The latest articles on DEV Community by Admas (@admasdev). https://dev.to/admasdev 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%2F4147295%2Fcdfad23e-2ca4-4240-b016-727315db1d05.png https://dev.to/admasdev en Admas Mon, 28 Sep 2026 14:15:32 +0000 https://dev.to/admasdev/ai-generated-the-app-how-do-you-know-it-actually-works-j8g https://dev.to/admasdev/ai-generated-the-app-how-do-you-know-it-actually-works-j8g <p>Generating code is no longer the difficult part.</p> <p>The harder question is: <strong>who verified the behavior of the application after the code was written?</strong></p> <p>A successful build proves that the project compiled or bundled. It does not prove that:</p> <ul> <li>A user can complete the main workflow.</li> <li>Invalid input is rejected correctly.</li> <li>Saved data survives a refresh.</li> <li>Unauthorized users cannot access protected actions.</li> <li>The preview matches the current code.</li> <li>A later change did not break something that previously worked.</li> </ul> <p>For an AI-built application, I believe “done” should require observable evidence from the running product.</p> <p>A simple verification standard can begin with four checks:</p> <h2> 1. Test the successful path </h2> <p>Complete the action exactly as a real user would.</p> <p>If it is a booking application, create a booking and verify that it appears in the owner dashboard.</p> <h2> 2. Test the failure path </h2> <p>Submit invalid information or attempt an action that should be rejected.</p> <p>A working success screen means little if the application accepts impossible dates, duplicate reservations, or malformed data.</p> <h2> 3. Test persistence </h2> <p>Refresh or reopen the application.</p> <p>If the data disappears, the interface may look complete while the actual product is not.</p> <h2> 4. Test a boundary </h2> <p>Attempt one action the current user should not be allowed to perform.</p> <p>Authentication is not the same as authorization. A user being signed in does not mean every record or action should be accessible.</p> <p>This verification loop is central to what we are building at Admas: <a href="https://www.admas.dev" rel="noopener noreferrer">https://www.admas.dev</a></p> <p>Admas works inside the project through planning, editing, running, testing, repairing, and previewing—not merely generating files and declaring completion.</p> <p>The larger principle applies to every AI software-development tool:</p> <blockquote> <p>Don’t ask only, “Did the agent write the code?”<br><br> Ask, “Can I reproduce both success and failure in the running application?”</p> </blockquote> <p>What evidence do you require before trusting an AI-generated application: a successful build, automated tests, a working preview, or something else?</p> <p>Disclosure: I’m part of the team building Admas.</p> ai webdev programming discuss