qu'une machine a remarqué le premier. C'est une belle vision.</p> <p>L'autre chemin mène à un avenir où les meilleurs outils n'appartiennent qu'à quelques-uns, où les enfants sont profilés et triés par des algorithmes opaques, où l'apprentissage est optimisé jusqu'à perdre son âme. C'est une vision d'avertissement.</p> <p>La route n'est pas choisie par la technologie. Elle est choisie par nous.</p> <p>La promesse de l'IA dans l'éducation est qu'elle pourrait enfin briser le cycle historique de déception qui a suivi chaque technologie éducative avant elle — qu'elle pourrait tenir le rêve d'un apprentissage qui se plie autour de l'apprenant. Le risque est qu'elle devienne une autre entrée dans ce cycle, un autre objet brillant qui a promis beaucoup et livré peu, ou pire, qu'elle livre sa promesse sélectivement à quelques privilégiés.</p> <p>Ni l'un ni l'autre de ces aboutissements n'est inévitable. Les deux sont possibles. L'avenir de l'apprentissage — que l'IA le rende plus large ou plus étroit, plus riche ou plus creux, plus juste ou moins — réside dans les choix que nous faisons maintenant.</p> <p>Ce serait une étrange trahison de notre humanité de faire ces choix avec négligence. Ce serait une trahison égale de refuser de les faire du tout.</p> <h2> 12. Références </h2> <p>Baker, R. S. (2016). Stupid tutoring systems, intelligent humans. <em>International Journal of Artificial Intelligence in Education</em>, 26(2), 600–614.</p> <p>Cuban, L. (1986). <em>Teachers and Machines: The Classroom Use of Technology Since 1920</em>. Teachers College Press.</p> <p>Kulik, J. A., &amp; Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. <em>Review of Educational Research</em>, 86(1), 42–78.</p> <p>Luckin, R., Holmes, W., Griffiths, M., &amp; Forcier, L. B. (2016). <em>Intelligence Unleashed: An Argument for AI in Education</em>. Pearson.</p> <p>Ma, W., Adesope, O. O., Nesbit, J. C., &amp; Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. <em>Journal of Educational Psychology</em>, 106(4), 901–918.</p> <p>Selwyn, N. (2019). <em>Should Robots Replace Teachers? AI and the Future of Education</em>. Polity Press.</p> <p>Skinner, B. F. (1958). Teaching machines. <em>Science</em>, 128(3330), 969–977.</p> <p>Williamson, B. (2017). <em>Big Data in Education: The Digital Future of Learning, Policy and Practice</em>. Sage.</p> <p>Zawacki-Richter, O., Marín, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. <em>International Journal of Educational Technology in Higher Education</em>, 16(1), 39.</p> ai education learning ogc16 Thu, 27 Aug 2026 12:27:40 +0000 https://dev.to/ogc16/-li-jie-ke-tang-ren-gong-zhi-neng-ru-he-zhong-su-wo-men-de-xue-xi-fang-shi-2oo9 https://dev.to/ogc16/-li-jie-ke-tang-ren-gong-zhi-neng-ru-he-zhong-su-wo-men-de-xue-xi-fang-shi-2oo9 <h2> 学位论文 </h2> <h2> 摘要 </h2> <p>人工智能已不再是遥远的承诺。它已经走进了教室。本论文考察了人工智能工具正在如何改变教育——从适应每位学习者的个性化辅导系统，到几秒钟内给出反馈的智能批改，再到伴随每项创新而来的那些无声的道德问题。</p> <p>这里的论点并不是说人工智能会取代教师。它不会。相反，本论文提出一个更细致的观点：人工智能如果被明智地使用，可以把教师从那些机器能做到的事情中解放出来——让他们去做教书育人、激励启发、建立联结这些机器做不到的事。然而，这一承诺完全取决于我们如何选择去构建和部署这些系统。</p> <p>接下来的章节将沿着一条路径展开：从教育技术的历史根源，到当前人工智能工具的版图，到其对学习的可衡量影响，最后深入到那些任何学生、教师或政策制定者都无法忽视的道德与政策问题。</p> <h2> 目录 </h2> <ol> <li>引言</li> <li>教育技术简史</li> <li>人工智能如何在课堂中运作</li> <li>个性化学习与自适应系统</li> <li>自动化评估与反馈</li> <li>人工智能时代教师的作用</li> <li>衡量成效：证据与结果</li> <li>伦理、隐私与偏见</li> <li>政策与实施</li> <li>人工智能在教育中的未来</li> <li>结论</li> <li>参考文献</li> </ol> <h2> 1. 引言 </h2> <p>我们生活在一个工具非凡的时代。</p> <p>驱动自动驾驶汽车和医疗诊断的同一项技术，如今也把一位永不疲倦、从不评判、记得学生每一次回答的导师交到了学习者手中。这是一个惊人的前景。而且它已经到来了。</p> <p>本论文所讨论的，正是当这个现实与课堂相遇时会发生什么。</p> <p>教育向来变革迟缓。黑板让位于白板。白板让位于投影仪。投影仪让位于平板电脑。但在每一层新硬件之下，其基本结构始终顽固地保持不变：一位教师、众多学生、一套固定的课程，还有一个对所有人都同样滴答作响的时钟。人工智能的出现，威胁着要打破这种结构。它提供了一种可能：让学习去围绕学习者弯曲，而不是强迫学习者去适应学习。</p> <p>指导本研究的核心研究问题简单却难答：人工智能能否让教育更有效、更公平？如果能，又是在什么条件下？</p> <p>要回答这个问题，我们必须先了解这些系统究竟在做什么。我们必须把真正的进步与营销炒作区分开来。我们必须直面关于数据、隐私的那些令人不安的真相，以及一个风险——那些出于良好意图的工具，可能会加深它们声称要消除的不平等。</p> <p>本论文分三个部分展开。</p> <p>首先，我们建立背景。我们回顾教育技术从何而来，以及为何此前的革命未能兑现其承诺。其次，我们审视当下。我们探索人工智能已经在课堂中具体运作的方式，并权衡其影响的证据。第三，我们展望未来。我们探讨伦理、政策与选择——正是这些将决定这场革命是服务于每一位学生，还是只服务于少数特权者。</p> <p>赌注很高。教育是机会的伟大引擎，是让家庭跨越世代实现跃迁的力量。如果人工智能让它变得更强，我们就获得了无法估量的财富。如果人工智能让它变得更加狭窄，我们失去的东西可能永远无法挽回。本论文正是为了理解我们正在建设的是哪一个未来。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 未来之一 未来之二 (机会) (不平等) | | | 每个头脑都被托举 | 最好的工具只属于少数 | 无人无声滑落 | 不透明的儿童画像分拣 | 反馈即刻到来 | 学习失去灵魂 | | \__________ ____________/ \/ / \ / 你 \ / 来抉择 \ /____________\ </code></pre> </div> <h2> 2. 教育技术简史 </h2> <p>要理解我们将走向何方，我们必须先理解我们曾走过怎样的路。</p> <p>教育中技术的故事，是一个循环的故事。一次又一次，新发明带着变革的宏大承诺到来。又一次又一次，它退居为佐助的角色——有用，却很少具有革命性。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 宏大承诺 | v 希望与狂热 | v 现实降临 | v 退居佐助角色 &lt;---- 机器并未统治课堂 </code></pre> </div> <p>想想广播。当广播信号在20世纪20年代传入美国家庭时，热衷者曾预言，全国的每个孩子都将很快由寥寥几位杰出的讲师授课，他们的声音被送到农舍厨房和城市公寓。这件事并未发生。广播成了补充，而不是替代。</p> <p>想想电视。在20世纪50年代和60年代，教学电视承诺把世界上最优秀的教师送进每个起居室。它同样悄然退居幕后，成为一种小众选择，而非新体系的基石。</p> <p>然后是计算机的到来，随之而来的是新一轮乐观情绪。</p> <p>程序教学，这是心理学家斯金纳在20世纪50年代提出的术语，提供了一个诱人的愿景：把内容分解为许多小步骤，让每个学生按自己的节奏推进，每一步都得到即时反馈。斯金纳的"教学机器"是机械的、笨拙的、有限的。但它们背后的思想——学习可以通过细致的排序和持续的强化来实现个性化——播下了一颗将生长数十年的种子。</p> <p>20世纪80年代个人电脑的到来，让计算机大规模进入学校。程序辅助教学出现在发达国家各地的实验室和教室中。然而，从大多数衡量标准看，结果却相当有限。许多机器尘封不用。许多软件无人问津。</p> <p>学者们提出了理解这种炒作与失望循环的方式。</p> <p>斯坦福大学教育史学家拉里·库班是最突出的声音之一。他的研究记录了一个不断重现的事实：学校对根本性的变革有着惊人的抵抗力，它们吸收新技术却不会被其改造。库班的分析表明，技术革新很少重塑学校教育；相反，是学校教育重塑了技术，把它折叠进现有的惯例、课程表和权力结构之中。</p> <p>正是在这种期望有限的环境中，人工智能到来了。</p> <p>人工智能与它的前辈有一个关键区别。早先的技术本质上是传递系统——它们把内容从源头送到学习者手中。而人工智能则可以回应。它可以适应。它可以对话。它可以生成。这些并不是无足轻重的差异。一本教科书看不到你在苦苦挣扎，也无法换一种解释。一个智能系统却可以。正是这一项能力，让许多观察者相信，人工智能将打破广播、电视和个人电脑都无法打破的历史循环。</p> <p>他们是否正确，正是本论文要考察的问题。</p> <h2> 3. 人工智能如何在课堂中运作 </h2> <p>人们很容易把教室中的人工智能想象成一个单一的、铁板一块的东西。事实并非如此。</p> <p>今天的教室更可能同时容纳许多不同的人工智能系统，各司其职。有些是看得见的。有些则看不见。</p> <p>想想记录出勤和成绩的学习管理系统。在它的表面之下，算法在静静地预测哪些学生有辍学风险，在问题变得明显之前就把他们的名字标记出来等待干预。这就是人工智能，尽管很少有人会这样称呼它。</p> <p>想想为学生划出语法错误并提出修改建议的写作助手。它使用的是在海量人类写作样本上训练出来的神经网络。这就是人工智能，只是嵌入得如此无缝，以至于让人觉得稀松平常。</p> <p>想想那个呈现一道题、等待解答、然后——仅仅依据这一个答案——决定接下来展示什么内容的辅导平台。这就是人工智能最直接的教育形态：一个在实时做出教学决策的系统。</p> <p>有一个重要的技术要点值得直白地说清楚。</p> <p>当前教育领域的这一波人工智能，主要建立于机器学习之上，尤其是深度学习。这些系统并不遵循手写的规则。它们从海量数据中学习规律。一个为作文生成反馈的语言模型，并非被当作一套原则来教授语法；它通过阅读数十亿个句子并内化其中的统计规律来吸收语法。</p> <p>这带来两个深远的后果。</p> <p>第一，这些系统能力惊人。它们能写出流畅的文章，能解数学题，能在令人眼花缭乱的知识范围内回答问题，还能进行常常显得自然的长时间对话。</p> <p>第二，这些系统又陌生得令人不安。因为它们是学习出来的而非编程出来的，其行为并不总是可预测或可解释的。一个系统可能前一秒给出精彩的解释，下一秒就报出一个信心满满的错误。它并不具备人类意义上的真正理解；它有的是规律，而规律虽然强大，却并不等同于理解。</p> <p>在课堂中，这些优点和缺点会转化为特定的现实。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 课堂中的人工智能 ________________________________________________________ | | | 学习管理系统 写作助手 | | 预测辍学风险 纠正语法 | | | | | | 辅导平台 评估引擎 | | 逐个调整个体问题 给作文打分 | | | | 全部使用：数据 --&gt; 学习规律 --&gt; 预测 | |________________________________________________________| </code></pre> </div> <p>优点使个性化练习、即时反馈成为可能，并提供了任何单个人类教师都无法同时为三十名学生提供的那种个别关注。缺点则制造了风险：强化错误的可能、决策过程的不透明，还有一个挥之不去的危险——某个系统会在完全不该有的自信中犯错，只因它见过一次正确答案，便从此认定它总是对的。</p> <p>对课堂中人工智能的诚实描述，必须同时握住这两个真相。这项技术确实前景可期。它也确确实实不完美。</p> <h2> 4. 个性化学习与自适应系统 </h2> <p>在教育技术界，很少有词像"个性化学习"这样被带着最高的敬意、却又最不精确地使用。</p> <p>然而，其核心理念却简单而吸引人。每个学生的学习方式都不同。每个学生带着不同的已有知识、不同的优势、不同的薄弱环节、不同的进步速度而来。一个能够感知这些差异并分别回应的系统，有潜力比一刀切的课程更好地方便每位学生。</p> <p>自适应学习系统把这个理念具体化为软件。</p> <p>这类系统通常以一个循环运作。它们呈现一个问题；观察学生的回应；更新对该学生知识状况的内部模型；再依据这一模型选择下一个问题。如果学生在分数上卡住了，系统就会在继续前进之前让他们在分数上做更多练习。如果学生已经掌握，系统就向前推进。这个反馈回路是整个设计的核心。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 呈现问题 观察回应 更新学生模型 \ / | \ / | \________/______ | | | 选择下一个问题 &lt;---------------+ </code></pre> </div> <p>在特定情境下，结果可能相当可观。</p> <p>在数学领域，自适应系统展现了最强的一些证据。针对算术练习和代数教学系统的研究，报告了学生成绩的显著提升，尤其是对那些起点落后于同伴的学习者。其中的机制在直觉上说得通：一个本来会因为掉队而放弃——或因同学仍在追赶而坐等无聊——的学生，如今被给予恰好与自己水平匹配的练习。那种心流体验，那种既不太难也不太容易的专注状态，得以保留。</p> <p>但证据并非全然正面，其原因很有启发性。</p> <p>第一，自适应系统的好坏，取决于其底层知识模型的好坏。一个学生可能因为错误的原因而答对一道题，也可能因为正确的原因而答错。系统只看到结果，看不到这背后的推理。当目标是程序性技能时，这或许无关紧要。当目标是概念性理解时，这就至关重要了。</p> <p>第二，节奏的个性化并不等于意义的个性化。一个系统可以调整难题的难度，却无法轻松理解一个学生为何失去了动力，无法识别个人的挣扎时刻，也无法提供只有关心他人的人才能给予的鼓励。学习的这些维度同样重要，而机器对它们仍然基本是盲的。</p> <p>第三，个性化存在着使抱负变窄而非变宽的风险。一个以掌握固定课程为优化目标的系统，可能会让苦苦挣扎的学生永远停留在练习同一种程序性技能上，永远不让他们接触那些可能真正点燃其兴趣的、更丰富更有创造性的方面。个性化做不好，就会变成牢笼。</p> <p>因此，关于自适应学习的文献得出了一个均衡的结论。这项技术是真实的，在适当条件下能产生可衡量的好处。但它是众多工具之一，而不是深思熟虑的教学法的替代品。它的价值一如既往，取决于使用它的技巧。</p> <h2> 5. 自动化评估与反馈 </h2> <p>批改作业是教学中最耗费时间、也最不受喜爱的任务之一。</p> <p>一个有上百名学生的教师，可能要用一个晚上来处理上百份书面作业。批改的需求与备课、准备、以及赋予教学意义的人际互动的需求直接竞争。难怪自动化会在这里强势登场。</p> <p>自动化评估有几种形态。</p> <p>最简单的是选择题和简答题，计算机在这一领域准确评分已有数十年。这些系统可靠、迅速，在许多情境下无可非议，尽管它们只考察学习中很窄的一块，奖励的是辨识而非推理。</p> <p>更有雄心的是那些旨在评估长篇文章的系统。</p> <p>这类系统使用自然语言处理，在清晰度、组织、语法和切题程度等维度上评估作文。在标准化考试中，这类系统已被证明能产生与人类评分者合理相关的分数。对于低风险的生成性反馈——帮助学生看出某段偏离主题、某处论点被埋没、某个过渡缺失——这些工具真的很有用，能提供任何教师都难以匹敌的规模性指导。</p> <p>但风险也很可观，值得仔细审视。</p> <p>在旧作文上训练出来的写作系统，学到的是旧作文的样子。它可能奖励刻板的行文结构，惩罚独特的文风，还可能被那些学会按算法偏好的模式来写作的学生钻空子。更令人不安的是，这些系统并不能真正阅读。它们检测的是统计信号，而不是意义。一篇论证优美却偏离常规的作文可能被苛刻打分；一篇空洞无物却符合预期形式的作文则可能得到慷慨的分数。</p> <p>还有一个问题：自动化会对反馈本身造成什么。</p> <p>当反馈即时、具体而鼓舞人心时，它最为有力。人工智能能提供前两者，在某种程度上也能提供第三者。但反馈同时也是一种关系。从机器那里得到一个分数和几句评论的学生，得到的是信息；与教师讨论自己作品的学生，得到的更多。前者有效率。后者有变革力。一个健康的教育体系两者都需要，并且必须抵制让效率挤掉变革的诱惑。</p> <p>最明智的道路似乎是混合式。</p> <p>让机器处理那些常规的、可重复的部分，处理那些把教师耗尽的部分。把需要理解的判断、需要共情的鼓励、需要在场陪伴的对话，留给人类。这种分工不是妥协；它是一种设计原则。</p> <h2> 6. 人工智能时代教师的作用 </h2> <p>每一次关于教育中人工智能的讨论，最终都会抵达同一个焦虑的问题：机器会取代教师吗？</p> <p>简短的回答是不会。更长一点的回答更有意思。</p> <p>教学不仅仅是信息的传递。如果仅仅如此，那么支持取代的论据确实很强，因为机器极擅长传递信息。但教学还是关系的培育、好奇心的示范、对某个正默默挣扎的学生的辨认、对某个终于理解的学生的庆祝。这些行为需要人的在场。没有哪个算法能看着学生的眼睛，说一句"我相信你"，并且是当真的。</p> <p>然而，换个角度，这个问题就变得合理了。</p> <p>教师不会被取代。但教学这项工作会改变，而且已经在改变了。</p> <p>那个必须花上数小时批改的教师，本可以把这些时间用来与学生面谈。那个无法给每个学生个别关注的教师，在人工智能的支持下，可以更接近做到这一点。当一个房间里知识随处可见时，作为知识唯一来源的教师，就变成了另一种角色：领航员、导师、教练。</p> <p>这种转变并不完全舒适。它要求教师交出对自己课堂的一定掌控，去信任那些他们并未构建、或许也并未完全理解的系统。它要求掌握新技能——评价工具质量的能力、识别其偏见的能力、截断其错误的能力。这些绝非轻而易举的要求。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 教师：以前 教师：有了人工智能 --------------------- --------------------- 知识的唯一来源 领航员 / 教练 批改每篇作文 只批改高风险之作 对所有人教同一课 指导学生个性化任务 唯一的关注给予者 把关注聚焦于真正的需求 </code></pre> </div> <p>一些教师会拥抱这种变化。另一些会抵制，而他们的抵制并不仅仅是固执。那是对一种他们未曾选择、也无法完全控制的力量的合情合理的警惕。人工智能在教育中的明智实施应承认这一点。它不应自上而下地强加技术，然后指望感恩。它应与教师合作、培训他们、倾听他们、把他们当作伙伴而非障碍。</p> <p>还有一个更深的、关于人的角色的观点。</p> <p>在一个年轻人生活中，除家庭以外，最强大的力量往往来自教师。一位教师能在难以衡量、也无法算法化的方面改变一个人的人生轨迹。人工智能的到来并不会削弱这种力量；如果说有什么影响，那就是把它映衬得更加鲜明。随着机器接管教学中的机械任务，那些真正属于人的工作——鼓励、信念、关系——变得清晰地成为这个职业的中心，而非边缘。</p> <p>换言之，教师不是人工智能时代中的牺牲品。教师是这个时代最不可或缺的资源。</p> <h2> 7. 衡量成效：证据与结果 </h2> <p>相信人工智能有助于学生学习是一回事。证明它则是另一回事。</p> <p>教育人工智能的证据库还很年轻，而且参差不齐。有些领域得到了广泛研究；另一些则建立在乐观的轶事和供应商的宣传之上。严肃的评估要求我们看看研究实际显示了什么。</p> <p>在数学和程序性学科中，证据最令人鼓舞。对智能辅导系统的元分析报告了学习结果的平均正向效应，一些研究发现的提升足以与人类个别辅导相媲美。这些结果在多项研究中的一致性，是真正值得乐观的理由。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 教育中人工智能的证据强度 数学 / 程序性 ████████████████ 强 阅读 / 语言 ███████████ 中等 写作 / 理解 ██████ 参差 创造 / 高级问题 ███ 薄弱 </code></pre> </div> <p>在阅读和语言学习方面，图景较为庞杂。提供词汇练习和适应性阅读练习的工具显示出益处。试图评估或改善高阶写作与理解力的工具，则显示出较不一致的结果，反映出底层任务的难度。</p> <p>在需要深概念理解或创造力的领域——人文学科、艺术、高级问题解决——证据最为稀薄。在这里，当前人工智能的局限，它无法真正理解这一事实，变得最为突出。很少有研究表明人工智能能在这些领域取代技艺娴熟的人类教学，许多研究则表明它目前还无法在优秀教师已能提供的东西之外增添多少。</p> <p>同样重要的是"它帮助的是谁"这个问题。</p> <p>有一种持续存在的担忧，并被一些研究所支持：自适应工具帮助的是最不需要帮助的学生，却未能触及最需要帮助的学生。自我调节能力强、读写能力好、家庭支持好的学生，可能从人工智能驱动的学习中获益最多，因为他们能独立驾驭它。缺乏这些优势的学生可能难以有效使用工具，反而掉得更远。技术有一种放大现有不平等而非抹平它们的倾向，而人工智能正显示出令人担忧的沿袭这一模式的迹象。</p> <p>此外还有证据本身的问题。</p> <p>这个领域的许多研究由销售这些工具的公司资助。很多研究是在短暂、人为的条件下进行的，而非在真实课堂中长期进行。发表偏倚——期刊倾向于发表正向结果的倾向——进一步扭曲了图景。诚实的评估必须以一定程度的怀疑来对待乐观的发现，并要求在真实环境中进行重复验证。</p> <p>总体结论既非狂喜，也非轻蔑。</p> <p>在合适条件下，人工智能能够改善学习。这些条件包括良好的教学法、设计精良的工具、受过训练的教师，以及对公平的关注。当这些要素齐备时，获益是真实的。当它们缺失时，这项技术几乎无益，甚至可能有害。决定成败的变量，不是算法的精妙程度，而是算法所运行环境的质量。</p> <h2> 8. 伦理、隐私与偏见 </h2> <p>对教育中人工智能的讨论，若不直面笼罩其上的伦理问题，便不算完整。这些问题不是事后补充。它们是核心。</p> <p>隐私是第一项、也是最紧迫的关切。</p> <p>教育人工智能建立于数据之上。为了个性化学习，系统会收集关于学生知道什么、如何学习、进步多快、在哪里卡住的信息。有时它收集的远不止这些：击键记录、阅读模式、从面部表情或打字速度推断出的情绪反应。这是珍贵的个人数据，来自儿童，而且常常并未获得有意义的同意。</p> <p>数据收集的规模令人震惊，而目的并非总是清晰。为改善学习而收集的数据，可以被用于其他目的——为学校排名、向学生营销、构建一个在人们离开课堂很久之后仍如影随形的画像。今天被画像的儿童，并没有能力决定其数据二十年后将被如何使用。这种权力与后果的不对称，所获得的关注远不及它应有的。</p> <p>然后是偏见。</p> <p>人工智能系统从数据中学习，而数据反映着世界，包括世界的不公。一个在某个群体学生的作文上训练出来的系统，可能有系统性地低估另一个群体写作的倾向。一个在资源充足学生的表现上优化的自适应系统，可能错误判断并亲视来自不同背景的学生。人工智能中的偏见不是一个偶然出现的故障；只要底层数据不均，它就是一个必然出现的结构性特征，而检测和纠正它的责任，不属于数据，而属于设计并部署系统的人。</p> <p>还有一个更微妙的伦理维度：监视的风险。</p> <p>一个每时每刻都被监视、每次错误都被记录、每次停顿都被计时的课堂，无论是否有意如此，都已改变了性质。处于持续观察之下的学生，学习的方式会不同。他们更少冒险。他们趋于顺从。恰恰是教育意在培养的那些品质——好奇、试验、敢于犯错的勇气——可能被一个盯得太紧的环境悄然压制。</p> <p>这一切并不意味着必须拒绝人工智能。但它确实意味着，采用人工智能需要精心构筑的伦理支架：关于数据的明确规则、强有力的同意保护、独立的监督，以及对这些系统如何处理所收集信息保持真正的透明。伦理不是后来再添加的功能。它是一项必须从一开始就内建其中的约束。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 我们必须保持的平衡 | ________/ \________ / \ 教育收益 人的代价 个性化学习 隐私的丧失 即时反馈 偏见与画像 每个学生都被服务 在监视下学习 ___________________ ___________________ ^ ^ |______ 同时盯住两者 ______| </code></pre> </div> <h2> 9. 政策与实施 </h2> <p>好的政策，是让工具服务于人而非统治人的关键。</p> <p>人工智能在教育的轨迹，并非由技术单独决定。它是由立法机关、部委、学区董事会和地区办公室做出的决定塑造的。这些决定决定了什么被允许、什么被资助、什么被教授、以及谁受到保护。</p> <p>第一项政策优先是数据保护。学生在数字空间中，应享有与物理空间相同的隐私保障。强有力的规则应规范哪些数据可以被收集、谁可以访问、可以保存多久、以及是否可以被用于学生自身教育之外的任何目的。这些规则应是可执行的，并应为违反行为带来真正的后果。</p> <p>第二项优先是公平。如果交给市场，最好的人工智能工具将流向能负担得起它们的学校，从而拉大富裕与贫困学区之间、有资源的国家与没有资源的国家之间的差距。公共政策应发挥作用，确保人工智能的益处能惠及最受益的学生。这可能要求公共投资、开源工具，以及优先考虑可及性而非利润的伙伴关系。</p> <p>第三项优先是教师支持。采用在课堂上成或败，而课堂由教师掌管。政策必须为培训提供资金、为试验留出时间，并尊重专业判断。一个被丢下一个工具便不管的教师，是一个会拒绝它的教师。一个被支持、被培训、被信任的教师，则可能让它发生转变。</p> <p>第四项优先是独立评估。政府和学校不应依赖供应商来证明它们自家产品的有效性。应有一个位置留给独立、严格、透明的评估，在这些系统扩展到数百万儿童之前，就询问关于有效性、安全性和公平性的硬问题。</p> <p>最后，政策必须保持敏捷。技术演化的速度快于法律跟进的步伐。监管者需要能够在不牺牲稳定所需保护的前提下应对变化的机制。这是一个艰难的平衡，而无论向哪个方向失准都代价高昂——太慢则造成伤害；太快则扼杀创新。</p> <p>现实的目标，不是完美地解决所有这些张力，而是审慎地、透明地、以学生的利益为中心去管理它们。</p> <h2> 10. 人工智能在教育中的未来 </h2> <p>预测未来是蠢人的差事。尽管如此，我们还是能辨认出一些值得认真对待的趋势。</p> <p>第一个趋势是能力。人工智能系统将继续变得更强大、更流畅、更多才多艺。那些今天看似令人印象深刻的系统，十年后将显得原始。这一点几乎可以肯定。</p> <p>第二个趋势是融合。人工智能将不再是一个独立的特性，而成为教育环境中不可见的一部分。它将存在于学生阅读的内容、参加的评估、获得的反馈里，甚至存在于教室的排课与管理中。问题将不再是人工智能是否在场，而是它被用得有多好。</p> <p>第三个趋势是个性化的深入。未来的系统可能建模的不仅是一个学生知道什么，还有他们如何学习得最好——其认知风格、动机画像、注意力模式。这可能为个体学习者提供非凡的支持。它也可能随之带来对隐私和自主性新的、严重的风险，因为心灵内在的轮廓变成了数据。</p> <p>还有一个可能是对教育目标本身的重新定义。</p> <p>如果人工智能能够完成我们目前教授的许多认知任务——计算、语法、检索，甚至某些形式的分析——那么问题就来了：学校应该教什么？答案可能转向那些独特属于人的东西：批判性思维、创造力、伦理、协作、韧性、善于提出好问题的能力，而不只是回答问题。人工智能或许并不会摧毁这些技能。它可能，通过让常规认知变得廉价，把它们提升到教育的中心。<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> 常规认知 学校所教的 _____________________ ______________________ 计算 XXXX 批判性思维 语法 XXXX 创造力 检索 XXXX 伦理与协作 基础分析 XXXX 韧性与好问题 </code></pre> </div> <p>然而。未来并未被写就。它是由选择造就的。</p> <p>当前这代教育者、技术人员和政策制定者所做的选择，将决定人工智能成为深化机会的力量，还是固化不平等的力量。两个未来都存在。两个未来都是由同一项技术支撑的。将它们区分开来的，是判断——谨慎构建、明智监管、警惕伤害，以及始终把人，而非机器，放在学习的中心。</p> <h2> 11. 结论 </h2> <p>人工智能已经进入教育。它不是将要到来；它已经在这里。</p> <p>本论文论证了：这项技术是真实的，它的承诺是真诚的，它的风险是严重的。它论证了人工智能能够改善学习，但仅仅是在合适条件下。它论证了教师不会被取代，但他们的工作会改变。它论证了由教育中人工智能所引发的最深层问题，不是技术性的，而是人性的。</p> <p>在某种意义上，我们正站在一个岔路口。</p> <p>一条路通向这样的未来：自适应系统托举每一个有能力的头脑，反馈即刻到来，没有学生会因机器的先觉而无声地坠入深渊。这是一个美丽的愿景。</p> <p>另一条路通向这样的未来：最好的工具只属于少数人，孩子被不透明的算法画像和分拣，学习被优化到失去了灵魂。这是一个警示性的愿景。</p> <p>选哪条路，不由技术决定。它由我们决定。</p> <p>教育中人工智能的承诺，是它也许能打破曾追随其前每项教育技术的历史性失望循环——是它也许能兑现让学习围绕学习者弯曲的梦想。风险在于它沦为这个循环中的又一条目，又一个承诺众多、兑现甚少的漂亮物件，甚至更糟：它只只惠及幸运的少数人。</p> <p>两种结局都不是必然。两种都皆有可能。学习的未来——人工智能是让它更宽还是更窄、更丰盈还是更空洞、更公正还是更不公——取决于我们此刻所做的选择。</p> <p>轻率地做出这些选择，将是对我们人性的一种奇怪的背叛。而完全拒绝做出它们，同样是背叛。</p> <h2> 12. 参考文献 </h2> <p>Baker, R. S. (2016). Stupid tutoring systems, intelligent humans. <em>International Journal of Artificial Intelligence in Education</em>, 26(2), 600–614.</p> <p>Cuban, L. (1986). <em>Teachers and Machines: The Classroom Use of Technology Since 1920</em>. Teachers College Press.</p> <p>Kulik, J. A., &amp; Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. <em>Review of Educational Research</em>, 86(1), 42–78.</p> <p>Luckin, R., Holmes, W., Griffiths, M., &amp; Forcier, L. B. (2016). <em>Intelligence Unleashed: An Argument for AI in Education</em>. Pearson.</p> <p>Ma, W., Adesope, O. O., Nesbit, J. C., &amp; Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. <em>Journal of Educational Psychology</em>, 106(4), 901–918.</p> <p>Selwyn, N. (2019). <em>Should Robots Replace Teachers? AI and the Future of Education</em>. Polity Press.</p> <p>Skinner, B. F. (1958). Teaching machines. <em>Science</em>, 128(3330), 969–977.</p> <p>Williamson, B. (2017). <em>Big Data in Education: The Digital Future of Learning, Policy and Practice</em>. Sage.</p> <p>Zawacki-Richter, O., Marín, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. <em>International Journal of Educational Technology in Higher Education</em>, 16(1), 39.</p> ai education learning ogc16 Thu, 27 Aug 2026 12:18:35 +0000 https://dev.to/ogc16/understanding-the-classroom-how-artificial-intelligence-is-rewriting-the-way-we-learn-cn1 https://dev.to/ogc16/understanding-the-classroom-how-artificial-intelligence-is-rewriting-the-way-we-learn-cn1 <h2> Abstract </h2> <p>Artificial intelligence is no longer a distant promise. It is in the classroom now. This thesis examines how AI-powered tools are reshaping education — from personalized tutoring systems that adapt to each learner, to intelligent grading that gives feedback in seconds, to the quiet ethical questions that follow every innovation.</p> <p>The argument here is not that AI will replace teachers. It will not. Instead, this thesis advances a more nuanced claim: that AI, used thoughtfully, can free teachers to do what machines cannot — mentor, inspire, and connect. Yet that promise depends entirely on how we choose to build and deploy these systems.</p> <p>The chapters that follow trace a path from the historical roots of educational technology, through the current landscape of AI tools, into the measurable effects on learning, and finally into the ethical and policy questions that no student, teacher, or policymaker can afford to ignore.</p> <h2> Table of Contents </h2> <ol> <li>Introduction</li> <li>A Brief History of Educational Technology</li> <li>How AI Works in the Classroom</li> <li>Personalized Learning and Adaptive Systems</li> <li>Automated Assessment and Feedback</li> <li>The Role of the Teacher in the Age of AI</li> <li>Measuring What Works: Evidence and Outcomes</li> <li>Ethics, Privacy, and Bias</li> <li>Policy and Implementation</li> <li>The Future of AI in Education</li> <li>Conclusion</li> <li>References</li> </ol> <h2> 1. Introduction </h2> <p>We live in an age of extraordinary tools.</p> <p>The same technology that powers self-driving cars and medical diagnostics now hands a student a tutor that never tires, never judges, and remembers every answer they have ever given. It is an astonishing prospect. And it is already here.</p> <p>This thesis is about what happens when that reality meets the classroom.</p> <p>Education has always been slow to change. Blackboards gave way to whiteboards. Whiteboards gave way to projectors. Projectors gave way to tablets. But beneath each layer of new hardware, the fundamental structure has remained stubbornly familiar: one teacher, many students, a fixed curriculum, and a clock that ticks the same for everyone. Artificial intelligence threatens to break that structure open. It offers the possibility of learning that bends around the learner, rather than forcing the learner to bend around the schedule.</p> <p>The central research question guiding this work is simple to state and difficult to answer: can artificial intelligence make education more effective and more equitable, and if so, under what conditions?</p> <p>To answer it, we must first understand what these systems actually do. We must separate genuine progress from marketing hype. We must confront uncomfortable truths about data, privacy, and the risk that well-intentioned tools might deepen the very inequalities they claim to erase.</p> <p>This thesis proceeds in three movements.</p> <p>First, we build context. We look at where educational technology came from and why previous revolutions fell short of their promises. Second, we examine the present. We explore the concrete ways AI is already operating inside classrooms, and we weigh the evidence for its effects. Third, we look forward. We grapple with the ethics, the policies, and the choices that will determine whether this revolution serves every student or only the privileged few.</p> <p>The stakes are high. Education is the great engine of opportunity, the force that lifts families across generations. If AI makes it stronger, we have gained something incalculable. If AI makes it narrower, we have lost something we may never get back. This thesis is an attempt to understand which future we are building.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> FUTURE ONE FUTURE TWO (opportunity) (inequality) | | | every mind stretched | best tools for the few | no one falls silent | opaque sorting of children | feedback is instant | learning loses its soul | | \____________ ___________/ \/ / \ / YOU \ / decide \ /____________\ </code></pre> </div> <h2> 2. A Brief History of Educational Technology </h2> <p>To understand where we are going, we must first understand where we have been.</p> <p>The story of technology in education is a story of cycles. Time and again, a new invention arrives wrapped in grand promises of transformation. Time and again, it settles into a supporting role — useful, but rarely revolutionary.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> GRAND PROMISE | v HOPE &amp; EUPHORIA | v REALITY SETS IN | v SETTLES INTO A SUPPORTING ROLE &lt;---- the machine does not rule the classroom </code></pre> </div> <p>Consider the radio. When radio broadcasts reached American homes in the 1920s, enthusiasts predicted that every child in the nation would soon be taught by a handful of brilliant lecturers, their voices carried to farmhouse kitchens and city apartments alike. It did not happen. Radio became a supplement, not a replacement.</p> <p>Consider television. In the 1950s and 1960s, instructional television promised to bring the world's best teachers into every living room. It too faded into the background, a niche offering rather than the foundation of a new system.</p> <p>Then came the computer, and with it a new wave of optimism.</p> <p>Programmed instruction, a term coined by the psychologist B. F. Skinner in the 1950s, offered a tantalizing vision: content broken into small steps, each student advancing at their own pace, with immediate feedback at every turn. Skinner's "teaching machines" were mechanical, clumsy, and limited. But the thinking behind them — that learning could be personalized through careful sequencing and constant reinforcement — planted a seed that would grow for decades.</p> <p>The arrival of the personal computer in the 1980s brought computers into schools at scale. Computer-assisted instruction appeared in labs and classrooms across the developed world. Yet the results were, by most measures, modest. Many machines gathered dust. Many software packages went unused.</p> <p>Scholars have offered ways to explain this pattern of hype and disappointment.</p> <p>Larry Cuban, a historian of education at Stanford, is perhaps the most prominent voice. His work documents a recurring truth: schools are remarkably resistant to fundamental change, absorbing new technologies without being transformed by them. Cuban's analysis suggests that technological innovation rarely reshapes schooling; instead, schooling reshapes the technology, bending it to fit existing routines, timetables, and power structures.</p> <p>Into this landscape of tempered expectations arrived artificial intelligence.</p> <p>AI is different from its predecessors in one crucial respect. Earlier technologies were essentially delivery systems — they carried content from a source to a learner. AI, by contrast, can respond. It can adapt. It can converse. It can generate. These are not trivial differences. A textbook cannot see that you are struggling and change its explanation. An intelligent system can. That single capability is why many observers believe AI will break the historical cycle that radio, television, and the personal computer could not.</p> <p>Whether they are right is the question this thesis exists to examine.</p> <h2> 3. How AI Works in the Classroom </h2> <p>It is tempting to think of the AI in a classroom as a single, monolithic thing. It is not.</p> <p>A classroom today is more likely to contain many different AI systems, each doing a different job. Some are visible. Some are not.</p> <p>Consider the learning management system that tracks attendance and grades. Beneath its surface, algorithms quietly predict which students are at risk of dropping out, flagging their names for intervention before the problem becomes obvious. This is AI, though few would call it by that name.</p> <p>Consider the writing assistant that underlines a student's grammar errors and suggests improvements. It uses neural networks trained on millions of examples of human writing. It is AI, embedded so seamlessly that it feels ordinary.</p> <p>Consider the tutoring platform that presents a problem, waits for an answer, and then — based on that single answer — decides what to show next. This is AI in its most direct educational form: a system making pedagogical decisions in real time.</p> <p>There is an important technical point worth stating plainly.</p> <p>The current wave of AI in education is built largely on machine learning, and in particular on deep learning. These systems do not follow hand-written rules. They learn patterns from vast amounts of data. A language model that generates feedback for student essays has not been taught grammar as a set of principles; it has absorbed grammar by reading billions of sentences and internalizing their statistical regularities.</p> <p>This has two profound consequences.</p> <p>First, these systems are remarkably capable. They can produce fluent prose, solve mathematical problems, answer questions across a staggering range of subjects, and engage in extended dialogue that often feels natural.</p> <p>Second, these systems are eerily unfamiliar. Because they learn rather than being programmed, their behavior is not always predictable or interpretable. A system may produce a brilliant explanation one moment and a confident error the next. It has no true understanding in the human sense; it has patterns, and patterns are powerful but not the same as comprehension.</p> <p>In the classroom, these strengths and weaknesses translate into specific realities.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ARTIFICIAL INTELLIGENCE IN THE CLASSROOM ________________________________________________________ | | | Learning Mgmt Sys Writing Assistant | | predicts drop-outs corrects grammar | | | | | | Tutoring Platform Assessment Engine | | adapts each problem scores essays | | | | ALL USE: DATA --&gt; LEARN PATTERNS --&gt; PREDICT | |________________________________________________________| </code></pre> </div> <p>The strengths enable personalized practice, instant feedback, and a degree of individual attention that no single human teacher could provide to thirty students simultaneously. The weaknesses create risks: the possibility of reinforcing errors, the opacity of decision-making, and the ever-present danger that a system will be wrong with the full, unwarranted confidence of a machine that has seen the correct answer once and now believes it always.</p> <p>An honest account of AI in education must hold both truths at once. The technology is genuinely promising. And it is genuinely imperfect.</p> <h2> 4. Personalized Learning and Adaptive Systems </h2> <p>Few phrases in educational technology are invoked with more reverence and less precision than "personalized learning."</p> <p>The core idea, however, is simple and attractive. Every student learns differently. Every student arrives with different prior knowledge, different strengths, different weaknesses, and different rates of progress. A system that can perceive these differences and respond to them individually has the potential to serve each student far better than a one-size-fits-all curriculum.</p> <p>Adaptive learning systems embody this idea in software.</p> <p>These systems typically work in a cycle. They present a problem; they observe the student's response; they update an internal model of that student's knowledge; and they choose the next problem based on that model. If a student struggles with fractions, the system gives them more practice with fractions before moving on. If a student has mastered them, the system advances. The feedback loop is the heart of the design.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> PRESENT OBSERVE UPDATE a problem the response the student model \ / | \ / | \________/______ | | | CHOOSE NEXT PROBLEM &lt;-----------+ </code></pre> </div> <p>The results can be striking in specific contexts.</p> <p>In mathematics, adaptive systems have shown some of the strongest evidence. Studies of systems used for arithmetic practice and algebra instruction have reported meaningful gains in student achievement, particularly for learners who start behind their peers. The mechanism makes intuitive sense: a student who would otherwise fall off the pace and give up — or who would sit bored while classmates catch up — is instead given work pitched precisely at their level. The experience of flow, of work that is neither too easy nor too hard, is preserved.</p> <p>Yet the evidence is not uniformly positive, and the reasons are instructive.</p> <p>First, adaptive systems are only as good as their underlying model of knowledge. A student may solve a problem correctly for the wrong reasons, or incorrectly for the right ones. The system sees only outcomes, not the reasoning behind them. When the goal is procedural skill, this may matter little. When the goal is conceptual understanding, it matters enormously.</p> <p>Second, personalization of pace is not the same as personalization of meaning. A system can adjust the difficulty of problems, but it cannot easily grasp why a student has lost motivation, or recognize a moment of personal struggle, or offer the encouragement that only a caring human can provide. These dimensions of learning matter, and machines remain largely blind to them.</p> <p>Third, there is a risk that personalization narrows rather than expands ambition. A system that optimizes for mastery of a fixed curriculum may keep a struggling student forever practicing the same procedural skills, never exposing them to the richer, more creative aspects of the subject that might truly ignite their interest. Personalization, done badly, can become a cage.</p> <p>The literature on adaptive learning therefore yields a balanced conclusion. The technology is real, and in the right conditions it produces measurable benefits. But it is a tool among tools, not a replacement for thoughtful pedagogy. Its value depends, as always, on the skill with which it is used.</p> <h2> 5. Automated Assessment and Feedback </h2> <p>Grading is among the most time-consuming and least beloved tasks in teaching.</p> <p>A teacher with a hundred students may spend an evening working through a hundred sets of written work. The demands of grading compete directly with the demands of planning, of preparation, of the human interactions that give teaching its meaning. It is little wonder that automation has arrived here with force.</p> <p>Automated assessment comes in several forms.</p> <p>The simplest involves multiple-choice and short-answer questions, where computers have scored responses accurately for decades. These systems are reliable, fast, and unobjectionable in many contexts, though they test a narrow slice of learning and reward recognition over reasoning.</p> <p>More ambitious are the systems designed to evaluate extended writing.</p> <p>These systems use natural language processing to assess essays on dimensions such as clarity, organization, grammar, and adherence to a prompt. On standardized tests, such systems have been shown to produce scores that correlate reasonably with human raters. For low-stakes formative feedback — helping a student see that a paragraph is off-topic, that a thesis is buried, that transitions are missing — the tools can be genuinely useful, offering guidance at a scale no teacher could match.</p> <p>But the risks are substantial, and they deserve scrutiny.</p> <p>A writing system trained on past essays learns what past essays looked like. It can reward formulaic structure, punish stylistic individuality, and be gamed by students who learn to write in whatever pattern the algorithm favors. More troubling, these systems cannot truly read. They detect statistical signals, not meaning. A beautifully argued essay that departs from convention may be scored harshly; a vacuous essay that mirrors the expected form may be scored generously.</p> <p>There is also the question of what automation does to feedback itself.</p> <p>Feedback is most powerful when it is immediate, specific, and encouraging. AI can deliver the first two, and to some extent the third. But feedback is also a relationship. A student who receives a score and a few comments from a machine has received information; a student who discusses their work with a teacher has received something more. The former is efficient. The latter is transformative. A healthy educational system needs both, and it must resist the temptation to let efficiency crowd out transformation.</p> <p>The most sensible path appears to be hybrid.</p> <p>Let machines handle the routine, the repeatable, the parts of assessment that wear teachers down. Reserve for humans the judgment that requires understanding, the encouragement that requires empathy, and the conversations that require presence. This division of labor is not a compromise; it is a design principle.</p> <h2> 6. The Role of the Teacher in the Age of AI </h2> <p>Every conversation about AI in education eventually arrives at the same anxious question: will machines replace teachers?</p> <p>The short answer is no. The longer answer is more interesting.</p> <p>Teaching is not merely the transmission of information. If it were, then yes, the argument for replacement would be strong, because machines are excellent at transmitting information. But teaching is also the cultivation of relationships, the modeling of curiosity, the recognition of a student who is quietly struggling, the celebration of a student who finally understands. These acts require a human presence. No algorithm can look a student in the eye and say, "I believe in you," and mean it.</p> <p>Yet the question, reframed, becomes fair.</p> <p>Teachers will not be replaced. But the work of teaching will change, and it is changing already.</p> <p>The teacher who must spend hours grading could instead spend those hours conferring with students. The teacher who cannot give every student individual attention could, with AI support, come closer to doing so. The teacher who is the sole source of knowledge in a room of learners becomes something else when knowledge is available everywhere: a navigator, a mentor, a coach.</p> <p>This shift is not entirely comfortable. It asks teachers to surrender a measure of control over their own classrooms, to trust systems they did not build and may not fully understand. It demands new skills — the ability to evaluate a tool's quality, to recognize its biases, to intercept its errors. These are not trivial demands.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> TEACHER: BEFORE TEACHER: WITH AI -------------------- -------------------- sole source of know. navigator / coach grades every essay grades high-stakes only same lesson for all directs personalized work sole attention-giver attention focused on real need </code></pre> </div> <p>Some teachers will embrace the change. Others will resist it, and their resistance is not mere stubbornness. It is an understandable wariness of a force they have not chosen and cannot fully control. The wise implementation of AI in education acknowledges this. It does not impose the technology from above and expect gratitude. It works with teachers, trains them, listens to them, and treats them as partners rather than obstacles.</p> <p>There is a further, deeper point about the human role.</p> <p>The most powerful force in a young person's life, after the family, is often a teacher. A teacher can change the trajectory of a life in ways that are difficult to measure and impossible to algorithmize. The arrival of AI does not diminish this power; if anything, it throws it into sharper relief. As machines take over the mechanical tasks of teaching, the genuinely human work — the encouragement, the belief, the relationship — becomes more clearly the center of the profession, not less.</p> <p>The teacher, in other words, is not the casualty of the AI era. The teacher is its most essential resource.</p> <h2> 7. Measuring What Works: Evidence and Outcomes </h2> <p>It is one thing to believe that AI helps students learn. It is another to demonstrate it.</p> <p>The evidence base for AI in education is young, and it is uneven. Some areas have been studied extensively; others rest on promising anecdotes and vendor claims. Serious evaluation demands that we look at what the research actually shows.</p> <p>In mathematics and procedural subjects, the evidence is most encouraging. Meta-analyses of intelligent tutoring systems have reported positive average effects on learning outcomes, with some studies finding gains comparable to those of individual human tutoring. The consistency of these results across multiple studies is a genuine reason for optimism.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> STRENGTH OF EVIDENCE FOR AI IN EDUCATION Math / procedural ████████████████ strong Reading / language ███████████ moderate Writing / comprehension ██████ mixed Creativity / advanced ███ thin </code></pre> </div> <p>In reading and language learning, the picture is more mixed. Tools that provide vocabulary practice and adaptive reading exercises show benefits. Tools that attempt to assess or improve higher-order writing and comprehension show less consistent results, reflecting the difficulty of the underlying task.</p> <p>In fields requiring deep conceptual understanding or creativity — the humanities, the arts, advanced problem-solving — the evidence is thinnest. Here the limitations of current AI, its inability to truly understand, become most salient. Few studies suggest that AI can replace skilled human instruction in these domains, and many suggest that it cannot yet add much beyond what a good teacher already does.</p> <p>Equally important is the question of <em>whom</em> the technology helps.</p> <p>There is a persistent worry, supported by some research, that adaptive tools help the students who need help least and fail to reach those who need help most. Students with well-developed self-regulation, strong literacy, and supportive homes may benefit most from AI-powered learning because they can navigate it independently. Students who lack these advantages may struggle to use the tools effectively, falling further behind. Technology has a tendency to amplify existing inequalities rather than erase them, and AI shows worrying signs of following the same pattern.</p> <p>There is also the challenge of the evidence itself.</p> <p>Much research in this area is funded by the companies that sell the tools. Much of it is conducted in short, artificial conditions rather than real classrooms over long periods. Publication bias — the tendency of journals to publish positive results — further distorts the picture. An honest assessment must treat promising findings with a degree of skepticism and demand replication in authentic settings.</p> <p>The overall conclusion is neither ecstatic nor dismissive.</p> <p>AI can improve learning under the right conditions. Those conditions include sound pedagogy, well-designed tools, trained teachers, and attention to equity. When these elements are present, the gains are real. When they are absent, the technology does little good and may do harm. The variable that determines success is not the sophistication of the algorithm. It is the quality of the environment in which the algorithm operates.</p> <h2> 8. Ethics, Privacy, and Bias </h2> <p>No discussion of AI in education would be complete without confronting the ethical questions that shadow the technology. These are not afterthoughts. They are central.</p> <p>Privacy is the first and most immediate concern.</p> <p>Educational AI is built on data. To personalize learning, a system collects information about what a student knows, how they learn, how fast they progress, where they struggle. Sometimes it collects far more: keystrokes, reading patterns, emotional responses inferred from facial expressions or typing speeds. This is intimate data, gathered from children, often without meaningful consent.</p> <p>The scale of collection is staggering, and the purposes are not always clear. Data collected to improve learning can be used for other ends — to rank schools, to market to students, to build profiles that follow a person long after they leave the classroom. The children being profiled today are not in a position to consent to how their data will be used in twenty years. This asymmetry of power and consequence deserves far more attention than it has received.</p> <p>Then there is bias.</p> <p>AI systems learn from data, and data reflects the world, including its injustices. A system trained on essays from one demographic may systematically undervalue the writing of another. An adaptive system optimized on the performance of well-resourced students may misjudge and underserve students from different backgrounds. Bias in AI is not a bug that occasionally appears; it is a structural feature that appears whenever the underlying data is uneven, and the burden of detecting and correcting it falls not on the data but on the humans who design and deploy the system.</p> <p>There is also a subtler ethical dimension: the risk of surveillance.</p> <p>A classroom in which every action is monitored, every mistake recorded, every pause timed, is a classroom that has changed in character, whether or not anyone intended it to. Students under constant observation learn differently. They take fewer risks. They conform. The very qualities that education is meant to cultivate — curiosity, experimentation, the courage to be wrong — can be quietly suppressed by an environment that watches too closely.</p> <p>None of this means AI must be rejected. But it does mean that its adoption requires deliberate ethical scaffolding: clear rules about data, strong consent protections, independent oversight, and genuine transparency about what these systems do with the information they gather. Ethics is not a feature to be added later. It is a constraint that must be built in from the start.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> THE BALANCE WE MUST STRIKE | ________/ \________ / \ EDUCATIONAL GAIN HUMAN COST personalized learn. loss of privacy immediate feedback bias &amp; profiling every student served learning under surveillance ___________________ ___________________ ^ ^ |______ keep an eye on both young ______| </code></pre> </div> <h2> 9. Policy and Implementation </h2> <p>Good policy is the difference between a tool that serves and a tool that dominates.</p> <p>The trajectory of AI in education is not determined by technology alone. It is shaped by decisions made in legislatures, ministries, school boards, and district offices. These decisions determine what is permitted, what is funded, what is taught, and who is protected.</p> <p>The first policy priority is data protection. Students deserve the same privacy guarantees in digital spaces that they already have in physical ones. Strong rules should govern what data may be collected, who may access it, how long it may be retained, and whether it may be used for any purpose beyond the student's own education. These rules should be enforceable, and they should carry real consequences for violation.</p> <p>The second priority is equity. Left to the market, the best AI tools will flow to the schools that can afford them, widening the gap between wealthy and poor districts, between countries with resources and countries without. Public policy has a role in ensuring that the benefits of AI reach the students who stand to gain the most. This may require public investment, open-source tools, and partnerships that prioritize access over profit.</p> <p>The third priority is teacher support. Adoption succeeds or fails in the classroom, and classrooms are run by teachers. Policies must include funding for training, time for experimentation, and respect for professional judgment. A teacher who is given a tool and abandoned is a teacher who will reject it. A teacher who is supported, trained, and trusted is a teacher who may transform it.</p> <p>The fourth priority is independent evaluation. Government and schools should not rely on vendors to prove that their own products work. There is a place for independent, rigorous, transparent evaluation that asks hard questions about effectiveness, safety, and equity before these systems are scaled to millions of children.</p> <p>Finally, policy must be nimble. The technology is evolving faster than laws can follow. Regulators need mechanisms that can respond to change without sacrificing the protections that stability requires. This is a difficult balance, and getting it wrong in either direction is costly — too slow, and harm is done; too fast, and innovation is smothered.</p> <p>The realistic goal is not to resolve all these tensions perfectly, but to manage them deliberately, transparently, and with the interests of students at the center.</p> <h2> 10. The Future of AI in Education </h2> <p>Predicting the future is a fool's errand. Still, we can discern trends worth taking seriously.</p> <p>The first trend is capability. AI systems will continue to grow more powerful, more fluent, and more versatile. The systems that seem impressive today will seem primitive in a decade. This much is almost certain.</p> <p>The second trend is integration. AI will stop being a distinct feature and become an invisible part of the educational environment. It will live in the content students read, the assessments they take, the feedback they receive, even in how classrooms are scheduled and managed. The question will no longer be whether AI is present, but how well it is used.</p> <p>The third trend is personalization gone deeper. Future systems may model not just what a student knows, but how they learn best — their cognitive style, their motivational profile, their attention patterns. This could yield extraordinary support for individual learners. It could also yield with it new and serious risks to privacy and autonomy, as the inner contours of a mind become data.</p> <p>There is also the possibility of a redefinition of the educational goal itself.</p> <p>If AI can perform many of the cognitive tasks we currently teach — computation, grammar, retrieval, even some forms of analysis — then the question arises: what should schools teach? The answer may shift toward the distinctly human: critical thinking, creativity, ethics, collaboration, resilience, the ability to ask good questions rather than merely answer them. AI might not destroy these skills. It might, by rendering routine cognition cheap, elevate them into the center of education.<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code> ROUTINE COGNITION WHAT SCHOOLS TEACH _____________________ ______________________ computation XXXX critical thinking grammar XXXX creativity retrieval XXXX ethics &amp; collaboration basic analysis XXXX resilience &amp; good questions </code></pre> </div> <p>And yet. The future is not written. It is made by choices.</p> <p>The choices made by the current generation of educators, technologists, and policymakers will determine whether AI becomes a force for deepening opportunity or for hardening inequality. Both futures are available. Both are supported by the same technology. What separates them is judgment — the judgment to build carefully, to regulate wisely, to watch for harm, and to keep the human being, not the machine, at the center of learning.</p> <h2> 11. Conclusion </h2> <p>Artificial intelligence has arrived in education. It is not coming; it is here.</p> <p>This thesis has argued that the technology is real, its promise is genuine, and its risks are serious. It has argued that AI can improve learning, but only under the right conditions. It has argued that teachers will not be replaced, but that their work will change. It has argued that the deepest questions raised by AI in education are not technical but human.</p> <p>We are, in a sense, standing at a fork in the road.</p> <p>One path leads to a future in which adaptive systems stretch every capable mind, in which feedback is immediate, in which no student falls silently through the cracks because a machine noticed first. This is a beautiful vision.</p> <p>The other path leads to a future in which the best tools belong only to the few, in which children are profiled and sorted by opaque algorithms, in which learning is optimized until it loses its soul. This is a cautionary vision.</p> <p>The road is not chosen by technology. It is chosen by us.</p> <p>The promise of AI in education is that it might finally break the historical cycle of disappointment that has followed every educational technology before it — that it might deliver on the dream of learning that bends around the learner. The risk is that it becomes another entry in that cycle, another shiny object that promised much and delivered little, or worse, that it delivers its promise selectively to a fortunate few.</p> <p>Neither outcome is inevitable. Both are possible. The future of learning — whether AI makes it wider or narrower, richer or more hollow, more just or less — lies in the choices we make now.</p> <p>It would be a strange betrayal of our humanity to make those choices carelessly. It would be an equal betrayal to refuse to make them at all.</p> <h2> 12. References </h2> <p>Baker, R. S. (2016). Stupid tutoring systems, intelligent humans. <em>International Journal of Artificial Intelligence in Education</em>, 26(2), 600–614.</p> <p>Cuban, L. (1986). <em>Teachers and Machines: The Classroom Use of Technology Since 1920</em>. Teachers College Press.</p> <p>Kulik, J. A., &amp; Fletcher, J. D. (2016). Effectiveness of intelligent tutoring systems: A meta-analytic review. <em>Review of Educational Research</em>, 86(1), 42–78.</p> <p>Luckin, R., Holmes, W., Griffiths, M., &amp; Forcier, L. B. (2016). <em>Intelligence Unleashed: An Argument for AI in Education</em>. Pearson.</p> <p>Ma, W., Adesope, O. O., Nesbit, J. C., &amp; Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. <em>Journal of Educational Psychology</em>, 106(4), 901–918.</p> <p>Selwyn, N. (2019). <em>Should Robots Replace Teachers? AI and the Future of Education</em>. Polity Press.</p> <p>Skinner, B. F. (1958). Teaching machines. <em>Science</em>, 128(3330), 969–977.</p> <p>Williamson, B. (2017). <em>Big Data in Education: The Digital Future of Learning, Policy and Practice</em>. Sage.</p> <p>Zawacki-Richter, O., Marín, V. I., Bond, M., &amp; Gouverneur, F. (2019). Systematic review of research on artificial intelligence applications in higher education. <em>International Journal of Educational Technology in Higher Education</em>, 16(1), 39.</p> ai learning programming ogc16 Mon, 17 Aug 2026 15:40:14 +0000 https://dev.to/ogc16/the-status-quo-of-ai-in-software-development-2026-17k8 https://dev.to/ogc16/the-status-quo-of-ai-in-software-development-2026-17k8 <h2> Artificial Intelligence in 2026: From Companion to Infrastructure </h2> <p>Artificial Intelligence has moved from being a futuristic concept to an everyday companion in software development. In 2026, the landscape is defined by rapid innovation, fierce competition, and unresolved challenges around governance, sustainability, and labor. Developers today are navigating both unprecedented opportunities and complex risks.</p> <h2> Industry Dominance </h2> <p>Over 90% of notable AI models now originate from industry rather than academia, signaling commercialization as the primary driver of innovation. Research labs continue to contribute breakthroughs, but the pace of deployment is overwhelmingly shaped by corporate priorities, venture capital, and cloud infrastructure.</p> <h2> Geopolitical Competition </h2> <p>The United States leads in model releases and data center infrastructure, while China dominates robotics and research output. This rivalry shapes the pace and direction of AI development. </p> <ul> <li>Europe has carved out a niche in regulation, with the AI Act setting global standards. </li> <li>Emerging economies in Africa and India are focusing on applied AI, building tools for agriculture, education, and healthcare. </li> </ul> <h2> Compute Explosion </h2> <p>Global AI compute capacity has grown more than threefold annually since 2022, powered largely by Nvidia GPUs. Data centers now consume nearly 30 GW of electricity — comparable to the peak demand of New York City. This raises urgent questions about sustainability and the environmental cost of progress.</p> <h2> The ChatGPT Moment </h2> <p>Artificial Intelligence has had many waves, but the one that truly captured global attention was the release of ChatGPT. What began as a conversational model quickly became a cultural phenomenon, reshaping how people interact with technology, learn, and even work. </p> <ul> <li> <strong>Disruption</strong>: It challenged traditional search engines, productivity tools, and educational practices. </li> <li> <strong>Social Acceptance</strong>: Within months, it was integrated into classrooms, offices, and personal devices. AI was no longer niche — it was part of everyday life. </li> </ul> <p>This acceptance marked a turning point: AI became a daily companion rather than a distant technology.</p> <h2> Competitors and Successors </h2> <p>The success of ChatGPT opened the floodgates for competitors, each pushing boundaries in unique ways: </p> <ul> <li> <strong>DeepSeek</strong>: Efficiency and cost‑effectiveness, democratizing access for startups and individuals. </li> <li> <strong>Claude (Anthropic)</strong>: Safety and alignment, trusted by enterprises and governments. </li> <li> <strong>OpenClaw</strong>: Multimodal capabilities, handling text, images, and structured data seamlessly. </li> </ul> <p>Together, they signal that the AI race is far from over — innovation is accelerating across domains.</p> <h2> Governance and Ethics </h2> <p>Despite progress, unresolved challenges loom large: </p> <ul> <li> <strong>Governance</strong>: Who sets the rules for safe deployment? </li> <li> <strong>Labor impact</strong>: Automation reshapes jobs, but reskilling remains uneven. </li> <li> <strong>Bias and trust</strong>: Ensuring fairness and reliability in outputs is still a work in progress. </li> </ul> <h2> Future Outlook </h2> <p>Artificial Intelligence in 2026 is no longer a tool but a partner. From classrooms to boardrooms, it shapes decisions, creativity, and governance. Yet the story is unfinished: </p> <ul> <li> <strong>AI as infrastructure</strong>: Becoming as essential as electricity or the internet. </li> <li> <strong>Hybrid intelligence</strong>: Humans and AI collaborating rather than competing. </li> <li> <strong>Next disruption</strong>: Multimodal reasoning, embodied AI, or quantum‑accelerated models. </li> </ul> <p><em>Artificial Intelligence has entered everyday life, but its trajectory will be defined not just by innovation, but by how societies govern, sustain, and integrate it into human progress.</em><br> ## Share Your Experience</p> <p>Artificial Intelligence is now part of everyday life, but usage varies widely.<br><br> Which AI do you rely on most in your daily work or personal projects?</p> <ul> <li> <a>ChatGPT</a> </li> <li> <a>Claude</a> </li> <li> <a>DeepSeek</a> </li> <li> <a>OpenClaw</a> </li> <li><a>Other AI tools</a></li> </ul> <p>Comment below or update your own README to let others know how you’re using AI day‑to‑day.</p> ai deepseek openclaw claude ogc16 Fri, 14 Aug 2026 13:08:56 +0000 https://dev.to/ogc16/deploying-iaas-platform-a-production-playbook-421f https://dev.to/ogc16/deploying-iaas-platform-a-production-playbook-421f <p><em>Learn how to go from <code>docker compose up</code> to production-grade deployments on Railway, Kubernetes, and beyond.</em></p> <blockquote> <p><strong>Follow-up to:</strong> <a href="https://dev.to/your-handle/building-multi-tenant-iaas-go">Building a Multi-Tenant IaaS Platform in Go</a></p> </blockquote> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fdeployment-hero.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fdeployment-hero.png" alt="deployment strategies" width="800" height="400"></a></p> <h2> Overview </h2> <p>Last week's post showed you the architecture of a production-ready IaaS platform. Today, we're going practical: <strong>how to actually run it in production</strong>.</p> <p>IaaS Platform ships with multiple deployment options because different teams have different constraints:</p> <ul> <li> <strong>Local dev</strong> — <code>docker compose</code> (5 seconds to running)</li> <li> <strong>Single host</strong> — systemd + PostgreSQL (small team, Hetzner/Linode)</li> <li> <strong>Containers</strong> — Docker image + Railway/Render (zero-ops, scales easily)</li> <li> <strong>Kubernetes</strong> — StatefulSet + Secrets (enterprise teams)</li> <li> <strong>Infrastructure-as-Code</strong> — Terraform (repeatable, auditable)</li> </ul> <p>This post walks through each, with real configs you can copy-paste.</p> <h2> Part 1: Local Development (5 Minutes) </h2> <p>Start here. No thinking required.</p> <h3> The Fastest Possible Start </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>git clone https://github.com/ogc16/iaas-platform.git <span class="nb">cd </span>iaas-platform docker compose up <span class="nt">-d</span> go run ./cmd/server </code></pre> </div> <p><strong>What just happened:</strong></p> <ul> <li>PostgreSQL 16 started in a Docker container (listening on <code>localhost:5432</code>)</li> <li>Server migrations ran automatically</li> <li>Control plane listening on <code>http://localhost:8080</code> </li> <li>Dashboard loaded at <code>/</code> </li> <li>API ready at <code>/api/v1</code> </li> </ul> <h3> What's Inside <code>docker-compose.yml</code> </h3> <div class="highlight js-code-highlight"> <pre class="highlight yaml"><code><span class="na">services</span><span class="pi">:</span> <span class="na">postgres</span><span class="pi">:</span> <span class="na">image</span><span class="pi">:</span> <span class="s">postgres:16-alpine</span> <span class="na">ports</span><span class="pi">:</span> <span class="pi">-</span> <span class="s2">"</span><span class="s">5432:5432"</span> <span class="na">environment</span><span class="pi">:</span> <span class="na">POSTGRES_USER</span><span class="pi">:</span> <span class="s">iaas</span> <span class="na">POSTGRES_PASSWORD</span><span class="pi">:</span> <span class="s">iaas</span> <span class="na">POSTGRES_DB</span><span class="pi">:</span> <span class="s">iaas</span> <span class="na">volumes</span><span class="pi">:</span> <span class="pi">-</span> <span class="s">pgdata:/var/lib/postgresql/data</span> <span class="na">healthcheck</span><span class="pi">:</span> <span class="na">test</span><span class="pi">:</span> <span class="pi">[</span><span class="s2">"</span><span class="s">CMD-SHELL"</span><span class="pi">,</span> <span class="s2">"</span><span class="s">pg_isready</span><span class="nv"> </span><span class="s">-U</span><span class="nv"> </span><span class="s">iaas</span><span class="nv"> </span><span class="s">-d</span><span class="nv"> </span><span class="s">iaas"</span><span class="pi">]</span> <span class="na">interval</span><span class="pi">:</span> <span class="s">5s</span> <span class="na">timeout</span><span class="pi">:</span> <span class="s">5s</span> <span class="na">retries</span><span class="pi">:</span> <span class="m">5</span> </code></pre> </div> <p><strong>Key takeaways:</strong></p> <ul> <li>Alpine Linux keeps the image tiny (~150 MB)</li> <li>Health checks ensure readiness probes pass</li> <li>Volume persists data across container restarts</li> <li>Default credentials are fine for local dev (but NOT production)</li> </ul> <h2> Part 2: Single-Host Deployment (1 Hour Setup) </h2> <p>You're ready to show this to early users. Use a cheap VPS: <strong>$5/month on Linode or Hetzner</strong>.</p> <h3> Step 1: Provision the Host </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># SSH into a fresh Ubuntu 22.04 server</span> ssh root@your-server-ip <span class="c"># Update everything</span> apt-get update <span class="o">&amp;&amp;</span> apt-get upgrade <span class="nt">-y</span> <span class="c"># Install Docker</span> curl <span class="nt">-fsSL</span> https://get.docker.com <span class="nt">-o</span> get-docker.sh sh get-docker.sh <span class="c"># Install PostgreSQL (system package, NOT container)</span> apt-get <span class="nb">install</span> <span class="nt">-y</span> postgresql postgresql-contrib <span class="c"># Create DB user for the app</span> <span class="nb">sudo</span> <span class="nt">-u</span> postgres psql <span class="o">&lt;&lt;</span><span class="no">EOF</span><span class="sh"> CREATE USER iaas WITH PASSWORD 'YOUR_SECURE_PASSWORD_HERE'; CREATE DATABASE iaas OWNER iaas; GRANT ALL PRIVILEGES ON DATABASE iaas TO iaas; </span><span class="no">EOF </span><span class="c"># Verify connection</span> psql <span class="nt">-h</span> localhost <span class="nt">-U</span> iaas <span class="nt">-d</span> iaas <span class="nt">-c</span> <span class="s2">"</span><span class="se">\d</span><span class="s2">t"</span> </code></pre> </div> <h3> Step 2: Deploy the Binary </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Download v0.1.0 release binary</span> wget https://github.com/ogc16/iaas-platform/releases/download/v0.1.0/iaas-platform-server-linux-amd64 <span class="nb">chmod</span> +x iaas-platform-server-linux-amd64 <span class="c"># Create a user to run the service</span> useradd <span class="nt">-r</span> <span class="nt">-s</span> /bin/false iaas-platform <span class="c"># Move binary to a standard location</span> <span class="nb">mkdir</span> <span class="nt">-p</span> /opt/iaas-platform/bin <span class="nb">mv </span>iaas-platform-server-linux-amd64 /opt/iaas-platform/bin/server <span class="nb">chown</span> <span class="nt">-R</span> iaas-platform:iaas-platform /opt/iaas-platform </code></pre> </div> <h3> Step 3: Create a systemd Service </h3> <p>Save this to <code>/etc/systemd/system/iaas-platform.service</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight ini"><code><span class="nn">[Unit]</span> <span class="py">Description</span><span class="p">=</span><span class="s">IaaS Platform Control Plane</span> <span class="py">After</span><span class="p">=</span><span class="s">network-online.target postgresql.service</span> <span class="py">Wants</span><span class="p">=</span><span class="s">network-online.target</span> <span class="nn">[Service]</span> <span class="py">Type</span><span class="p">=</span><span class="s">simple</span> <span class="py">User</span><span class="p">=</span><span class="s">iaas-platform</span> <span class="py">Group</span><span class="p">=</span><span class="s">iaas-platform</span> <span class="c"># Load environment variables from a file </span><span class="py">EnvironmentFile</span><span class="p">=</span><span class="s">/etc/iaas-platform/env</span> <span class="py">ExecStart</span><span class="p">=</span><span class="s">/opt/iaas-platform/bin/server</span> <span class="c"># Auto-restart if the process crashes </span><span class="py">Restart</span><span class="p">=</span><span class="s">on-failure</span> <span class="py">RestartSec</span><span class="p">=</span><span class="s">5</span> <span class="c"># Limits </span><span class="py">LimitNOFILE</span><span class="p">=</span><span class="s">65536</span> <span class="py">TimeoutStopSec</span><span class="p">=</span><span class="s">30</span> <span class="nn">[Install]</span> <span class="py">WantedBy</span><span class="p">=</span><span class="s">multi-user.target</span> </code></pre> </div> <h3> Step 4: Configure Environment Variables </h3> <p>Create <code>/etc/iaas-platform/env</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Production hardening</span> <span class="nv">ENV</span><span class="o">=</span>production <span class="nv">PORT</span><span class="o">=</span>8080 <span class="c"># Database (use the user/password from Step 1)</span> <span class="nv">DATABASE_URL</span><span class="o">=</span>postgres://iaas:YOUR_SECURE_PASSWORD_HERE@localhost:5432/iaas?sslmode<span class="o">=</span>require <span class="c"># Generate a strong secret: openssl rand -hex 32</span> <span class="nv">JWT_SECRET</span><span class="o">=</span>xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx <span class="c"># Password hashing</span> <span class="nv">BCRYPT_COST</span><span class="o">=</span>12 <span class="c"># Email (optional; without SMTP, password-reset links are logged to stdout)</span> <span class="nv">SMTP_HOST</span><span class="o">=</span>smtp.sendgrid.net <span class="nv">SMTP_PORT</span><span class="o">=</span>587 <span class="nv">SMTP_USERNAME</span><span class="o">=</span>apikey <span class="nv">SMTP_PASSWORD</span><span class="o">=</span>SG.xxxxxxxxxxxxx <span class="nv">SMTP_FROM</span><span class="o">=</span>noreply@yourdomain.com <span class="c"># Base URL for password-reset links</span> <span class="nv">APP_BASE_URL</span><span class="o">=</span>https://iaas.yourdomain.com <span class="c"># Restrict read access</span> <span class="nb">chmod </span>600 /etc/iaas-platform/env </code></pre> </div> <h3> Step 5: Enable and Start </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>systemctl daemon-reload systemctl <span class="nb">enable </span>iaas-platform systemctl start iaas-platform systemctl status iaas-platform <span class="c"># Verify it's running</span> </code></pre> </div> <h3> Step 6: Nginx Reverse Proxy </h3> <p>Put this in <code>/etc/nginx/sites-available/iaas</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight nginx"><code><span class="k">upstream</span> <span class="s">iaas_platform</span> <span class="p">{</span> <span class="kn">server</span> <span class="nf">127.0.0.1</span><span class="p">:</span><span class="mi">8080</span><span class="p">;</span> <span class="p">}</span> <span class="c1"># HTTP → HTTPS redirect</span> <span class="k">server</span> <span class="p">{</span> <span class="kn">listen</span> <span class="mi">80</span><span class="p">;</span> <span class="kn">server_name</span> <span class="s">iaas.yourdomain.com</span><span class="p">;</span> <span class="kn">return</span> <span class="mi">301</span> <span class="s">https://</span><span class="nv">$server_name$request_uri</span><span class="p">;</span> <span class="p">}</span> <span class="c1"># HTTPS server</span> <span class="k">server</span> <span class="p">{</span> <span class="kn">listen</span> <span class="mi">443</span> <span class="s">ssl</span> <span class="s">http2</span><span class="p">;</span> <span class="kn">server_name</span> <span class="s">iaas.yourdomain.com</span><span class="p">;</span> <span class="c1"># Use Let's Encrypt (or your CA)</span> <span class="kn">ssl_certificate</span> <span class="n">/etc/letsencrypt/live/iaas.yourdomain.com/fullchain.pem</span><span class="p">;</span> <span class="kn">ssl_certificate_key</span> <span class="n">/etc/letsencrypt/live/iaas.yourdomain.com/privkey.pem</span><span class="p">;</span> <span class="c1"># Security headers (some added by our app, but reinforce here)</span> <span class="kn">add_header</span> <span class="s">Strict-Transport-Security</span> <span class="s">"max-age=63072000</span><span class="p">;</span> <span class="kn">includeSubDomains</span><span class="p">;</span> <span class="kn">preload"</span> <span class="s">always</span><span class="p">;</span> <span class="kn">add_header</span> <span class="s">X-Content-Type-Options</span> <span class="s">"nosniff"</span> <span class="s">always</span><span class="p">;</span> <span class="kn">add_header</span> <span class="s">X-Frame-Options</span> <span class="s">"DENY"</span> <span class="s">always</span><span class="p">;</span> <span class="kn">add_header</span> <span class="s">Referrer-Policy</span> <span class="s">"no-referrer"</span> <span class="s">always</span><span class="p">;</span> <span class="c1"># Request body size limit (prevents DoS; issue #13 will add this to the app)</span> <span class="kn">client_max_body_size</span> <span class="mi">1M</span><span class="p">;</span> <span class="c1"># Proxy to our app</span> <span class="kn">location</span> <span class="n">/</span> <span class="p">{</span> <span class="kn">proxy_pass</span> <span class="s">http://iaas_platform</span><span class="p">;</span> <span class="kn">proxy_set_header</span> <span class="s">Host</span> <span class="nv">$host</span><span class="p">;</span> <span class="kn">proxy_set_header</span> <span class="s">X-Real-IP</span> <span class="nv">$remote_addr</span><span class="p">;</span> <span class="kn">proxy_set_header</span> <span class="s">X-Forwarded-For</span> <span class="nv">$proxy_add_x_forwarded_for</span><span class="p">;</span> <span class="kn">proxy_set_header</span> <span class="s">X-Forwarded-Proto</span> <span class="nv">$scheme</span><span class="p">;</span> <span class="c1"># Timeouts for long-polling (health checks, etc.)</span> <span class="kn">proxy_read_timeout</span> <span class="s">30s</span><span class="p">;</span> <span class="kn">proxy_connect_timeout</span> <span class="s">10s</span><span class="p">;</span> <span class="p">}</span> <span class="p">}</span> </code></pre> </div> <h3> Step 7: Get an SSL Certificate </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>apt-get <span class="nb">install</span> <span class="nt">-y</span> certbot python3-certbot-nginx certbot certonly <span class="nt">--nginx</span> <span class="nt">-d</span> iaas.yourdomain.com <span class="c"># Auto-renew</span> systemctl <span class="nb">enable </span>certbot.timer </code></pre> </div> <h3> Step 8: Verify It Works </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Check the app is running</span> curl https://iaas.yourdomain.com/healthz <span class="c"># Should return: OK</span> <span class="c"># Check readiness (pings the database)</span> curl https://iaas.yourdomain.com/readyz <span class="c"># Should return: OK</span> <span class="c"># Open in browser</span> open https://iaas.yourdomain.com </code></pre> </div> <p><strong>Cost:</strong> ~$5/month (VPS) + ~$10/year (domain) + free SSL (Let's Encrypt).</p> <h2> Part 3: Container Deployment (No Server Management) </h2> <p>Use <strong>Railway</strong> or <strong>Render</strong> for zero-ops deployment. We'll show Railway because it's fast.</p> <h3> Step 1: Containerize the App </h3> <p>The repo includes a multi-stage <code>Dockerfile</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight docker"><code><span class="c"># Build stage</span> <span class="k">FROM</span><span class="w"> </span><span class="s">golang:1.26-alpine</span><span class="w"> </span><span class="k">AS</span><span class="w"> </span><span class="s">build</span> <span class="k">WORKDIR</span><span class="s"> /src</span> <span class="k">COPY</span><span class="s"> go.mod go.sum ./</span> <span class="k">RUN </span>go mod download <span class="k">COPY</span><span class="s"> . .</span> <span class="k">RUN </span><span class="nv">CGO_ENABLED</span><span class="o">=</span>0 <span class="nv">GOOS</span><span class="o">=</span>linux go build <span class="nt">-trimpath</span> <span class="nt">-ldflags</span><span class="o">=</span><span class="s2">"-s -w"</span> <span class="nt">-o</span> /out/server ./cmd/server <span class="c"># Runtime stage (distroless = minimal, no shell, no package manager)</span> <span class="k">FROM</span><span class="s"> gcr.io/distroless/static-debian12:nonroot</span> <span class="k">COPY</span><span class="s"> --from=build /out/server /usr/local/bin/</span> <span class="k">EXPOSE</span><span class="s"> 8080</span> <span class="k">ENTRYPOINT</span><span class="s"> ["/usr/local/bin/server"]</span> </code></pre> </div> <p><strong>Why distroless?</strong></p> <ul> <li>✅ Image size: ~30 MB (not 800 MB with Alpine)</li> <li>✅ No shell or package manager = smaller attack surface</li> <li>✅ Runs as non-root by default (<code>uid 65532</code>)</li> <li>✅ Security scanning doesn't find phantom vulnerabilities</li> </ul> <h3> Step 2: Push to Container Registry </h3> <p>Option A: GitHub Container Registry (free, no auth needed for public repos):<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>docker build <span class="nt">-t</span> ghcr.io/YOUR_GITHUB_USERNAME/iaas-platform:latest <span class="nb">.</span> docker push ghcr.io/YOUR_GITHUB_USERNAME/iaas-platform:latest </code></pre> </div> <p>Option B: Docker Hub:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>docker build <span class="nt">-t</span> YOUR_DOCKER_USERNAME/iaas-platform:latest <span class="nb">.</span> docker push YOUR_DOCKER_USERNAME/iaas-platform:latest </code></pre> </div> <h3> Step 3: Deploy to Railway </h3> <ol> <li>Go to <a href="https://railway.app" rel="noopener noreferrer">railway.app</a> </li> <li>Click <strong>"New Project"</strong> → <strong>"Deploy from GitHub"</strong> </li> <li>Select your <code>iaas-platform</code> fork</li> <li>Railway auto-detects the Dockerfile</li> <li> <strong>Add PostgreSQL 16</strong> plugin</li> <li>Set environment variables: </li> </ol> <div class="highlight js-code-highlight"> <pre class="highlight conf"><code> <span class="n">ENV</span>=<span class="n">production</span> <span class="n">JWT_SECRET</span>=&lt;<span class="n">generate</span>: <span class="n">openssl</span> <span class="n">rand</span> -<span class="n">hex</span> <span class="m">32</span>&gt; <span class="n">DATABASE_URL</span>=&lt;<span class="n">Railway</span> <span class="n">generates</span> <span class="n">this</span> <span class="n">automatically</span>&gt; <span class="n">BCRYPT_COST</span>=<span class="m">12</span> </code></pre> </div> <ol> <li>Deploy button → <strong>Wait 2 minutes</strong> </li> </ol> <p><strong>That's it.</strong> Railway provides:</p> <ul> <li>✅ HTTPS by default (auto-generated domain)</li> <li>✅ PostgreSQL managed for you</li> <li>✅ Auto-scaling</li> <li>✅ $5/month free tier (then $5/gb RAM/month)</li> </ul> <p><strong>Your app is now live at:</strong> <code>https://yourdomain-production-xxxx.railway.app</code></p> <h2> Part 4: Kubernetes Deployment (Enterprise) </h2> <p>For large teams, Kubernetes gives you <strong>orchestration, auto-scaling, and self-healing</strong>.</p> <h3> Prerequisites </h3> <ul> <li>A Kubernetes cluster (EKS, GKE, DigitalOcean, or local <code>minikube</code>)</li> <li> <code>kubectl</code> CLI</li> <li> <code>helm</code> (optional but recommended)</li> </ul> <h3> Step 1: Create Secrets </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Generate a strong JWT secret</span> <span class="nv">JWT_SECRET</span><span class="o">=</span><span class="si">$(</span>openssl rand <span class="nt">-hex</span> 32<span class="si">)</span> <span class="c"># Create a Kubernetes secret</span> kubectl create secret generic iaas-platform-secrets <span class="se">\</span> <span class="nt">--from-literal</span><span class="o">=</span><span class="nv">JWT_SECRET</span><span class="o">=</span><span class="nv">$JWT_SECRET</span> <span class="se">\</span> <span class="nt">--from-literal</span><span class="o">=</span><span class="nv">DATABASE_URL</span><span class="o">=</span>postgres://iaas:password@postgres-svc:5432/iaas <span class="se">\</span> <span class="nt">--from-literal</span><span class="o">=</span><span class="nv">BCRYPT_COST</span><span class="o">=</span>12 <span class="se">\</span> <span class="nt">--from-literal</span><span class="o">=</span><span class="nv">ENV</span><span class="o">=</span>production </code></pre> </div> <h3> Step 2: Deploy PostgreSQL (if not using managed RDS) </h3> <p>Create <code>postgres.yaml</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight yaml"><code><span class="na">apiVersion</span><span class="pi">:</span> <span class="s">v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">ConfigMap</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres-config</span> <span class="na">data</span><span class="pi">:</span> <span class="na">POSTGRES_USER</span><span class="pi">:</span> <span class="s">iaas</span> <span class="na">POSTGRES_DB</span><span class="pi">:</span> <span class="s">iaas</span> <span class="nn">---</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">PersistentVolumeClaim</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres-pvc</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">accessModes</span><span class="pi">:</span> <span class="pi">-</span> <span class="s">ReadWriteOnce</span> <span class="na">resources</span><span class="pi">:</span> <span class="na">requests</span><span class="pi">:</span> <span class="na">storage</span><span class="pi">:</span> <span class="s">10Gi</span> <span class="nn">---</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">apps/v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">Deployment</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">replicas</span><span class="pi">:</span> <span class="m">1</span> <span class="na">selector</span><span class="pi">:</span> <span class="na">matchLabels</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">postgres</span> <span class="pi">}</span> <span class="na">template</span><span class="pi">:</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">labels</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">postgres</span> <span class="pi">}</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">containers</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres</span> <span class="na">image</span><span class="pi">:</span> <span class="s">postgres:16-alpine</span> <span class="na">envFrom</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">configMapRef</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres-config</span> <span class="na">env</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">name</span><span class="pi">:</span> <span class="s">POSTGRES_PASSWORD</span> <span class="na">valueFrom</span><span class="pi">:</span> <span class="na">secretKeyRef</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres-secrets</span> <span class="na">key</span><span class="pi">:</span> <span class="s">password</span> <span class="na">ports</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">containerPort</span><span class="pi">:</span> <span class="m">5432</span> <span class="na">volumeMounts</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">name</span><span class="pi">:</span> <span class="s">data</span> <span class="na">mountPath</span><span class="pi">:</span> <span class="s">/var/lib/postgresql/data</span> <span class="na">livenessProbe</span><span class="pi">:</span> <span class="na">exec</span><span class="pi">:</span> <span class="na">command</span><span class="pi">:</span> <span class="pi">[</span><span class="s2">"</span><span class="s">pg_isready"</span><span class="pi">,</span> <span class="s2">"</span><span class="s">-U"</span><span class="pi">,</span> <span class="s2">"</span><span class="s">iaas"</span><span class="pi">]</span> <span class="na">initialDelaySeconds</span><span class="pi">:</span> <span class="m">30</span> <span class="na">periodSeconds</span><span class="pi">:</span> <span class="m">10</span> <span class="na">volumes</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">name</span><span class="pi">:</span> <span class="s">data</span> <span class="na">persistentVolumeClaim</span><span class="pi">:</span> <span class="na">claimName</span><span class="pi">:</span> <span class="s">postgres-pvc</span> <span class="nn">---</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">Service</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">postgres-svc</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">selector</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">postgres</span> <span class="pi">}</span> <span class="na">ports</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">port</span><span class="pi">:</span> <span class="m">5432</span> <span class="na">targetPort</span><span class="pi">:</span> <span class="m">5432</span> </code></pre> </div> <h3> Step 3: Deploy the App </h3> <p>Create <code>iaas-platform.yaml</code>:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight yaml"><code><span class="na">apiVersion</span><span class="pi">:</span> <span class="s">apps/v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">Deployment</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">iaas-platform</span> <span class="na">labels</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">iaas-platform</span> <span class="pi">}</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">replicas</span><span class="pi">:</span> <span class="m">2</span> <span class="na">selector</span><span class="pi">:</span> <span class="na">matchLabels</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">iaas-platform</span> <span class="pi">}</span> <span class="na">template</span><span class="pi">:</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">labels</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">iaas-platform</span> <span class="pi">}</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">securityContext</span><span class="pi">:</span> <span class="na">runAsNonRoot</span><span class="pi">:</span> <span class="kc">true</span> <span class="na">runAsUser</span><span class="pi">:</span> <span class="m">65532</span> <span class="na">containers</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">name</span><span class="pi">:</span> <span class="s">server</span> <span class="na">image</span><span class="pi">:</span> <span class="s">ghcr.io/YOUR_GITHUB_USERNAME/iaas-platform:latest</span> <span class="na">imagePullPolicy</span><span class="pi">:</span> <span class="s">Always</span> <span class="na">ports</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">containerPort</span><span class="pi">:</span> <span class="m">8080</span> <span class="na">name</span><span class="pi">:</span> <span class="s">http</span> <span class="na">envFrom</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">secretRef</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">iaas-platform-secrets</span> <span class="na">resources</span><span class="pi">:</span> <span class="na">requests</span><span class="pi">:</span> <span class="na">memory</span><span class="pi">:</span> <span class="s2">"</span><span class="s">256Mi"</span> <span class="na">cpu</span><span class="pi">:</span> <span class="s2">"</span><span class="s">250m"</span> <span class="na">limits</span><span class="pi">:</span> <span class="na">memory</span><span class="pi">:</span> <span class="s2">"</span><span class="s">512Mi"</span> <span class="na">cpu</span><span class="pi">:</span> <span class="s2">"</span><span class="s">500m"</span> <span class="na">livenessProbe</span><span class="pi">:</span> <span class="na">httpGet</span><span class="pi">:</span> <span class="na">path</span><span class="pi">:</span> <span class="s">/healthz</span> <span class="na">port</span><span class="pi">:</span> <span class="m">8080</span> <span class="na">initialDelaySeconds</span><span class="pi">:</span> <span class="m">10</span> <span class="na">periodSeconds</span><span class="pi">:</span> <span class="m">10</span> <span class="na">readinessProbe</span><span class="pi">:</span> <span class="na">httpGet</span><span class="pi">:</span> <span class="na">path</span><span class="pi">:</span> <span class="s">/readyz</span> <span class="na">port</span><span class="pi">:</span> <span class="m">8080</span> <span class="na">initialDelaySeconds</span><span class="pi">:</span> <span class="m">5</span> <span class="na">periodSeconds</span><span class="pi">:</span> <span class="m">5</span> <span class="nn">---</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">Service</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">iaas-platform-svc</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">selector</span><span class="pi">:</span> <span class="pi">{</span> <span class="nv">app</span><span class="pi">:</span> <span class="nv">iaas-platform</span> <span class="pi">}</span> <span class="na">type</span><span class="pi">:</span> <span class="s">LoadBalancer</span> <span class="na">ports</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">port</span><span class="pi">:</span> <span class="m">80</span> <span class="na">targetPort</span><span class="pi">:</span> <span class="m">8080</span> <span class="na">protocol</span><span class="pi">:</span> <span class="s">TCP</span> <span class="nn">---</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">autoscaling/v2</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">HorizontalPodAutoscaler</span> <span class="na">metadata</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">iaas-platform-hpa</span> <span class="na">spec</span><span class="pi">:</span> <span class="na">scaleTargetRef</span><span class="pi">:</span> <span class="na">apiVersion</span><span class="pi">:</span> <span class="s">apps/v1</span> <span class="na">kind</span><span class="pi">:</span> <span class="s">Deployment</span> <span class="na">name</span><span class="pi">:</span> <span class="s">iaas-platform</span> <span class="na">minReplicas</span><span class="pi">:</span> <span class="m">2</span> <span class="na">maxReplicas</span><span class="pi">:</span> <span class="m">10</span> <span class="na">metrics</span><span class="pi">:</span> <span class="pi">-</span> <span class="na">type</span><span class="pi">:</span> <span class="s">Resource</span> <span class="na">resource</span><span class="pi">:</span> <span class="na">name</span><span class="pi">:</span> <span class="s">cpu</span> <span class="na">target</span><span class="pi">:</span> <span class="na">type</span><span class="pi">:</span> <span class="s">Utilization</span> <span class="na">averageUtilization</span><span class="pi">:</span> <span class="m">70</span> </code></pre> </div> <h3> Step 4: Apply and Verify </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Create namespace</span> kubectl create namespace iaas <span class="c"># Apply configs</span> kubectl apply <span class="nt">-f</span> postgres.yaml <span class="nt">-n</span> iaas kubectl apply <span class="nt">-f</span> iaas-platform.yaml <span class="nt">-n</span> iaas <span class="c"># Wait for rollout</span> kubectl rollout status deployment/iaas-platform <span class="nt">-n</span> iaas <span class="c"># Get the LoadBalancer IP</span> kubectl get svc iaas-platform-svc <span class="nt">-n</span> iaas <span class="c"># Test it</span> curl http://&lt;LOAD_BALANCER_IP&gt;/healthz </code></pre> </div> <p><strong>Kubernetes gives you:</strong></p> <ul> <li>✅ Auto-scaling (CPU-based HPA)</li> <li>✅ Self-healing (pod restarts on failure)</li> <li>✅ Rolling updates (zero downtime deploys)</li> <li>✅ Multi-region ready (Federated clusters)</li> </ul> <h2> Part 5: Infrastructure-as-Code (Terraform) </h2> <p>For repeatable, auditable infrastructure.</p> <h3> What's Included </h3> <p>The repo has <code>examples/terraform/</code> with configs for <strong>single-host Docker deployment</strong> (great for staging):<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight hcl"><code><span class="c1"># examples/terraform/main.tf</span> <span class="nx">terraform</span> <span class="p">{</span> <span class="nx">required_version</span> <span class="p">=</span> <span class="s2">"&gt;= 1.5"</span> <span class="nx">required_providers</span> <span class="p">{</span> <span class="nx">docker</span> <span class="p">=</span> <span class="p">{</span> <span class="nx">source</span> <span class="p">=</span> <span class="s2">"kreuzwerker/docker"</span> <span class="nx">version</span> <span class="p">=</span> <span class="s2">"~&gt; 3.0"</span> <span class="p">}</span> <span class="p">}</span> <span class="p">}</span> <span class="nx">provider</span> <span class="s2">"docker"</span> <span class="p">{}</span> <span class="c1"># PostgreSQL container</span> <span class="nx">resource</span> <span class="s2">"docker_container"</span> <span class="s2">"postgres"</span> <span class="p">{</span> <span class="nx">name</span> <span class="p">=</span> <span class="s2">"iaas-platform-postgres"</span> <span class="nx">image</span> <span class="p">=</span> <span class="nx">docker_image</span><span class="p">.</span><span class="nx">postgres</span><span class="p">.</span><span class="nx">image_id</span> <span class="nx">env</span> <span class="p">=</span> <span class="p">[</span> <span class="s2">"POSTGRES_USER=iaas"</span><span class="p">,</span> <span class="s2">"POSTGRES_PASSWORD=${var.database_password}"</span><span class="p">,</span> <span class="s2">"POSTGRES_DB=iaas"</span><span class="p">,</span> <span class="p">]</span> <span class="nx">ports</span> <span class="p">{</span> <span class="nx">internal</span> <span class="p">=</span> <span class="mi">5432</span> <span class="nx">external</span> <span class="p">=</span> <span class="nx">var</span><span class="p">.</span><span class="nx">database_port</span> <span class="p">}</span> <span class="nx">volumes</span> <span class="p">{</span> <span class="nx">volume_name</span> <span class="p">=</span> <span class="nx">docker_volume</span><span class="p">.</span><span class="nx">postgres_data</span><span class="p">.</span><span class="nx">name</span> <span class="nx">container_path</span> <span class="p">=</span> <span class="s2">"/var/lib/postgresql/data"</span> <span class="p">}</span> <span class="p">}</span> <span class="c1"># App server container</span> <span class="nx">resource</span> <span class="s2">"docker_container"</span> <span class="s2">"server"</span> <span class="p">{</span> <span class="nx">name</span> <span class="p">=</span> <span class="s2">"iaas-platform-server"</span> <span class="nx">image</span> <span class="p">=</span> <span class="nx">docker_image</span><span class="p">.</span><span class="nx">iaas_platform</span><span class="p">.</span><span class="nx">image_id</span> <span class="nx">env</span> <span class="p">=</span> <span class="p">[</span> <span class="s2">"ENV=production"</span><span class="p">,</span> <span class="s2">"HTTP_ADDR=:${var.server_port}"</span><span class="p">,</span> <span class="s2">"DATABASE_URL=postgres://iaas:${var.database_password}@postgres:5432/iaas?sslmode=disable"</span><span class="p">,</span> <span class="s2">"JWT_SECRET=${var.jwt_secret}"</span><span class="p">,</span> <span class="p">]</span> <span class="nx">ports</span> <span class="p">{</span> <span class="nx">internal</span> <span class="p">=</span> <span class="nx">var</span><span class="p">.</span><span class="nx">server_port</span> <span class="nx">external</span> <span class="p">=</span> <span class="nx">var</span><span class="p">.</span><span class="nx">server_port</span> <span class="p">}</span> <span class="nx">depends_on</span> <span class="p">=</span> <span class="p">[</span><span class="nx">docker_container</span><span class="p">.</span><span class="nx">postgres</span><span class="p">]</span> <span class="p">}</span> </code></pre> </div> <h3> Deploy with Terraform </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="nb">cd </span>examples/terraform terraform init terraform apply <span class="nt">-var</span> <span class="nv">server_image_tag</span><span class="o">=</span>v0.1.0 </code></pre> </div> <p><strong>Benefits:</strong></p> <ul> <li>✅ Repeatable (same config → same infrastructure)</li> <li>✅ Auditable (git history of changes)</li> <li>✅ Shareable (colleague runs <code>terraform apply</code>, gets identical setup)</li> <li>✅ Destroyable (tear down with <code>terraform destroy</code>)</li> </ul> <h2> Production Readiness Checklist </h2> <p>Before launching, verify:</p> <h3> Security </h3> <ul> <li>[ ] <code>ENV=production</code> (refuses weak JWT secrets)</li> <li>[ ] <code>JWT_SECRET</code> is ≥32 bytes random (run: <code>openssl rand -hex 32</code>)</li> <li>[ ] TLS enabled (HTTPS only, no HTTP)</li> <li>[ ] Database password strong</li> <li>[ ] SMTP credentials NOT committed to Git (use env or secrets manager)</li> <li>[ ] API key hashing enabled (default)</li> </ul> <h3> Operations </h3> <ul> <li>[ ] Health probes working (<code>/healthz</code>, <code>/readyz</code>)</li> <li>[ ] Logs flowing to stdout (JSON structured via slog)</li> <li>[ ] Backups scheduled (<code>pg_dump</code> daily)</li> <li>[ ] Rollback plan documented</li> </ul> <h3> Testing </h3> <ul> <li>[ ] <code>api-usage.sh</code> works against production</li> <li>[ ] Dashboard loads</li> <li>[ ] Can create org, instance, and generate invoice</li> <li>[ ] Password reset emails working (if SMTP enabled)</li> </ul> <h2> Cost Comparison </h2> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Platform</th> <th>Cost/Month</th> <th>Setup Time</th> <th>Scaling</th> <th>Best For</th> </tr> </thead> <tbody> <tr> <td><strong>Single Host (systemd)</strong></td> <td>$5-15</td> <td>1 hour</td> <td>Manual</td> <td>Early-stage startups, learning</td> </tr> <tr> <td><strong>Railway/Render</strong></td> <td>$5-50</td> <td>10 min</td> <td>Auto</td> <td>Fast iteration, small teams</td> </tr> <tr> <td><strong>Kubernetes (GKE/EKS)</strong></td> <td>$100+</td> <td>2 hours</td> <td>Auto</td> <td>Enterprise, high traffic</td> </tr> <tr> <td><strong>Terraform (staging)</strong></td> <td>$10-50</td> <td>30 min</td> <td>Repeatable</td> <td>CI/CD integration, testing</td> </tr> </tbody> </table></div> <h2> Deployment Workflow (Complete Example) </h2> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># 1. Build and test locally</span> make <span class="nb">test </span>make race docker build <span class="nt">-t</span> iaas-platform:latest <span class="nb">.</span> <span class="c"># 2. Push to registry</span> docker tag iaas-platform:latest ghcr.io/you/iaas-platform:v0.1.1 docker push ghcr.io/you/iaas-platform:v0.1.1 <span class="c"># 3. Deploy to staging (Terraform)</span> <span class="nb">cd </span>examples/terraform terraform apply <span class="nt">-var</span> <span class="nv">server_image_tag</span><span class="o">=</span>v0.1.1 <span class="c"># 4. Smoke test</span> curl http://staging-server/healthz bash examples/api-usage.sh <span class="c"># 5. Deploy to production (Railway or K8s)</span> <span class="c"># Railway: auto-deploys on git push</span> <span class="c"># K8s: kubectl set image deployment/iaas-platform \</span> <span class="c"># server=ghcr.io/you/iaas-platform:v0.1.1</span> <span class="c"># 6. Verify</span> curl https://iaas.yourdomain.com/healthz </code></pre> </div> <h2> Monitoring in Production </h2> <p>IaaS Platform logs structured JSON (via Go's <code>slog</code>). Aggregate with:</p> <ul> <li> <strong>ELK Stack</strong> — Elasticsearch + Logstash + Kibana</li> <li> <strong>Datadog</strong> — SaaS APM (integrates with Kubernetes)</li> <li> <strong>Grafana Loki</strong> — Lightweight log aggregation</li> <li> <strong>CloudWatch</strong> — If using AWS</li> </ul> <p>Every request includes an <code>X-Request-ID</code> for correlation.</p> <p><strong>Missing:</strong> Prometheus metrics. That's issue #14; see <a href="https://github.com/ogc16/iaas-platform/blob/master/ROADMAP.md" rel="noopener noreferrer">ROADMAP.md</a>.</p> <h2> Next Steps </h2> <ol> <li> <strong>Pick your platform</strong> — Start with Railway if unsure</li> <li> <strong>Follow the steps above</strong> — Copy-paste the configs</li> <li> <strong>Test the API</strong> — Run <code>examples/api-usage.sh</code> </li> <li> <strong>Explore the dashboard</strong> — Create an org, spin up an instance</li> <li> <strong>Open an issue</strong> — Something missing? File it on GitHub</li> </ol> <h2> Troubleshooting </h2> <h3> "healthz returns 503" </h3> <p>Database isn't reachable. Check:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># Is PostgreSQL running?</span> docker ps | <span class="nb">grep </span>postgres <span class="c"># Can the app reach it?</span> curl <span class="nv">$DATABASE_URL</span> <span class="c"># Should fail gracefully, not hang</span> </code></pre> </div> <h3> "502 Bad Gateway" </h3> <p>App crashed. Check logs:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># If systemd:</span> journalctl <span class="nt">-u</span> iaas-platform <span class="nt">-f</span> <span class="c"># If Docker:</span> docker logs iaas-platform-server <span class="c"># If Kubernetes:</span> kubectl logs <span class="nt">-f</span> deployment/iaas-platform <span class="nt">-n</span> iaas </code></pre> </div> <h3> "Certificate verification failed" </h3> <p>TLS misconfigured. For development:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>curl <span class="nt">--insecure</span> https://localhost:8080/healthz </code></pre> </div> <p>For production, ensure Let's Encrypt cert is valid:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>openssl x509 <span class="nt">-in</span> /etc/letsencrypt/live/yourdomain.com/fullchain.pem <span class="nt">-text</span> <span class="nt">-noout</span> </code></pre> </div> <h2> What's Next? </h2> <p>This series has covered:</p> <ol> <li> <strong><a href="https://dev.to/...">Part 1</a> - Architecture</strong>: Multi-tenant design, billing engine, async lifecycle</li> <li> <strong><a href="https://dev.to/...">Part 2</a> - Deployment</strong> (this post): Local, single-host, containerized, K8s</li> <li> <strong>Part 3 (coming soon)</strong> - Scaling beyond v0.1.0: Prometheus metrics, webhook notifications, real compute backends</li> </ol> <h2> Join the Community </h2> <p>The IaaS Platform is open source and growing:</p> <ul> <li>⭐ <a href="https://github.com/ogc16/iaas-platform" rel="noopener noreferrer">GitHub</a> </li> <li>💬 <a href="https://github.com/ogc16/iaas-platform/discussions" rel="noopener noreferrer">Discussions</a> </li> <li>🐛 <a href="https://github.com/ogc16/iaas-platform/issues" rel="noopener noreferrer">Open Issues</a> (5 good-first-issues for contributors)</li> </ul> <p><strong>Have deployment questions?</strong> Reply in the comments below or open an issue.</p> <p>Happy deploying! 🚀</p> iaas saas devops docker ogc16 Fri, 14 Aug 2026 12:20:24 +0000 https://dev.to/ogc16/learn-how-to-build-a-cloud-control-panel-with-authentication-billing-and-resource-management--530e https://dev.to/ogc16/learn-how-to-build-a-cloud-control-panel-with-authentication-billing-and-resource-management--530e <h2> Building a Multi-Tenant IaaS Platform in Go: From Zero to Production </h2> <p><em>Learn how to build a cloud control plane with authentication, billing, and resource management — complete with runnable code.</em></p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fyour-image.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fyour-image.png" alt="Hero Image"></a></p> <blockquote> <p><strong>TL;DR:</strong> I built an open-source Infrastructure-as-a-Service platform in Go with multi-tenant orgs, async compute lifecycle, usage-based billing, and production-grade security. It's ~2,000 lines of code, fully tested, and ready to deploy. <a href="https://github.com/ogc16/iaas-platform" rel="noopener noreferrer">Check it out on GitHub</a>.</p> </blockquote> <h2> Why I Built This </h2> <p>A few months ago, I was trying to understand how cloud platforms like AWS, DigitalOcean, and Heroku work under the hood. Specifically:</p> <ul> <li>How do they isolate customer data across multiple organizations?</li> <li>How does the billing engine track usage and generate invoices?</li> <li>How do they manage async compute lifecycle (start/stop/terminate)?</li> <li>What does secure multi-tenant architecture actually look like?</li> </ul> <p>I couldn't find a reference implementation that was both <strong>simple enough to learn from</strong> and <strong>complete enough to be production-like</strong>. So I built one.</p> <p>The result is <strong>IaaS Platform</strong>: a fully-functional control plane that handles authentication, multi-tenancy, compute resource management, and usage-based billing — all in ~2,000 lines of clean, tested Go code.</p> <h2> What Does It Do? </h2> <p>Imagine you're building your own version of AWS or Heroku. You need:</p> <h3> 1️⃣ <strong>Authentication &amp; Tenancy</strong> </h3> <ul> <li>Users sign up with email/password</li> <li>Each user can create multiple organizations</li> <li>Organizations have members with roles (admin/member)</li> <li>Member suspensions with auto-expiry</li> <li>API keys for programmatic access</li> </ul> <h3> 2️⃣ <strong>Compute Lifecycle</strong> </h3> <ul> <li>Users request compute instances</li> <li>Instances move through states asynchronously: <code>pending → running → stopping → stopped</code> </li> <li>Per-organization quotas (max 20 instances, 16 vCPU, 32 GB RAM, 500 GB disk)</li> <li>Per-region capacity enforcement (e.g., us-east-1 has 64 vCPU total)</li> <li>Background reconciler settles transient states</li> </ul> <h3> 3️⃣ <strong>Usage-Based Billing</strong> </h3> <ul> <li>Track CPU hours, memory GB-hours, disk GB-hours</li> <li>Pricing: $0.50/core-hour, $0.20/GB-hour, $0.01/GB-hour</li> <li>Generate invoices monthly with line items</li> <li>Audit-logged every transaction</li> </ul> <h3> 4️⃣ <strong>Production Features</strong> </h3> <ul> <li>JWT + API key authentication (both Bearer and X-API-Key)</li> <li>Rate limiting (token bucket, 1 req/sec per IP/API key)</li> <li>Security headers (CSP, HSTS, X-Frame-Options, etc.)</li> <li>Bcrypt password hashing (cost 12)</li> <li>API keys stored as SHA-256 hashes at rest</li> <li>CodeQL + gitleaks in CI</li> <li>Full test suite with integration tests</li> <li>OpenAPI 3.0 spec included</li> </ul> <h2> Architecture Overview </h2> <p>Here's how the pieces fit together:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight plaintext"><code>┌─────────────────────────────────────────────────────┐ │ HTTP Client (curl, SDK, Dashboard) │ │ with JWT Bearer or X-API-Key │ └─────────────────┬───────────────────────────────────┘ │ ▼ ┌─────────────────────────────────────────────────────┐ │ Chi Router + Middleware Layer │ │ • CORS, Logging, Rate Limiting, Security Headers │ │ • Request ID propagation │ └─────────────────┬───────────────────────────────────┘ │ ▼ ┌─────────────────────────────────────────────────────┐ │ Service Layer (Business Logic) │ │ • AuthService (JWT, bcrypt, API keys) │ │ • ComputeService (quota, capacity, state machine) │ │ • BillingService (usage tracking, invoices) │ │ • OrganizationService (multi-tenancy) │ └─────────────────┬───────────────────────────────────┘ │ ▼ ┌─────────────────────────────────────────────────────┐ │ PostgreSQL 16 (pgx driver) │ │ • users, organizations, members, instances │ │ • usage_records, invoices, invoices_line_items │ │ • quotas, region_capacity │ └─────────────────────────────────────────────────────┘ </code></pre> </div> <p><strong>Key insight:</strong> The API layer is intentionally thin. All business rules (quota checks, capacity enforcement, multi-tenancy gates) live in the <em>service layer</em>.</p> <h2> Let's Look at the Code </h2> <h3> 1. Multi-Tenant Quota Enforcement </h3> <p>Here's how we prevent one organization from consuming all resources:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight go"><code><span class="c">// internal/compute/service.go</span> <span class="k">func</span> <span class="p">(</span><span class="n">s</span> <span class="o">*</span><span class="n">Service</span><span class="p">)</span> <span class="n">enforceQuota</span><span class="p">(</span><span class="n">ctx</span> <span class="n">context</span><span class="o">.</span><span class="n">Context</span><span class="p">,</span> <span class="n">orgID</span> <span class="kt">int64</span><span class="p">,</span> <span class="n">cpu</span><span class="p">,</span> <span class="n">mem</span><span class="p">,</span> <span class="n">disk</span> <span class="kt">int</span><span class="p">)</span> <span class="kt">error</span> <span class="p">{</span> <span class="n">quota</span><span class="p">,</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">quotas</span><span class="o">.</span><span class="n">Get</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">orgID</span><span class="p">)</span> <span class="k">if</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"get quota: %w"</span><span class="p">,</span> <span class="n">err</span><span class="p">)</span> <span class="p">}</span> <span class="c">// Sum active instances for this org</span> <span class="n">usage</span><span class="p">,</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">repo</span><span class="o">.</span><span class="n">SumActiveByOrg</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">orgID</span><span class="p">)</span> <span class="k">if</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"sum org usage: %w"</span><span class="p">,</span> <span class="n">err</span><span class="p">)</span> <span class="p">}</span> <span class="c">// Check each dimension: instances, CPU, memory, disk</span> <span class="k">if</span> <span class="n">usage</span><span class="o">.</span><span class="n">Count</span><span class="o">+</span><span class="m">1</span> <span class="o">&gt;</span> <span class="n">quota</span><span class="o">.</span><span class="n">MaxInstances</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"%w: would exceed quota"</span><span class="p">,</span> <span class="n">ErrQuotaExceeded</span><span class="p">)</span> <span class="p">}</span> <span class="k">if</span> <span class="n">usage</span><span class="o">.</span><span class="n">CPUCores</span><span class="o">+</span><span class="kt">int64</span><span class="p">(</span><span class="n">cpu</span><span class="p">)</span> <span class="o">&gt;</span> <span class="n">quota</span><span class="o">.</span><span class="n">MaxCPUCores</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"%w: CPU exceeds quota (%d max)"</span><span class="p">,</span> <span class="n">ErrQuotaExceeded</span><span class="p">,</span> <span class="n">quota</span><span class="o">.</span><span class="n">MaxCPUCores</span><span class="p">)</span> <span class="p">}</span> <span class="k">if</span> <span class="n">usage</span><span class="o">.</span><span class="n">MemoryMB</span><span class="o">+</span><span class="kt">int64</span><span class="p">(</span><span class="n">mem</span><span class="p">)</span> <span class="o">&gt;</span> <span class="n">quota</span><span class="o">.</span><span class="n">MaxMemoryMB</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"%w: memory exceeds quota"</span><span class="p">,</span> <span class="n">ErrQuotaExceeded</span><span class="p">)</span> <span class="p">}</span> <span class="k">if</span> <span class="n">usage</span><span class="o">.</span><span class="n">DiskGB</span><span class="o">+</span><span class="kt">int64</span><span class="p">(</span><span class="n">disk</span><span class="p">)</span> <span class="o">&gt;</span> <span class="n">quota</span><span class="o">.</span><span class="n">MaxDiskGB</span> <span class="p">{</span> <span class="k">return</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"%w: disk exceeds quota"</span><span class="p">,</span> <span class="n">ErrQuotaExceeded</span><span class="p">)</span> <span class="p">}</span> <span class="k">return</span> <span class="no">nil</span> <span class="p">}</span> </code></pre> </div> <p><strong>What's important here:</strong></p> <ul> <li>Multi-dimensional quotas (not just "max instances", but CPU, memory, disk separately)</li> <li>Check enforcement at creation time (fail fast)</li> <li>Each org has independent limits</li> <li>Graceful error messages that tell users exactly what they hit</li> </ul> <h3> 2. Async State Machine </h3> <p>Instances don't provision instantly. Here's how we handle that without blocking:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight go"><code><span class="c">// internal/compute/service.go</span> <span class="k">var</span> <span class="n">instanceTransitions</span> <span class="o">=</span> <span class="k">map</span><span class="p">[</span><span class="kt">string</span><span class="p">]</span><span class="k">map</span><span class="p">[</span><span class="kt">string</span><span class="p">]</span><span class="kt">bool</span><span class="p">{</span> <span class="s">"pending"</span><span class="o">:</span> <span class="p">{</span> <span class="s">"running"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="c">// reconciler advances this</span> <span class="s">"terminating"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="c">// user can terminate anytime</span> <span class="p">},</span> <span class="s">"running"</span><span class="o">:</span> <span class="p">{</span> <span class="s">"stopping"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="s">"terminating"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="p">},</span> <span class="s">"stopping"</span><span class="o">:</span> <span class="p">{</span> <span class="s">"stopped"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="c">// reconciler advances this</span> <span class="s">"terminating"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="p">},</span> <span class="s">"stopped"</span><span class="o">:</span> <span class="p">{</span> <span class="s">"pending"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="c">// user restarts</span> <span class="s">"terminating"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="p">},</span> <span class="s">"terminating"</span><span class="o">:</span> <span class="p">{</span> <span class="s">"terminated"</span><span class="o">:</span> <span class="no">true</span><span class="p">,</span> <span class="c">// reconciler advances this</span> <span class="p">},</span> <span class="p">}</span> <span class="k">func</span> <span class="p">(</span><span class="n">s</span> <span class="o">*</span><span class="n">Service</span><span class="p">)</span> <span class="n">Create</span><span class="p">(</span><span class="n">ctx</span> <span class="n">context</span><span class="o">.</span><span class="n">Context</span><span class="p">,</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">userID</span> <span class="kt">int64</span><span class="p">,</span> <span class="n">req</span> <span class="n">models</span><span class="o">.</span><span class="n">CreateInstanceRequest</span><span class="p">)</span> <span class="p">(</span><span class="o">*</span><span class="n">models</span><span class="o">.</span><span class="n">ComputeInstance</span><span class="p">,</span> <span class="kt">error</span><span class="p">)</span> <span class="p">{</span> <span class="c">// ... validation &amp; quota checks ...</span> <span class="n">inst</span> <span class="o">:=</span> <span class="o">&amp;</span><span class="n">models</span><span class="o">.</span><span class="n">ComputeInstance</span><span class="p">{</span> <span class="n">OrganizationID</span><span class="o">:</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">Status</span><span class="o">:</span> <span class="s">"pending"</span><span class="p">,</span> <span class="c">// Start here</span> <span class="c">// ... specs ...</span> <span class="p">}</span> <span class="c">// Immediately persist to DB</span> <span class="k">if</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">repo</span><span class="o">.</span><span class="n">Create</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">inst</span><span class="p">);</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">err</span> <span class="p">}</span> <span class="c">// Return 202 Accepted to client</span> <span class="c">// Background reconciler advances pending → running</span> <span class="k">return</span> <span class="n">inst</span><span class="p">,</span> <span class="no">nil</span> <span class="p">}</span> </code></pre> </div> <p><strong>Why this matters:</strong></p> <ul> <li>API returns immediately (202 Accepted, not blocking)</li> <li>Client polls <code>/orgs/{id}/instances/{id}</code> to watch status</li> <li>Background reconciler (runs every 2 seconds) advances <code>pending → running</code> </li> <li>Prevents one slow provision from blocking other requests</li> </ul> <h3> 3. Usage-Based Billing </h3> <p>Here's how we track usage and generate invoices:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight go"><code><span class="c">// internal/billing/service.go</span> <span class="k">var</span> <span class="n">unitPrices</span> <span class="o">=</span> <span class="k">map</span><span class="p">[</span><span class="kt">string</span><span class="p">]</span><span class="kt">int64</span><span class="p">{</span> <span class="s">"cpu_hours"</span><span class="o">:</span> <span class="m">50</span><span class="p">,</span> <span class="c">// $0.50 per CPU hour</span> <span class="s">"memory_gb_hours"</span><span class="o">:</span> <span class="m">20</span><span class="p">,</span> <span class="c">// $0.20 per GB-hour</span> <span class="s">"disk_gb_hours"</span><span class="o">:</span> <span class="m">1</span><span class="p">,</span> <span class="c">// $0.01 per GB-hour</span> <span class="p">}</span> <span class="k">func</span> <span class="p">(</span><span class="n">s</span> <span class="o">*</span><span class="n">Service</span><span class="p">)</span> <span class="n">GenerateInvoice</span><span class="p">(</span><span class="n">ctx</span> <span class="n">context</span><span class="o">.</span><span class="n">Context</span><span class="p">,</span> <span class="n">orgID</span> <span class="kt">int64</span><span class="p">)</span> <span class="p">(</span><span class="o">*</span><span class="n">models</span><span class="o">.</span><span class="n">Invoice</span><span class="p">,</span> <span class="kt">error</span><span class="p">)</span> <span class="p">{</span> <span class="n">now</span> <span class="o">:=</span> <span class="n">time</span><span class="o">.</span><span class="n">Now</span><span class="p">()</span><span class="o">.</span><span class="n">UTC</span><span class="p">()</span> <span class="n">periodStart</span> <span class="o">:=</span> <span class="n">time</span><span class="o">.</span><span class="n">Date</span><span class="p">(</span><span class="n">now</span><span class="o">.</span><span class="n">Year</span><span class="p">(),</span> <span class="n">now</span><span class="o">.</span><span class="n">Month</span><span class="p">(),</span> <span class="m">1</span><span class="p">,</span> <span class="m">0</span><span class="p">,</span> <span class="m">0</span><span class="p">,</span> <span class="m">0</span><span class="p">,</span> <span class="m">0</span><span class="p">,</span> <span class="n">time</span><span class="o">.</span><span class="n">UTC</span><span class="p">)</span> <span class="n">periodEnd</span> <span class="o">:=</span> <span class="n">periodStart</span><span class="o">.</span><span class="n">AddDate</span><span class="p">(</span><span class="m">0</span><span class="p">,</span> <span class="m">1</span><span class="p">,</span> <span class="m">0</span><span class="p">)</span> <span class="c">// Get 30-day usage summary</span> <span class="n">usage</span><span class="p">,</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">usageRepo</span><span class="o">.</span><span class="n">GetSummary</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">periodStart</span><span class="p">)</span> <span class="k">if</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">err</span> <span class="p">}</span> <span class="c">// Calculate amounts in cents</span> <span class="n">items</span> <span class="o">:=</span> <span class="p">[]</span><span class="k">struct</span> <span class="p">{</span> <span class="n">resourceType</span> <span class="kt">string</span> <span class="n">quantity</span> <span class="kt">float64</span> <span class="n">description</span> <span class="kt">string</span> <span class="p">}{</span> <span class="p">{</span><span class="s">"cpu_hours"</span><span class="p">,</span> <span class="n">usage</span><span class="o">.</span><span class="n">CPUHours</span><span class="p">,</span> <span class="s">"CPU Hours"</span><span class="p">},</span> <span class="p">{</span><span class="s">"memory_gb_hours"</span><span class="p">,</span> <span class="n">usage</span><span class="o">.</span><span class="n">MemoryGBHours</span><span class="p">,</span> <span class="s">"Memory GB-hours"</span><span class="p">},</span> <span class="p">{</span><span class="s">"disk_gb_hours"</span><span class="p">,</span> <span class="n">usage</span><span class="o">.</span><span class="n">DiskGBHours</span><span class="p">,</span> <span class="s">"Disk GB-hours"</span><span class="p">},</span> <span class="p">}</span> <span class="k">var</span> <span class="n">totalCents</span> <span class="kt">int64</span> <span class="k">for</span> <span class="n">_</span><span class="p">,</span> <span class="n">item</span> <span class="o">:=</span> <span class="k">range</span> <span class="n">items</span> <span class="p">{</span> <span class="k">if</span> <span class="n">item</span><span class="o">.</span><span class="n">quantity</span> <span class="o">&gt;</span> <span class="m">0</span> <span class="p">{</span> <span class="n">totalCents</span> <span class="o">+=</span> <span class="kt">int64</span><span class="p">(</span><span class="n">item</span><span class="o">.</span><span class="n">quantity</span> <span class="o">*</span> <span class="kt">float64</span><span class="p">(</span><span class="n">unitPrices</span><span class="p">[</span><span class="n">item</span><span class="o">.</span><span class="n">resourceType</span><span class="p">]))</span> <span class="p">}</span> <span class="p">}</span> <span class="c">// Create invoice and line items</span> <span class="n">inv</span> <span class="o">:=</span> <span class="o">&amp;</span><span class="n">models</span><span class="o">.</span><span class="n">Invoice</span><span class="p">{</span> <span class="n">OrganizationID</span><span class="o">:</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">AmountCents</span><span class="o">:</span> <span class="n">totalCents</span><span class="p">,</span> <span class="n">PeriodStart</span><span class="o">:</span> <span class="n">periodStart</span><span class="p">,</span> <span class="n">PeriodEnd</span><span class="o">:</span> <span class="n">periodEnd</span><span class="p">,</span> <span class="n">Status</span><span class="o">:</span> <span class="s">"pending"</span><span class="p">,</span> <span class="p">}</span> <span class="k">if</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">invoiceRepo</span><span class="o">.</span><span class="n">Create</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">inv</span><span class="p">);</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">err</span> <span class="p">}</span> <span class="c">// Add detailed line items for transparency</span> <span class="k">for</span> <span class="n">_</span><span class="p">,</span> <span class="n">item</span> <span class="o">:=</span> <span class="k">range</span> <span class="n">items</span> <span class="p">{</span> <span class="k">if</span> <span class="n">item</span><span class="o">.</span><span class="n">quantity</span> <span class="o">&gt;</span> <span class="m">0</span> <span class="p">{</span> <span class="n">li</span> <span class="o">:=</span> <span class="o">&amp;</span><span class="n">models</span><span class="o">.</span><span class="n">InvoiceLineItem</span><span class="p">{</span> <span class="n">InvoiceID</span><span class="o">:</span> <span class="n">inv</span><span class="o">.</span><span class="n">ID</span><span class="p">,</span> <span class="n">Description</span><span class="o">:</span> <span class="n">item</span><span class="o">.</span><span class="n">description</span><span class="p">,</span> <span class="n">ResourceType</span><span class="o">:</span> <span class="n">item</span><span class="o">.</span><span class="n">resourceType</span><span class="p">,</span> <span class="n">Quantity</span><span class="o">:</span> <span class="n">item</span><span class="o">.</span><span class="n">quantity</span><span class="p">,</span> <span class="n">UnitPriceCents</span><span class="o">:</span> <span class="n">unitPrices</span><span class="p">[</span><span class="n">item</span><span class="o">.</span><span class="n">resourceType</span><span class="p">],</span> <span class="n">AmountCents</span><span class="o">:</span> <span class="kt">int64</span><span class="p">(</span><span class="n">item</span><span class="o">.</span><span class="n">quantity</span> <span class="o">*</span> <span class="kt">float64</span><span class="p">(</span><span class="n">unitPrices</span><span class="p">[</span><span class="n">item</span><span class="o">.</span><span class="n">resourceType</span><span class="p">])),</span> <span class="p">}</span> <span class="n">s</span><span class="o">.</span><span class="n">invoiceRepo</span><span class="o">.</span><span class="n">AddLineItem</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">li</span><span class="p">)</span> <span class="p">}</span> <span class="p">}</span> <span class="k">return</span> <span class="n">inv</span><span class="p">,</span> <span class="no">nil</span> <span class="p">}</span> </code></pre> </div> <p><strong>Key details:</strong></p> <ul> <li>Usage tracked in 30-day windows</li> <li>Money always in cents (integer math, no float rounding errors)</li> <li>Line items for transparency (customers can see CPU hours vs memory hours)</li> <li>Idempotent (calling twice for same month doesn't double-charge)</li> </ul> <h3> 4. Multi-Tenant Security </h3> <p>Every service layer operation checks membership before proceeding:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight go"><code><span class="c">// Pattern repeated across compute, billing, organizations services</span> <span class="k">func</span> <span class="p">(</span><span class="n">s</span> <span class="o">*</span><span class="n">Service</span><span class="p">)</span> <span class="n">Get</span><span class="p">(</span><span class="n">ctx</span> <span class="n">context</span><span class="o">.</span><span class="n">Context</span><span class="p">,</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">instanceID</span><span class="p">,</span> <span class="n">userID</span> <span class="kt">int64</span><span class="p">)</span> <span class="p">(</span><span class="o">*</span><span class="n">models</span><span class="o">.</span><span class="n">ComputeInstance</span><span class="p">,</span> <span class="kt">error</span><span class="p">)</span> <span class="p">{</span> <span class="c">// ALWAYS check membership first</span> <span class="k">if</span> <span class="n">_</span><span class="p">,</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">orgRepo</span><span class="o">.</span><span class="n">FindMember</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">orgID</span><span class="p">,</span> <span class="n">userID</span><span class="p">);</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">if</span> <span class="n">errors</span><span class="o">.</span><span class="n">Is</span><span class="p">(</span><span class="n">err</span><span class="p">,</span> <span class="n">database</span><span class="o">.</span><span class="n">ErrNotFound</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">ErrNotInOrg</span> <span class="p">}</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">fmt</span><span class="o">.</span><span class="n">Errorf</span><span class="p">(</span><span class="s">"check membership: %w"</span><span class="p">,</span> <span class="n">err</span><span class="p">)</span> <span class="p">}</span> <span class="c">// Then get the resource</span> <span class="n">inst</span><span class="p">,</span> <span class="n">err</span> <span class="o">:=</span> <span class="n">s</span><span class="o">.</span><span class="n">repo</span><span class="o">.</span><span class="n">FindByID</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">instanceID</span><span class="p">)</span> <span class="k">if</span> <span class="n">err</span> <span class="o">!=</span> <span class="no">nil</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">err</span> <span class="p">}</span> <span class="c">// Verify it belongs to the right org</span> <span class="k">if</span> <span class="n">inst</span><span class="o">.</span><span class="n">OrganizationID</span> <span class="o">!=</span> <span class="n">orgID</span> <span class="p">{</span> <span class="k">return</span> <span class="no">nil</span><span class="p">,</span> <span class="n">ErrNotFound</span> <span class="p">}</span> <span class="k">return</span> <span class="n">inst</span><span class="p">,</span> <span class="no">nil</span> <span class="p">}</span> </code></pre> </div> <p><strong>Why this is critical:</strong></p> <ul> <li>Multi-tenancy is enforced at the <em>service layer</em>, not just the router</li> <li>Prevents SQL injection or logic bugs from leaking data across orgs</li> <li>Defense in depth: router checks + service checks</li> </ul> <h2> Getting Started (5 Minutes) </h2> <h3> Prerequisites </h3> <ul> <li>Docker &amp; Docker Compose</li> <li>Go 1.26+</li> <li>curl &amp; jq (for examples)</li> </ul> <h3> Quick Start </h3> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code><span class="c"># 1. Clone the repo</span> git clone https://github.com/ogc16/iaas-platform.git <span class="nb">cd </span>iaas-platform <span class="c"># 2. Start PostgreSQL</span> docker compose up <span class="nt">-d</span> <span class="c"># 3. Run the server</span> go run ./cmd/server <span class="c"># 4. Try the API</span> curl <span class="nt">-X</span> POST http://localhost:8080/api/v1/auth/signup <span class="se">\</span> <span class="nt">-H</span> <span class="s1">'Content-Type: application/json'</span> <span class="se">\</span> <span class="nt">-d</span> <span class="s1">'{"email":"test@example.com","password":"changeme123","name":"Demo User"}'</span> <span class="c"># 5. Explore the dashboard</span> open http://localhost:8080 </code></pre> </div> <p>Or run the full end-to-end demo:<br> </p> <div class="highlight js-code-highlight"> <pre class="highlight shell"><code>bash examples/quick-start.sh </code></pre> </div> <p>This will:</p> <ol> <li>Start PostgreSQL</li> <li>Boot the server</li> <li>Create a user account</li> <li>Create an organization</li> <li>Provision an instance (watch it move from <code>pending</code> → <code>running</code>)</li> <li>Record usage and generate an invoice</li> <li>Clean up everything</li> </ol> <h2> Production Readiness Checklist </h2> <p>The platform is designed to be deployed to production. Here's what we included:</p> <h3> Security ✅ </h3> <ul> <li>Passwords: bcrypt (cost 12)</li> <li>API keys: stored as SHA-256 hashes</li> <li>JWT: HS256, 24h expiry, configurable</li> <li>Headers: CSP, HSTS, X-Frame-Options, etc.</li> <li>Secret scanning: gitleaks in CI</li> <li>Code scanning: CodeQL on every PR</li> <li>Audit logging: every action recorded with actor/IP/timestamp</li> </ul> <h3> Testing ✅ </h3> <ul> <li>45%+ code coverage</li> <li>Unit tests for all services</li> <li>Integration tests against PostgreSQL</li> <li>Race detector enabled in CI</li> <li> <code>go vet</code> &amp; <code>gofmt</code> checks</li> </ul> <h3> Operations ✅ </h3> <ul> <li>Health probes: <code>/healthz</code> (always 200) and <code>/readyz</code> (checks DB)</li> <li>Structured logging with request IDs</li> <li>OpenAPI 3.0 spec (machine-readable)</li> <li>Docker multi-stage build (distroless runtime)</li> <li>Graceful shutdown</li> <li>Connection pooling (pgx)</li> </ul> <h3> Deployment ✅ </h3> <ul> <li>Environment-based config (12-factor app)</li> <li>TLS support (in-process or reverse proxy)</li> <li>Horizontal scaling ready (but bring your own shared rate-limit store)</li> <li>Kubernetes-friendly (health probes, graceful shutdown, resource requests)</li> </ul> <h2> What's Next? </h2> <p>This is v0.1.0. The roadmap includes:</p> <h3> v1.0 </h3> <ul> <li>[ ] Prometheus metrics export (<code>/metrics</code>)</li> <li>[ ] CORS allowlist (currently allows all origins)</li> <li>[ ] Request-body size limits</li> <li>[ ] API key rotation endpoint</li> <li>[ ] golangci-lint integration</li> <li>[ ] Transaction support for multi-step writes</li> <li>[ ] Database migrations versioning</li> </ul> <h3> v2.0 </h3> <ul> <li>[ ] Real compute backends (Docker, Podman, AWS EC2)</li> <li>[ ] Webhook notifications for lifecycle events</li> <li>[ ] Per-account rate limiting beyond global bucket</li> <li>[ ] Control plane dashboard v2 (real-time charts, cost forecasts)</li> <li>[ ] Audit logs UI</li> <li>[ ] Usage aggregation jobs (move billing out of request path)</li> </ul> <h2> Key Takeaways </h2> <p>Building this taught me several things I wish I'd known earlier:</p> <ol> <li><p><strong>Service layer is the security perimeter.</strong> Router auth checks are necessary but not sufficient. Every service method should re-verify authorization.</p></li> <li><p><strong>Async operations unblock your API.</strong> Don't provision instances synchronously. Return 202 Accepted, then have a background reconciler advance state.</p></li> <li><p><strong>Track money in cents, not dollars.</strong> Float precision issues will haunt you. Use integers for financial calculations.</p></li> <li><p><strong>Multi-dimensionality matters.</strong> Quotas aren't just "max instances". Real systems need CPU, memory, disk, and network quotas.</p></li> <li><p><strong>Observability from day one.</strong> Structured logging, health probes, and OpenAPI specs aren't afterthoughts. Build them in.</p></li> <li><p><strong>Test with real PostgreSQL.</strong> Unit tests are great, but integration tests against your actual database catch the bugs that matter.</p></li> </ol> <h2> Try It Out </h2> <p><strong>GitHub:</strong> <a href="https://github.com/ogc16/iaas-platform" rel="noopener noreferrer">ogc16/iaas-platform</a></p> <ul> <li>⭐ Star if you find it useful</li> <li>🐛 Open issues for bugs or feature requests</li> <li>🤝 PRs welcome — see CONTRIBUTING.md</li> <li>💬 Questions? Start a discussion</li> </ul> <h2> Further Reading </h2> <p>If you want to dig deeper:</p> <ul> <li> <a href="https://github.com/ogc16/iaas-platform/blob/master/docs/ARCHITECTURE.md" rel="noopener noreferrer">Architecture Deep Dive</a> — detailed system design</li> <li> <a href="https://github.com/ogc16/iaas-platform/blob/master/SECURITY.md" rel="noopener noreferrer">Security Model</a> — threat analysis &amp; configuration hardening</li> <li> <a href="https://github.com/ogc16/iaas-platform/blob/master/openapi.yaml" rel="noopener noreferrer">OpenAPI Spec</a> — every endpoint documented</li> <li> <a href="https://github.com/ogc16/iaas-platform/blob/master/docs/DEPLOYMENT.md" rel="noopener noreferrer">Deployment Guide</a> — production hardening</li> </ul> <h2> Let's Connect </h2> <p>What aspects of building multi-tenant systems interest you most? I'd love to hear your thoughts in the comments:</p> <ul> <li>Are you building a SaaS platform?</li> <li>Which part would you like me to explain in more detail?</li> <li>What did you learn from this project?</li> </ul> <p>Tweet <a href="https://twitter.com/ogc16" rel="noopener noreferrer">@ogc16</a> or reply here.</p> devops go ogc16 Fri, 28 Mar 2025 08:43:28 +0000 https://dev.to/ogc16/mcp-servers-explained-4p7a https://dev.to/ogc16/mcp-servers-explained-4p7a <p>MCP Servers are systems that provide context, tools, and prompts to AI clients. <a href="https://mcp.so/" rel="noopener noreferrer">See</a></p> <p>MCP is a standard open-source protocol developed for AI to securely access external data and tools, enabling consistent and interoperable interactions<br> using a client-server system with standardized requests and secure OAuth authorization, allowing AI to access external tools and data sources.</p> ai opensource oauth automation ogc16 Fri, 28 Mar 2025 07:20:08 +0000 https://dev.to/ogc16/how-are-you-adapting-as-a-developer-pc4 https://dev.to/ogc16/how-are-you-adapting-as-a-developer-pc4 <p>With a few prompts you are able to achieve amazing products like:</p> <p><em>Streaming platform</em></p> <p>The objective is to create a streaming service offering a diverse library of content, personalized recommendations, and innovative features that enhance the viewing experience. This platform will differentiate itself through a focus on niche content, community engagement, and flexible pricing models while maintaining affordability and accessibility for users worldwide.</p> <p><strong>Key Goals and Objectives</strong><br> <strong>&gt; Primary Goals</strong>:</p> <ul> <li>Provide a vast library of movies, TV shows, documentaries, and original content.</li> <li>Offer advanced personalization through AI-driven recommendations based on user behavior and preferences.</li> <li>Support community features like live watch parties, discussion forums, and user-generated content.</li> </ul> <p><strong>&gt; Secondary Goals</strong>:</p> <ul> <li> Introduce flexible pricing models (e.g., à la carte content selection, tiered subscriptions).</li> <li>Ensure seamless integration with various devices (smart TVs, mobile apps, gaming consoles).</li> <li> Offer high-quality video streaming with options for 4K, HDR, and Dolby Atmos.</li> </ul> <p><strong>Target Audience</strong></p> <p><strong>&gt;Primary Users</strong>:</p> <ol> <li> Young adults (18–35) seeking diverse content options and social features.</li> <li> Niche audiences interested in independent films, documentaries, or international content.</li> <li> Families looking for affordable entertainment solutions with parental controls.</li> </ol> <p><strong>&gt;Demographics</strong>:</p> <blockquote> <p>Global reach with a focus on urban, tech-savvy individuals who value content discovery and community engagement.</p> </blockquote> <p><strong>Competitor Analysis</strong><br> <em>Netflix:</em></p> <blockquote> <p>Strengths: Extensive content library, strong brand recognition, user-friendly interface.<br> Weaknesses: Increasing costs; limited focus on niche content or community features.</p> </blockquote> <p><em>Amazon Prime Video:</em></p> <blockquote> <p>Strengths: Integration with Amazon ecosystem; additional benefits like free shipping.<br> Weaknesses: Confusing interface; less emphasis on social features.</p> </blockquote> <p><em>Hulu:</em></p> <blockquote> <p>Strengths: Strong focus on TV shows; affordable pricing options.<br> Weaknesses: Limited international content; fewer original titles compared to Netflix.</p> </blockquote> <p><em>Disney+:</em></p> <blockquote> <p>Strengths: Exclusive Disney, Pixar, Marvel, and Star Wars content; family-friendly focus.<br> Weaknesses: Limited content diversity beyond Disney franchises.</p> </blockquote> <p><strong>&gt; Design Requirements</strong></p> <p><strong><em>Functional Needs:</em></strong></p> <ul> <li> Advanced search and recommendation algorithms based on user viewing history and ratings.</li> <li> Community features like live watch parties with real-time chat, user reviews, and ratings.</li> <li> Flexible pricing models allowing users to choose specific content bundles or à la carte options.</li> <li> High-quality video streaming with support for 4K, HDR, and Dolby Atmos.</li> </ul> <p><strong>&gt; Visual Design:</strong></p> <ul> <li>Clean interface with customizable themes (light/dark mode).</li> <li>Dynamic content showcases highlighting new releases, user favorites, and trending titles.</li> <li>Intuitive navigation for discovering new content or accessing personalized recommendations.</li> </ul> <p><em>Regards</em> <a href="//x.com/d4m1n/status/1905285032849129795">Ref</a></p> <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fy0pd9nm3nsphfsiqp5r6.jpg" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fy0pd9nm3nsphfsiqp5r6.jpg" alt="Image " width="800" height="800"></a></p> softwareengineering softwaredevelopment ai machinelearning ogc16 Wed, 26 Mar 2025 19:10:42 +0000 https://dev.to/ogc16/can-ai-outsmart-human-beings-l31 https://dev.to/ogc16/can-ai-outsmart-human-beings-l31 <h2> Software's Future: Developer's Role Evolves </h2> <p>Software drives our world, from phones to global systems. Developers are key, but tech advances bring change.<br> The world is increasingly powered by software. From the smartphones in our pockets to the complex systems running global infrastructure, software developers are the architects of our digital reality. But as technology advances at an unprecedented pace, what does the future hold for these vital professionals? It's a landscape of both immense opportunity and evolving challenges.</p> <p><strong>The Rise of AI and Automation:</strong></p> <p>Perhaps the most significant shift is the increasing integration of artificial intelligence (AI) and automation into the development workflow. AI-powered tools are already assisting with code generation, bug detection, and testing, streamlining the development process. In the future, we can expect:</p> <ul> <li> <strong>AI-assisted coding:</strong> AI models will become more sophisticated, capable of generating complex code snippets and even entire applications based on high-level specifications. This doesn't mean developers will be replaced, but rather that their role will evolve towards higher-level design and architecture.</li> <li> <strong>Automated testing and deployment:</strong> Continuous integration and continuous deployment (CI/CD) pipelines, powered by AI, will become even more prevalent, allowing for faster and more reliable software releases.</li> <li> <strong>Low-code/no-code platforms:</strong> These platforms, which enable non-developers to create applications, will become more powerful and user-friendly. Developers will likely work alongside these platforms, integrating them into larger systems and building custom extensions.</li> </ul> <p><strong>The Expanding Scope of Software Development:</strong></p> <p>Software is permeating every aspect of our lives, leading to a broader range of development domains:</p> <ul> <li> <strong>The Internet of Things (IoT):</strong> With the proliferation of connected devices, developers will be crucial in building secure and efficient IoT applications, from smart home systems to industrial automation.</li> <li> <strong>Edge computing:</strong> As data processing moves closer to the source, developers will need to master edge computing technologies to build low-latency and responsive applications.</li> <li> <strong>Quantum computing:</strong> While still in its early stages, quantum computing holds immense potential for solving complex problems. Developers who specialize in quantum algorithms and software will be highly sought after.</li> <li> <strong>Virtual and Augmented Reality (VR/AR):</strong> The metaverse and immersive experiences are rapidly expanding, creating demand for developers who can build engaging and interactive VR/AR applications.</li> <li> <strong>Cybersecurity:</strong> As software becomes more integrated into critical infrastructure, cybersecurity will become paramount. Developers will need to prioritize security best practices and understand how to build resilient systems.</li> </ul> <p><strong>The Importance of Soft Skills and Adaptability:</strong></p> <p>Technical skills remain essential, but soft skills are becoming increasingly crucial:</p> <ul> <li> <strong>Communication and collaboration:</strong> Developers will need to effectively communicate with stakeholders, collaborate with cross-functional teams, and explain complex technical concepts to non-technical audiences.</li> <li> <strong>Problem-solving and critical thinking:</strong> The ability to analyze complex problems, identify root causes, and develop innovative solutions will be highly valued.</li> <li> <strong>Continuous learning and adaptability:</strong> The technology landscape is constantly evolving, so developers must be lifelong learners, willing to embrace new technologies and adapt to changing requirements.</li> <li> <strong>Ethical considerations:</strong> As software becomes more powerful, developers will need to consider the ethical implications of their work, ensuring that their creations are used responsibly and for the benefit of society.</li> </ul> <p><strong>The Rise of Specialized Roles:</strong></p> <p>As the field of software development becomes more complex, specialization will become increasingly important:</p> <ul> <li> <strong>AI/ML engineers:</strong> Developers with expertise in machine learning and deep learning will be in high demand.</li> <li> <strong>Cloud architects:</strong> Designing and managing cloud-based infrastructure will require specialized skills.</li> <li> <strong>DevOps engineers:</strong> Streamlining the development and deployment process will remain a critical role.</li> <li> <strong>Security engineers:</strong> Protecting software systems from cyber threats will be a top priority.</li> <li> <strong>Data scientists and engineers:</strong> The ability to process and analyze large datasets will be essential for building data-driven applications.</li> </ul> <p><strong>Conclusion:</strong></p> <p>The future for software developers is bright, albeit one that requires constant adaptation. While AI and automation may reshape the development process, they will also create new opportunities for developers to focus on higher-level tasks and innovate in emerging fields. The key to success will be a combination of strong technical skills, essential soft skills, and a willingness to embrace continuous learning. By staying ahead of the curve and adapting to the evolving landscape, software developers can continue to play a vital role in shaping the future of technology.<br> Verdict: There are a lot tasks to be completed by human beings.</p> programming ai cybersecurity futurechallenge ogc16 Wed, 26 Mar 2025 18:56:54 +0000 https://dev.to/ogc16/-softwares-future-developers-role-evolves-software-drives-our-world-from-phones-to-global-533j https://dev.to/ogc16/-softwares-future-developers-role-evolves-software-drives-our-world-from-phones-to-global-533j softwaredevelopment developer discuss