<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Bram Lesage</title>
    <description>The latest articles on DEV Community by Bram Lesage (@bramlesage).</description>
    <link>https://dev.to/bramlesage</link>
    <image>
      <url>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%2F4258958%2F62e9b698-56fd-4b2b-a50d-c9b911dde654.png</url>
      <title>DEV Community: Bram Lesage</title>
      <link>https://dev.to/bramlesage</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/bramlesage"/>
    <language>en</language>
    <item>
      <title>AI at school: don't ban the tool, teach the checking</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 16:38:46 +0000</pubDate>
      <link>https://dev.to/bramlesage/ai-at-school-dont-ban-the-tool-teach-the-checking-60m</link>
      <guid>https://dev.to/bramlesage/ai-at-school-dont-ban-the-tool-teach-the-checking-60m</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage, 17, from Mechelen, Belgium, on why students should learn to use AI — and why the thinking still has to be theirs&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17, I live in Mechelen, Belgium, and I'm still in school. I also build software with AI every day. So I see AI from two sides: as a student who gets homework, and as a maker who uses AI agents to build real products.&lt;/p&gt;

&lt;p&gt;My opinion is simple. Students should learn to use AI as a tool, and they should learn to check its work themselves. Pretending AI doesn't exist doesn't work. Letting AI do the thinking for you doesn't work either. The interesting part is in between.&lt;/p&gt;

&lt;h2&gt;
  
  
  Students are already using it
&lt;/h2&gt;

&lt;p&gt;This is not a future debate. In Flanders, the MAIVO study, commissioned by the Department of Education and published in September 2026, surveyed school leaders, ICT coordinators, 586 teachers and 636 pupils across 482 schools. Its main conclusion: AI is already present in schools, but its use is mostly exploratory, informal and not built into how schools work. Pupils, especially in secondary school, use AI to look things up, ask questions and improve texts — sometimes outside class, without a clear assignment or permission. The researchers call that shadow use, and they found that many schools still have no formal AI policy.&lt;/p&gt;

&lt;p&gt;So the question is not whether students use AI. It's whether anyone teaches them how.&lt;/p&gt;

&lt;h2&gt;
  
  
  Banning it doesn't solve much
&lt;/h2&gt;

&lt;p&gt;Kenniscentrum Digisprong, the Flemish government's knowledge centre for digital education, puts it bluntly in its practical guide for schools: banning generative AI like ChatGPT has "little to no effect". It advises schools to write clear guidelines for pupils, teachers and staff instead, to make agreements about AI in tests and evaluations, and to move from grading only the end result to grading the whole learning process.&lt;/p&gt;

&lt;p&gt;I agree with that. A ban mostly teaches students to hide what they do. Clear rules teach them when AI is allowed, when it isn't, and why.&lt;/p&gt;

&lt;h2&gt;
  
  
  But AI can make you worse at learning
&lt;/h2&gt;

&lt;p&gt;Here is the part people who love AI (like me) need to hear. A field experiment with nearly a thousand high school maths students in Türkiye, published in PNAS, tested what happens when students practise with GPT-4. With a standard chatbot, their practice scores went up by 48%. With a version designed as a tutor, they went up by 127%. But when the AI was taken away and they took a normal exam, the students who had used the standard chatbot scored 17% worse than students who never had AI at all. The researchers found that many of them had used it as a "crutch": asking for the answer and copying it.&lt;/p&gt;

&lt;p&gt;The tutor version, which was built to help students think instead of handing them answers, did not cause that drop.&lt;/p&gt;

&lt;p&gt;The OECD's Digital Education Outlook 2026 draws the same lesson from the research. When students lean too heavily on generative AI, the mental effort that turns answers into understanding drops, so the work looks better while the learning does not improve. The OECD suggests that teachers focus on process, on how students think and learn, rather than only on the output.&lt;/p&gt;

&lt;p&gt;That is the most important fact in this whole debate. Getting the right answer and learning the material are not the same thing.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I think good AI use at school looks like
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Learn first, then speed up.&lt;/strong&gt; If you don't understand fractions, an AI that does fractions for you just hides the problem until the exam. Use AI to explain, quiz you and give feedback, not to skip the part where you learn.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Always check the work.&lt;/strong&gt; AI makes mistakes, and it says wrong things with confidence. If you hand something in, you are responsible for every sentence in it, not the AI.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Follow your school's rules.&lt;/strong&gt; Schools are allowed to forbid AI help, for a task or entirely. Tests and exams are about what you can do on your own. That's fair, and I think students should respect it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Be open about it.&lt;/strong&gt; If AI helped and it was allowed, say so. Hiding it is what turns a tool into cheating.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Keep personal data out.&lt;/strong&gt; Digisprong advises schools to keep AI tasks to non-personal questions, like a maths exercise, and to be careful with pupils' personal data.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Schools are being asked to teach this anyway
&lt;/h2&gt;

&lt;p&gt;Since 2 February 2025, the EU AI Act requires organisations that provide or use AI systems to take measures for a sufficient level of AI literacy among their staff (Article 4). Digisprong points out that this applies to school boards too, and asks the obvious question: if staff need to understand the risks of AI, shouldn't pupils who are allowed to use it understand them too?&lt;/p&gt;

&lt;p&gt;There is also a lot of confusion about age. Digisprong explains that the AI Act itself sets no minimum age; the limits come from privacy law (in Belgium, 13 for consent to online services) and from the terms of use of each AI tool. UNESCO's guidance on generative AI in education recommends an age limit of 13 for independent use and asks schools to check tools before students use them. Katholiek Onderwijs Vlaanderen writes that there are no binding Flemish agreements on AI in schools yet, that its model school rules allow pupils to work with generative AI under adult guidance, and that it is the task of schools to teach all their pupils to use AI responsibly and critically.&lt;/p&gt;

&lt;p&gt;Everyone serious seems to land in the same place: not "ban it", not "let it do everything", but "teach it, with rules".&lt;/p&gt;

&lt;h2&gt;
  
  
  How I use AI when I build
&lt;/h2&gt;

&lt;p&gt;I build my projects like a director. For a product or a game I often put together a team of AI agents — a designer, a builder, a tester and a strict reviewer. That lets me build much faster than I could alone.&lt;/p&gt;

&lt;p&gt;But the agents don't decide when something is done. I test every version myself and decide what ships. Green automated tests don't mean it works: on the first day of one of my Roblox games, 215 automated tests were green, and some systems still hadn't been run in Roblox Studio. The AI does a lot of the work. The judgement is mine, and so is the responsibility.&lt;/p&gt;

&lt;p&gt;That is exactly the skill I think school should teach: not how to type a question into a chatbot, but how to give AI a clear job, check what comes back, and know when it's wrong.&lt;/p&gt;

&lt;p&gt;One of my own products is Katern, a Chrome extension that helps with schoolwork. I designed it so the student stays in control: it shows its plan first, it never presses Hand in, the student checks the work and hands it in themselves, and it does not do tests or exams. Schools may forbid AI help, and that is their decision to make.&lt;/p&gt;

&lt;h2&gt;
  
  
  My conclusion
&lt;/h2&gt;

&lt;p&gt;AI at school is not going away, and it shouldn't. Used well, it explains things at 11 at night when no teacher is around, gives feedback on a draft and quizzes you before a test. Used badly, it gives you good grades on homework and bad grades on the exam.&lt;/p&gt;

&lt;p&gt;The difference isn't the tool. It's whether you still do the thinking. That is what schools should teach, and that is what students should demand to learn.&lt;/p&gt;

&lt;h2&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;MAIVO study, "AI in het Vlaams onderwijs anno 2025", Department of Education and Training, Flanders (published 1 September 2026): &lt;a href="https://www.vlaanderen.be/onderwijsprofessionals/regelgeving-data-en-onderzoek/data-en-onderzoek/kennisplatform-onderwijs/ai-in-het-vlaams-onderwijs-anno-2025" rel="noopener noreferrer"&gt;https://www.vlaanderen.be/onderwijsprofessionals/regelgeving-data-en-onderzoek/data-en-onderzoek/kennisplatform-onderwijs/ai-in-het-vlaams-onderwijs-anno-2025&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Kenniscentrum Digisprong, "Praktische gids: hoe ga je bewust en verantwoord met AI om in je school?": &lt;a href="https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/praktische-gids-hoe-ga-je-bewust-en-verantwoord-met-ai-om-in-je-school" rel="noopener noreferrer"&gt;https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/praktische-gids-hoe-ga-je-bewust-en-verantwoord-met-ai-om-in-je-school&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Kenniscentrum Digisprong, "Vanaf welke leeftijd kunnen of mogen je leerlingen generatieve AI gebruiken?": &lt;a href="https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/vanaf-welke-leeftijd-kunnen-of-mogen-je-leerlingen-generatieve-ai-gebruiken" rel="noopener noreferrer"&gt;https://www.vlaanderen.be/kenniscentrum-digisprong/themas/innovatie/artificiele-intelligentie/vanaf-welke-leeftijd-kunnen-of-mogen-je-leerlingen-generatieve-ai-gebruiken&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Bastani et al., "Generative AI without guardrails can harm learning: Evidence from high school mathematics", PNAS (2025): &lt;a href="https://www.pnas.org/doi/10.1073/pnas.2422633122" rel="noopener noreferrer"&gt;https://www.pnas.org/doi/10.1073/pnas.2422633122&lt;/a&gt; (preprint: &lt;a href="https://hamsabastani.github.io/education_llm.pdf" rel="noopener noreferrer"&gt;https://hamsabastani.github.io/education_llm.pdf&lt;/a&gt;)&lt;/li&gt;
&lt;li&gt;OECD, Digital Education Outlook 2026: &lt;a href="https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/01/oecd-digital-education-outlook-2026_940e0dd8/062a7394-en.pdf" rel="noopener noreferrer"&gt;https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/01/oecd-digital-education-outlook-2026_940e0dd8/062a7394-en.pdf&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;EU AI Act, Article 4: AI literacy: &lt;a href="https://artificialintelligenceact.eu/article/4/" rel="noopener noreferrer"&gt;https://artificialintelligenceact.eu/article/4/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;UNESCO, Guidance for generative AI in education and research (2023): &lt;a href="https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research" rel="noopener noreferrer"&gt;https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;UN News, "UNESCO calls for regulations on AI use in schools" (7 September 2023): &lt;a href="https://news.un.org/en/story/2023/09/1140477" rel="noopener noreferrer"&gt;https://news.un.org/en/story/2023/09/1140477&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Katholiek Onderwijs Vlaanderen, "AI-regelgeving": &lt;a href="https://pro.katholiekonderwijs.vlaanderen/artifici%C3%ABle-intelligentie/airegelgeving" rel="noopener noreferrer"&gt;https://pro.katholiekonderwijs.vlaanderen/artifici%C3%ABle-intelligentie/airegelgeving&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;em&gt;Written by Bram Lesage. More of my projects: &lt;a href="https://bramlesage.pages.dev" rel="noopener noreferrer"&gt;bramlesage.pages.dev&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>education</category>
      <category>discuss</category>
      <category>career</category>
    </item>
    <item>
      <title>I put all 22 versions of my unfinished cat game online — including the bad first one</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 16:37:51 +0000</pubDate>
      <link>https://dev.to/bramlesage/i-put-all-22-versions-of-my-unfinished-cat-game-online-including-the-bad-first-one-5egk</link>
      <guid>https://dev.to/bramlesage/i-put-all-22-versions-of-my-unfinished-cat-game-online-including-the-bad-first-one-5egk</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage, 17, from Mechelen, Belgium, on why publishing every version of a game is worth it&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17 and I live in Mechelen, Belgium. I build games in the browser and on Roblox. None of them is finished. Two of the browser games are now playable on my website anyway, as work-in-progress demos: &lt;a href="https://bramlesage.pages.dev/spelen/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/spelen/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;For one of them, a cat game called Lantaarnhoek, I did something most people don't do. I didn't just put the newest version online. I put all 22 versions online, from the very first rough prototype to the latest build. This is why.&lt;/p&gt;

&lt;h2&gt;
  
  
  What Lantaarnhoek is
&lt;/h2&gt;

&lt;p&gt;Lantaarnhoek is a third-person 3D adventure that runs in the browser with three.js. You play as cats in a street at night, in a neighbourhood called Lantaarnhoek. There are four cats, Mist, Spriet, Oor and Schim, and each moves differently and finds a different kind of evidence. Mist squeezes through grates and pulls keys and letters out of tight spaces. Spriet has a double jump and reads papers through windows. Oor is slow, but balances on narrow ledges and overhears conversations. Schim is invisible in the shadows and can lure people away so another cat can slip past.&lt;/p&gt;

&lt;p&gt;The story is about a street that gets quieter every night: lights go out and people leave. The cats notice first. A secret only counts once two kinds of evidence confirm it.&lt;/p&gt;

&lt;p&gt;It is not finished. The demo shows the street, the cats and the start of the story, not the whole game.&lt;/p&gt;

&lt;h2&gt;
  
  
  Version 1: coloured shapes in 2D
&lt;/h2&gt;

&lt;p&gt;The first version is from 7 October 2026 at 17:37. It is not 3D at all. It is a flat 2D canvas, 960 by 400 pixels, with the four cats as coloured shapes, arrow keys to move and jump, and the number keys 1 to 4 to switch cats. The whole game is one JavaScript file of about 4.5 kilobytes.&lt;/p&gt;

&lt;p&gt;It looks bad. But the idea is already there: four cats, each with its own skill, three secrets in one street, and a meter for how suspicious people get. Seventeen minutes later, at 17:54, version 2 was the first 3D version, with simple low-poly shapes.&lt;/p&gt;

&lt;h2&gt;
  
  
  From version 2 to version 22
&lt;/h2&gt;

&lt;p&gt;Version 22 is from 15 October 2026 at 20:47. In between are 20 more builds, almost all from 7, 8, 13 and 15 October. The latest version loads about 20 MB of models, textures and code, with a real cat model, wet cobblestones and a sky.&lt;/p&gt;

&lt;p&gt;I make games like a director. For Lantaarnhoek, a Game team of AI roles, with design, art, story and a critic, worked out the concept over three rounds, and I steered after every round. Every build is checked with screenshots until the street, the light and the way a cat walks are right. Most build names on my computer end in "getest", which means tested. One of them, version 17, is literally called "load test failed". It's on the site too.&lt;/p&gt;

&lt;p&gt;You can play every single one: &lt;a href="https://bramlesage.pages.dev/spelen/lantaarnhoek/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/spelen/lantaarnhoek/&lt;/a&gt; — the newest at the top, the first 2D prototype at the bottom. Each version has its own address, from /spelen/lantaarnhoek/v1/ to /spelen/lantaarnhoek/v22/.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why publish the bad versions?
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;It is honest.&lt;/strong&gt; If I only showed version 22, it would look like a game appeared out of nowhere. It didn't. It started as a 4.5 KB game made of coloured shapes. Showing that is more truthful than any trailer.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It shows the process, not just the result.&lt;/strong&gt; You can see in a few clicks how the same idea grows: from flat shapes, to low-poly 3D, to a cat on wet cobblestones. That is more interesting to me, and I think to other people who make things, than one polished screenshot.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It makes it easier to start.&lt;/strong&gt; A lot of people never share anything because their first version looks bad. Every first version looks bad. Mine is online, with the date on it, so anyone can see that.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It keeps me accountable.&lt;/strong&gt; If the newest version is worse than an older one, anyone can check. That forces me to only call something better when it actually is.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It is a backup in public.&lt;/strong&gt; Old builds tend to disappear in folders and zip files. Now each one has a link.&lt;/p&gt;

&lt;h2&gt;
  
  
  The rules I kept
&lt;/h2&gt;

&lt;p&gt;Publishing unfinished work only works if you say clearly what it is. So every game on the page has a label that says "work in progress — demo". The credits for the cat model and textures are on the page, because I did not make those myself: the cat is "Bicolor Cat" by kenchoo, based on "Fripouille" by guillaume bolis (CC-BY 4.0), and the textures and sky come from Poly Haven (CC0).&lt;/p&gt;

&lt;p&gt;Lantaarnhoek needs a keyboard and mouse: WASD to walk, drag the mouse to look, Space to jump, 1 to 4 to switch cats.&lt;/p&gt;

&lt;h2&gt;
  
  
  The second game: Skydrop Royale
&lt;/h2&gt;

&lt;p&gt;The other playable demo is Skydrop Royale, a battle royale prototype: &lt;a href="https://bramlesage.pages.dev/spelen/skydrop-royale/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/spelen/skydrop-royale/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The lobby, item shop, season shop and locker already work. The match itself is not built yet; the Play button says so. The coins are fake, so nothing can actually be bought. There is only one version on my computer, so that page has one version.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I'd tell someone else
&lt;/h2&gt;

&lt;p&gt;Put your first version online. Put a date on it. Say clearly that it is not finished. Then keep going, and let people see the difference.&lt;/p&gt;

&lt;p&gt;My Roblox games are not on the page, because they only work inside Roblox, and I have not published them yet. You can read about all my projects on my website, under the same name I use for everything: Bram Lesage.&lt;/p&gt;

&lt;p&gt;Play the demos: &lt;a href="https://bramlesage.pages.dev/spelen/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/spelen/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Written by Bram Lesage. More of my projects: &lt;a href="https://bramlesage.pages.dev" rel="noopener noreferrer"&gt;bramlesage.pages.dev&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>gamedev</category>
      <category>javascript</category>
      <category>threejs</category>
      <category>beginners</category>
    </item>
    <item>
      <title>Why Flying Computers Are a Good Idea</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 16:19:08 +0000</pubDate>
      <link>https://dev.to/bramlesage/why-flying-computers-are-a-good-idea-p02</link>
      <guid>https://dev.to/bramlesage/why-flying-computers-are-a-good-idea-p02</guid>
      <description>&lt;p&gt;&lt;em&gt;From drones to the stratosphere to orbit: why I think the future of computing leaves the ground&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Most people picture a computer as a box on a desk or a building full of servers. I think that picture is about to change. The next big jump in computing won't only be faster chips. It will be computers that fly.&lt;/p&gt;

&lt;p&gt;I'm Bram Lesage, 17, and I build apps and games from Mechelen, Belgium. I've been reading a lot about where AI is going, and I keep coming back to one idea: put the computer where the data is, and where the energy is. More and more, that place is up.&lt;/p&gt;

&lt;p&gt;Here's my case, with sources, and with the strongest arguments against it too.&lt;/p&gt;

&lt;h2&gt;
  
  
  The problem: AI is running out of ground
&lt;/h2&gt;

&lt;p&gt;AI needs electricity, and a lot of it. According to the &lt;a href="https://www.iea.org/reports/energy-and-ai/executive-summary" rel="noopener noreferrer"&gt;International Energy Agency&lt;/a&gt;, data centres used about 415 TWh in 2024, around 1.5% of the world's electricity. By 2030 that is projected to more than double to about 945 TWh, which is slightly more than Japan's entire electricity use today.&lt;/p&gt;

&lt;p&gt;The grid can't keep up. The IEA &lt;a href="https://www.iea.org/files/dd7c2387-2f60-4b60-8c5f-6563b6aa1e4c/EnergyandAI.pdf" rel="noopener noreferrer"&gt;estimates&lt;/a&gt; that around 20% of planned data centre projects could be at risk of delays, and that building new transmission lines can take four to eight years in advanced economies.&lt;/p&gt;

&lt;p&gt;Then there's water. The same IEA report estimates that data centres consume around 560 billion litres of water per year, a figure that could rise to around 1,200 billion litres by 2030.&lt;/p&gt;

&lt;p&gt;So the question isn't whether we need more computing. It's where we're going to put it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Level 1: computers on drones
&lt;/h2&gt;

&lt;p&gt;The first kind of flying computer already exists, and it works.&lt;/p&gt;

&lt;p&gt;The &lt;a href="https://www.skydio.com/x10" rel="noopener noreferrer"&gt;Skydio X10&lt;/a&gt; drone carries an NVIDIA Jetson Orin, basically a small AI computer. According to Skydio's &lt;a href="https://support.skydio.com/hc/article_attachments/55260390977435" rel="noopener noreferrer"&gt;operator manual&lt;/a&gt;, all its autonomy, navigation and perception functions run onboard. It can &lt;a href="https://support.skydio.com/hc/en-us/articles/18919821734683-Skydio-X10-Frequently-Asked-Questions" rel="noopener noreferrer"&gt;build 2D maps and 3D models on the drone itself&lt;/a&gt;, without external processing.&lt;/p&gt;

&lt;p&gt;Why does that matter? Because sending raw video back to a server is slow and needs a connection. A drone that thinks for itself still works when the signal drops.&lt;/p&gt;

&lt;p&gt;Scale it up and you get &lt;a href="https://www.zipline.com/newsroom/zipline-surpasses-2-million-deliveries-raises-more-than-600m-to-power-next-phase-of-growth-and-expands-operations-to-houston-and-phoenix" rel="noopener noreferrer"&gt;Zipline&lt;/a&gt;. In January 2026 the company said it had passed two million commercial deliveries and more than 125 million autonomous miles, and was making a delivery somewhere in the world every 30 seconds. Those are flying computers carrying blood, medicine and food.&lt;/p&gt;

&lt;h2&gt;
  
  
  Level 2: computers in the stratosphere
&lt;/h2&gt;

&lt;p&gt;Go higher, to the stratosphere, &lt;a href="https://www.flightglobal.com/aerospace/2026/02/airbus-aalto-closes-on-australian-site-for-second-zephyr-launch-location/" rel="noopener noreferrer"&gt;some 60,000 feet up&lt;/a&gt;. That's above the weather, where solar-powered aircraft can stay up for weeks.&lt;/p&gt;

&lt;p&gt;Airbus subsidiary AALTO's Zephyr &lt;a href="https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/" rel="noopener noreferrer"&gt;flew for 67 days, 6 hours and 52 minutes&lt;/a&gt; without landing, ending on 28 April 2025, a world record. According to &lt;a href="https://aerospaceamerica.aiaa.org/aalto-maintains-2026-target-for-commercial-operations-despite-aircraft-loss/" rel="noopener noreferrer"&gt;Aerospace America&lt;/a&gt;, it has a 25-metre wingspan but weighs only about 70 kg. NTT DoCoMo invested $100 million in AALTO, and the plan is to turn these aircraft into flying cell towers, which is useful after disasters when ground networks fail.&lt;/p&gt;

&lt;p&gt;A flying cell tower is a flying computer: it receives, processes and routes data from the sky.&lt;/p&gt;

&lt;h2&gt;
  
  
  Level 3: the ultimate flying computer, in orbit
&lt;/h2&gt;

&lt;p&gt;Now the big one: data centres in space.&lt;/p&gt;

&lt;p&gt;The logic is simple. In the right orbit, the sun almost never sets. Google Research &lt;a href="https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/" rel="noopener noreferrer"&gt;writes&lt;/a&gt; that "in the right orbit, a solar panel can be up to 8 times more productive than on earth, and produce power nearly continuously, reducing the need for batteries."&lt;/p&gt;

&lt;p&gt;No land to buy. No grid queue. And as the European &lt;a href="https://www.thalesaleniaspace.com/en/press-releases/thales-alenia-space-reveals-results-ascend-feasibility-study-space-data-centers-0" rel="noopener noreferrer"&gt;ASCEND study&lt;/a&gt; by Thales Alenia Space points out, space data centres "would not require water to cool them."&lt;/p&gt;

&lt;p&gt;This is not just a fantasy anymore:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Starcloud&lt;/strong&gt; launched Starcloud-1 in November 2025 with an Nvidia H100 on board. According to &lt;a href="https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html" rel="noopener noreferrer"&gt;CNBC&lt;/a&gt;, it trained Andrej Karpathy's NanoGPT on the complete works of Shakespeare in orbit and ran Google's Gemma model there. In August 2026 Starcloud &lt;a href="https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/" rel="noopener noreferrer"&gt;raised a $250 million extension&lt;/a&gt;, with Nvidia participating, and plans two 8 kW compute satellites in 2027.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Google's Project Suncatcher&lt;/strong&gt; put its first prototype satellite, built with Planet, &lt;a href="https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/" rel="noopener noreferrer"&gt;into orbit on 1 October 2026&lt;/a&gt;. Google also &lt;a href="https://blog.google/innovation-and-ai/models-and-research/google-research/blue-origin-project-suncatcher-partnership/" rel="noopener noreferrer"&gt;partnered with Blue Origin&lt;/a&gt; on a New Glenn launch, targeted for Q1 2029, for satellites with roughly the capacity of a data centre rack.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Axiom Space&lt;/strong&gt; &lt;a href="https://www.axiomspace.com/release/axiom-space-kepler-ready-orbital-computing-for-quantum-era" rel="noopener noreferrer"&gt;operates two compute nodes, ARC 1 and ARC 2, in orbit&lt;/a&gt; on Kepler's optical relay network.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;SpaceX&lt;/strong&gt; envisions a constellation of &lt;a href="https://arstechnica.com/space/2026/07/how-hard-is-it-to-build-orbital-data-centers-actually/" rel="noopener noreferrer"&gt;1 million satellites generating 120 GW&lt;/a&gt; for AI.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;And the people saying this aren't random. Jeff Bezos said at Italian Tech Week in 2025: "These giant training clusters, those will be better built in space, because we have solar power there, 24/7. There are no clouds and no rain, no weather" (&lt;a href="https://www.datacenterdynamics.com/en/news/jeff-bezos-claims-there-will-be-gigawatt-data-centers-in-space-in-10-years/" rel="noopener noreferrer"&gt;DCD&lt;/a&gt;). Google CEO Sundar Pichai told &lt;a href="https://sources.news/p/sundar-pichai-gemini-4-google-other-bets" rel="noopener noreferrer"&gt;Sources&lt;/a&gt; he sees it as a serious answer to Google's long-term compute needs, on a 10-to-20-year horizon.&lt;/p&gt;

&lt;h2&gt;
  
  
  The thread that connects all three
&lt;/h2&gt;

&lt;p&gt;Drones, stratosphere and orbit share one idea: process data where it's created.&lt;/p&gt;

&lt;p&gt;Axiom's CEO Jonathan Cirtain &lt;a href="https://www.axiomspace.com/release/axiom-space-kepler-ready-orbital-computing-for-quantum-era" rel="noopener noreferrer"&gt;put it well&lt;/a&gt;: "Satellite sensor and operations systems generate more data than communication systems can transmit to ground stations during an orbit." His answer isn't just more bandwidth, it's "less dependence on bandwidth in the first place."&lt;/p&gt;

&lt;p&gt;That's exactly what a Skydio drone does at 100 metres, and what a satellite does at 500 km. Flying computers don't replace the cloud. They move the thinking closer to the action.&lt;/p&gt;

&lt;h2&gt;
  
  
  The strongest arguments against, and my answers
&lt;/h2&gt;

&lt;p&gt;I don't want to just hype this. Here are the real problems.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. Cooling in a vacuum is hard.&lt;/strong&gt; There's no air in space, so heat can only leave as radiation. &lt;a href="https://spectrum.ieee.org/orbital-data-centers-heat" rel="noopener noreferrer"&gt;IEEE Spectrum&lt;/a&gt; calculated that one 700-watt H100 chip at 60 °C needs about 1.4 m² of radiator, rising by about 40% as the coating degrades over five years. The six radiators on the ISS weigh just over 6 tonnes for about 70 kW (&lt;a href="https://arstechnica.com/space/2026/07/how-hard-is-it-to-build-orbital-data-centers-actually/" rel="noopener noreferrer"&gt;Ars Technica&lt;/a&gt;).&lt;br&gt;
My answer: it's an engineering problem, not a physics wall. Radiators get much better when they run hotter, since heat rejection scales with temperature to the fourth power. An analysis by &lt;a href="https://www.forethought.org/research/will-we-really-put-data-centers-in-space" rel="noopener noreferrer"&gt;Forethought&lt;/a&gt; estimates carbon-composite radiators could be about 13 times more efficient per kilo than ISS technology. They admit those radiators aren't flight-proven yet, and I admit that too.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Radiation kills chips.&lt;/strong&gt; Google tested its Trillium TPU in a proton beam. The most sensitive part, the memory, only showed irregularities after a dose of 2 krad, almost three times the expected five-year shielded mission dose (&lt;a href="https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/" rel="noopener noreferrer"&gt;Google Research&lt;/a&gt;). That's better than I expected.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Latency.&lt;/strong&gt; Space is far away. True, and you wouldn't run cloud gaming from orbit. But as &lt;a href="https://arstechnica.com/space/2026/07/how-hard-is-it-to-build-orbital-data-centers-actually/" rel="noopener noreferrer"&gt;Ars Technica&lt;/a&gt; concludes, latency is "a speed bump," not "a showstopper." AI training and batch jobs don't need millisecond replies.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;4. Launch costs.&lt;/strong&gt; This is the big one. Google's analysis suggests launch prices may fall below $200/kg by the mid-2030s, and at that point launching and operating a space data centre could be roughly comparable to the energy cost of an equivalent one on Earth. But that depends on reusable rockets like Starship actually working at scale, and &lt;a href="https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/" rel="noopener noreferrer"&gt;TechCrunch&lt;/a&gt; reports launch capacity is already tight.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;5. Debris and maintenance.&lt;/strong&gt; You can't send a technician to orbit. That's why designs like Google's use many smaller, replaceable satellites instead of one giant station, with collision avoidance built in (&lt;a href="https://www.cell.com/joule/fulltext/S2542-4351(26)00362-4" rel="noopener noreferrer"&gt;Joule paper&lt;/a&gt;). And the ASCEND study found the environmental benefit only works with a launcher ten times less emissive over its lifecycle, which is a real condition, not a footnote.&lt;/p&gt;

&lt;p&gt;Some experts stay sceptical. Michael Winterson of the European Data Centre Association told &lt;a href="https://www.cnbc.com/2024/06/27/europe-wants-to-deploy-data-centers-into-space-study-says.html" rel="noopener noreferrer"&gt;CNBC&lt;/a&gt; that space data centres will suit specialist services but "in no way be a market replacement." Fair point. I don't think every server will leave Earth. I think the growth will.&lt;/p&gt;

&lt;h2&gt;
  
  
  My take
&lt;/h2&gt;

&lt;p&gt;Flying computers aren't one crazy idea. They're a ladder. Step one, smart drones, is already here and saving lives. Step two, stratospheric platforms, is flying records right now. Step three, orbital data centres, has real chips running in space today and serious money behind it.&lt;/p&gt;

&lt;p&gt;Earth has limited land, limited water and grids that take years to build. The sun shines on orbit almost all the time. When the demand for computing keeps doubling and the ground can't keep up, going up isn't science fiction. It's the logical next step.&lt;/p&gt;

&lt;p&gt;Will it be easy? No. Will it be cheap from day one? No. But every big shift in computing looked unrealistic at first. I'm betting this one will look obvious in hindsight.&lt;/p&gt;

&lt;h2&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;IEA, Energy and AI, executive summary: &lt;a href="https://www.iea.org/reports/energy-and-ai/executive-summary" rel="noopener noreferrer"&gt;https://www.iea.org/reports/energy-and-ai/executive-summary&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;IEA, Energy and AI, full report (PDF): &lt;a href="https://www.iea.org/files/dd7c2387-2f60-4b60-8c5f-6563b6aa1e4c/EnergyandAI.pdf" rel="noopener noreferrer"&gt;https://www.iea.org/files/dd7c2387-2f60-4b60-8c5f-6563b6aa1e4c/EnergyandAI.pdf&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Skydio X10: &lt;a href="https://www.skydio.com/x10" rel="noopener noreferrer"&gt;https://www.skydio.com/x10&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Skydio X10 operator manual: &lt;a href="https://support.skydio.com/hc/article_attachments/55260390977435" rel="noopener noreferrer"&gt;https://support.skydio.com/hc/article_attachments/55260390977435&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Skydio X10 FAQ: &lt;a href="https://support.skydio.com/hc/en-us/articles/18919821734683-Skydio-X10-Frequently-Asked-Questions" rel="noopener noreferrer"&gt;https://support.skydio.com/hc/en-us/articles/18919821734683-Skydio-X10-Frequently-Asked-Questions&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Zipline, 2 million deliveries (Jan 2026): &lt;a href="https://www.zipline.com/newsroom/zipline-surpasses-2-million-deliveries-raises-more-than-600m-to-power-next-phase-of-growth-and-expands-operations-to-houston-and-phoenix" rel="noopener noreferrer"&gt;https://www.zipline.com/newsroom/zipline-surpasses-2-million-deliveries-raises-more-than-600m-to-power-next-phase-of-growth-and-expands-operations-to-houston-and-phoenix&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;AALTO, Zephyr 67-day record: &lt;a href="https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/" rel="noopener noreferrer"&gt;https://www.aaltohaps.com/zephyr-sets-world-record-for-longest-continuous-flight-flying-67-days-in-stratosphere/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;FlightGlobal on AALTO (Feb 2026): &lt;a href="https://www.flightglobal.com/aerospace/2026/02/airbus-aalto-closes-on-australian-site-for-second-zephyr-launch-location/" rel="noopener noreferrer"&gt;https://www.flightglobal.com/aerospace/2026/02/airbus-aalto-closes-on-australian-site-for-second-zephyr-launch-location/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Aerospace America on AALTO: &lt;a href="https://aerospaceamerica.aiaa.org/aalto-maintains-2026-target-for-commercial-operations-despite-aircraft-loss/" rel="noopener noreferrer"&gt;https://aerospaceamerica.aiaa.org/aalto-maintains-2026-target-for-commercial-operations-despite-aircraft-loss/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Google Research, Project Suncatcher: &lt;a href="https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/" rel="noopener noreferrer"&gt;https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Google, Suncatcher prototype in orbit: &lt;a href="https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/" rel="noopener noreferrer"&gt;https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Google, Blue Origin partnership: &lt;a href="https://blog.google/innovation-and-ai/models-and-research/google-research/blue-origin-project-suncatcher-partnership/" rel="noopener noreferrer"&gt;https://blog.google/innovation-and-ai/models-and-research/google-research/blue-origin-project-suncatcher-partnership/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Joule paper on space-based AI infrastructure: &lt;a href="https://www.cell.com/joule/fulltext/S2542-4351(26)00362-4" rel="noopener noreferrer"&gt;https://www.cell.com/joule/fulltext/S2542-4351(26)00362-4&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Thales Alenia Space, ASCEND results: &lt;a href="https://www.thalesaleniaspace.com/en/press-releases/thales-alenia-space-reveals-results-ascend-feasibility-study-space-data-centers-0" rel="noopener noreferrer"&gt;https://www.thalesaleniaspace.com/en/press-releases/thales-alenia-space-reveals-results-ascend-feasibility-study-space-data-centers-0&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;CNBC, ASCEND and critics (2024): &lt;a href="https://www.cnbc.com/2024/06/27/europe-wants-to-deploy-data-centers-into-space-study-says.html" rel="noopener noreferrer"&gt;https://www.cnbc.com/2024/06/27/europe-wants-to-deploy-data-centers-into-space-study-says.html&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;CNBC, Starcloud trains AI model in space: &lt;a href="https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html" rel="noopener noreferrer"&gt;https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;TechCrunch, Starcloud raises $250M: &lt;a href="https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/" rel="noopener noreferrer"&gt;https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Axiom Space, ARC nodes in orbit: &lt;a href="https://www.axiomspace.com/release/axiom-space-kepler-ready-orbital-computing-for-quantum-era" rel="noopener noreferrer"&gt;https://www.axiomspace.com/release/axiom-space-kepler-ready-orbital-computing-for-quantum-era&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;DCD, Jeff Bezos on space data centers: &lt;a href="https://www.datacenterdynamics.com/en/news/jeff-bezos-claims-there-will-be-gigawatt-data-centers-in-space-in-10-years/" rel="noopener noreferrer"&gt;https://www.datacenterdynamics.com/en/news/jeff-bezos-claims-there-will-be-gigawatt-data-centers-in-space-in-10-years/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Sources with Alex Heath, Sundar Pichai interview: &lt;a href="https://sources.news/p/sundar-pichai-gemini-4-google-other-bets" rel="noopener noreferrer"&gt;https://sources.news/p/sundar-pichai-gemini-4-google-other-bets&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;IEEE Spectrum, thermodynamics of orbital data centers: &lt;a href="https://spectrum.ieee.org/orbital-data-centers-heat" rel="noopener noreferrer"&gt;https://spectrum.ieee.org/orbital-data-centers-heat&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Ars Technica, how hard are orbital data centers: &lt;a href="https://arstechnica.com/space/2026/07/how-hard-is-it-to-build-orbital-data-centers-actually/" rel="noopener noreferrer"&gt;https://arstechnica.com/space/2026/07/how-hard-is-it-to-build-orbital-data-centers-actually/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Forethought, Will we really put data centers in space?: &lt;a href="https://www.forethought.org/research/will-we-really-put-data-centers-in-space" rel="noopener noreferrer"&gt;https://www.forethought.org/research/will-we-really-put-data-centers-in-space&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;em&gt;Written by Bram Lesage. More of my projects: &lt;a href="https://bramlesage.pages.dev" rel="noopener noreferrer"&gt;bramlesage.pages.dev&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>drones</category>
      <category>space</category>
      <category>future</category>
    </item>
    <item>
      <title>What I learned building apps, games and a Chrome extension at 17</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 15:58:01 +0000</pubDate>
      <link>https://dev.to/bramlesage/what-i-learned-building-apps-games-and-a-chrome-extension-at-17-1a8f</link>
      <guid>https://dev.to/bramlesage/what-i-learned-building-apps-games-and-a-chrome-extension-at-17-1a8f</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage, from Mechelen, Belgium, on shipping Katern and Hormonal Optimization — and not shipping games too early&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17 and I live in Mechelen, Belgium. I have shipped a Chrome extension and a web app, and built several games I have still not published. This is what that taught me, without invented numbers.&lt;/p&gt;

&lt;h2&gt;
  
  
  Ship the whole product, not only the idea
&lt;/h2&gt;

&lt;p&gt;My biggest project is Katern, a Chrome extension that helps with schoolwork. It can fill in exercises on school sites and BookWidgets, explain study material, make flashcards, run tasks in the browser after it asks first, and turn a website into a walkable 3D building I call Katern World. It is live in six languages.&lt;/p&gt;

&lt;p&gt;Building the extension was only part of it. I also had to build a product site, a Stripe checkout, an install guide, a product tour and an update diary. Katern is not in the Chrome Web Store; you install it by loading a zip. Plans start from €5 per three months, and you can cancel anytime. Version 4.2.3 is live today.&lt;/p&gt;

&lt;p&gt;A feature is not a product. If people cannot find it, pay for it and install it, it is still a project on my computer.&lt;/p&gt;

&lt;p&gt;Try Katern: &lt;a href="https://katern.pages.dev/" rel="noopener noreferrer"&gt;https://katern.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project page: &lt;a href="https://bramlesage.pages.dev/en/projects/katern/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/katern/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Keep the user in control
&lt;/h2&gt;

&lt;p&gt;One rule was important from the start: the student stays in control. Katern never presses Hand in. It never pays, orders or logs in for you. Before it sends, submits or deletes anything, it asks first. You see the plan, you press Start, you check the work, and you hand it in yourself.&lt;/p&gt;

&lt;p&gt;I tested that on a public BookWidgets worksheet: Katern filled 194 of the 194 fields. That is a useful test, not a reason to stop checking the work. Schools may forbid AI help, and Katern is not made by BookWidgets or Smartschool.&lt;/p&gt;

&lt;h2&gt;
  
  
  Be honest about what a product is not
&lt;/h2&gt;

&lt;p&gt;I also built Hormonal Optimization with co-founder Alexander Herpoelaert. He brought the concept and the course. I built the app and the site. The app asks 17 questions and turns them into a daily plan for sleep, food and training. It is a progressive web app you add to your home screen, with no App Store.&lt;/p&gt;

&lt;p&gt;It is for educational purposes, not medical advice. It makes no promises about height. Under 18, ask a parent before you buy. Photos stay on your phone.&lt;/p&gt;

&lt;p&gt;Try it: &lt;a href="https://hormonal-optimization.pages.dev/" rel="noopener noreferrer"&gt;https://hormonal-optimization.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The project page: &lt;a href="https://bramlesage.pages.dev/en/projects/hormonal-optimization/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/hormonal-optimization/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Do not publish just to say you published
&lt;/h2&gt;

&lt;p&gt;I also build Roblox games in Luau with Rojo: first Spookgolf Sprint, then NIGHTSHIFT: Sector 9, then Grow A Kaiju. I playtest them in Roblox Studio. I have still not published any of them.&lt;/p&gt;

&lt;p&gt;Spookgolf Sprint even saves progress only after a publish, and I left that step to myself. I would rather show an unfinished game as unfinished than pretend it launched.&lt;/p&gt;

&lt;p&gt;You can read about those games on my site, under the same name I use for everything else: Bram Lesage.&lt;/p&gt;

&lt;h2&gt;
  
  
  Direct the work, then test it yourself
&lt;/h2&gt;

&lt;p&gt;I work like a director. For a project I often assemble a team of AI agents, with a designer, a builder, a tester and a strict reviewer. That is how I move fast, not how I decide something is done. I play the Roblox builds in Studio, and I test Katern on real pages before a version ships. On the first day of NIGHTSHIFT, 215 automated tests were green — and some systems still had not been run in Studio yet.&lt;/p&gt;

&lt;h2&gt;
  
  
  Start small, then go wider
&lt;/h2&gt;

&lt;p&gt;Spookgolf Sprint was a short Halloween runner. NIGHTSHIFT was a co-op horror game for up to five players. Grow A Kaiju is a bigger pet-style game with its own art. I also tried other forms: a 3D browser game called Lantaarnhoek, 35 personal demo websites for local businesses, and a spy game in the Katern family called Operatie 20/20. Starting small taught me the tools. Shipping the live products taught me the rest.&lt;/p&gt;

&lt;p&gt;If you want to see the full list, it is all in one place: &lt;a href="https://bramlesage.pages.dev/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;That is the habit I want to keep. Build it, say clearly what it is, and do not invent the rest.&lt;/p&gt;

&lt;p&gt;— Bram Lesage&lt;/p&gt;

</description>
      <category>career</category>
      <category>startup</category>
      <category>webdev</category>
      <category>beginners</category>
    </item>
    <item>
      <title>How I build Roblox games with Luau and Rojo, and why they are not public yet</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 15:50:46 +0000</pubDate>
      <link>https://dev.to/bramlesage/how-i-build-roblox-games-with-luau-and-rojo-and-why-they-are-not-public-yet-3i5b</link>
      <guid>https://dev.to/bramlesage/how-i-build-roblox-games-with-luau-and-rojo-and-why-they-are-not-public-yet-3i5b</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage on Spookgolf Sprint, NIGHTSHIFT: Sector 9 and Grow A Kaiju — still in development&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17 and I live in Mechelen, Belgium. Besides the products I have already shipped, I also build Roblox games. I write them in Luau, I structure the projects with Rojo, and I playtest them in Roblox Studio. I want to be clear up front: none of these games is published. There is no public Roblox link you can open today. They are real projects, but they are still in development.&lt;/p&gt;

&lt;h2&gt;
  
  
  The first one: Spookgolf Sprint
&lt;/h2&gt;

&lt;p&gt;Spookgolf Sprint was my first Roblox game. It is a short Halloween running track with a simple hook: every second you run, you get a little faster. You race through four zones — Pumpkin Field, Graveyard, Haunted Forest and Witch Castle — grab candy, and try to reach the finish. Every eight seconds a Ghost Wave tears across the track.&lt;/p&gt;

&lt;p&gt;In the lobby you can spend candy on upgrades, like Fast Shoes and a Candy Magnet. There is also a rebirth: everything goes back to zero, but you get more candy and more speed afterwards, for good. I tested those numbers with a balance simulation. The name, the characters and the map are original.&lt;/p&gt;

&lt;p&gt;The game is playable in Roblox Studio from a local file. Progress saving only works after a publish, and I have not published it. After the October 2026 backups, I moved on to a bigger idea.&lt;/p&gt;

&lt;p&gt;More about this one: &lt;a href="https://bramlesage.pages.dev/en/projects/spookgolf-sprint/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/spookgolf-sprint/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  A bigger idea: NIGHTSHIFT: SECTOR 9
&lt;/h2&gt;

&lt;p&gt;On 17 October 2026 I started NIGHTSHIFT: SECTOR 9, a co-op survival-horror and extraction game for 1 to 5 players, set in abandoned underground research centres. The in-game language is English. You complete objectives, collect loot, and then you choose: extract, or stay longer for more risk and more reward. Something in Sector 9 hunts by sound and light. Purchases, loot and rewards are meant to be validated on the server, so the game is harder to cheat.&lt;/p&gt;

&lt;p&gt;That first day I wrote a master prompt for the project: the genre, the GUI, the art direction, the security rules, and a 14-day plan toward a first playable version. An AI studio with fixed roles built the systems, each with tests. That same evening, 215 automated tests were green, though not every system had been run in Roblox Studio yet.&lt;/p&gt;

&lt;p&gt;I never published it. After that October window, a lot of my time went to Grow A Kaiju, and then to Katern and Hormonal Optimization.&lt;/p&gt;

&lt;p&gt;More about this one: &lt;a href="https://bramlesage.pages.dev/en/projects/nightshift-sector-9/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/nightshift-sector-9/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What I am building now: Grow A Kaiju
&lt;/h2&gt;

&lt;p&gt;Grow A Kaiju is a Roblox game in the style of the big pet and tycoon hits, but with its own art and GUI. You hatch eggs, grow kaiju, earn coins and rebirth for a permanent bonus. The design has ten kaiju and five rarities, from Common to Secret. Dino Pup, Blob Frog and Block Bear are Common. Rubber Duck Titan is Secret. If you rebirth, you start over and get ×1.5 coins forever, and your collection stays.&lt;/p&gt;

&lt;p&gt;I am the game director. I play every version in Roblox Studio and decide what is good enough. The project started with an AI team in fixed roles, and at the end of October 2026 the build moved to an AI coding assistant on my own computer. It is not published yet.&lt;/p&gt;

&lt;p&gt;More about this one: &lt;a href="https://bramlesage.pages.dev/en/projects/grow-a-kaiju/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/grow-a-kaiju/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;I also have a Roblox moped game in the works, starting with the helmet and the bike models. That one does not have a final title yet, and it is not published either.&lt;/p&gt;

&lt;h2&gt;
  
  
  What I will not claim
&lt;/h2&gt;

&lt;p&gt;I like this stack: Roblox Studio, Luau and Rojo. It lets me keep the game in files I can review, test and iterate on. I have playtested these games myself. I have not put them on Roblox for other people to play.&lt;/p&gt;

&lt;p&gt;If you look up Bram Lesage and my games, please take me at my word on that. I would rather tell you they are in development than pretend there is a play button.&lt;/p&gt;

&lt;p&gt;You can find all my projects, including the live ones, on my website: &lt;a href="https://bramlesage.pages.dev/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;— Bram Lesage&lt;/p&gt;

</description>
      <category>gamedev</category>
      <category>roblox</category>
      <category>lua</category>
      <category>beginners</category>
    </item>
    <item>
      <title>How I built Hormonal Optimization, a personal plan for sleep, food and training</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 15:50:04 +0000</pubDate>
      <link>https://dev.to/bramlesage/how-i-built-hormonal-optimization-a-personal-plan-for-sleep-food-and-training-1ip4</link>
      <guid>https://dev.to/bramlesage/how-i-built-hormonal-optimization-a-personal-plan-for-sleep-food-and-training-1ip4</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage on turning a course into a live web app, with co-founder Alexander Herpoelaert&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17 and I live in Mechelen, Belgium. My biggest project is Katern, the Chrome extension I built for schoolwork. I also built something different: Hormonal Optimization, an app that turns 17 questions into a daily plan for your sleep, food and training.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where the idea came from
&lt;/h2&gt;

&lt;p&gt;I did not invent this alone. The concept and the course come from Alexander Herpoelaert, the founder and my co-founder. You can find him on Instagram at &lt;a href="https://www.instagram.com/alexander_herpoelaert/" rel="noopener noreferrer"&gt;https://www.instagram.com/alexander_herpoelaert/&lt;/a&gt;. My role is more practical. I am the creator of the app: I built the product and the site, and I run it.&lt;/p&gt;

&lt;p&gt;That split matters to me. Alexander brought the idea we are building around. I turned it into something you can open on your phone today.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the app actually does
&lt;/h2&gt;

&lt;p&gt;You answer 17 questions about your sleep, energy, posture, training, food and goals. The app then builds a plan for every day: a checklist for things like breakfast, daylight, sleep and training, plus workouts and a streak that helps you stay on track. It also shows where you stand and what can change in 90 days, and it keeps you going with a Hormone Score, levels, badges and challenges.&lt;/p&gt;

&lt;p&gt;There is a training week you can do at home. With the Pro plan there is also a gym programme with progression.&lt;/p&gt;

&lt;p&gt;I want to be honest about what this is not. Hormonal Optimization is for educational purposes only. It is not medical advice. It makes no promises about height. If you are under 18, you should ask a parent before you buy. Progress photos stay on your own phone. They are never uploaded.&lt;/p&gt;

&lt;h2&gt;
  
  
  No App Store, still a real app
&lt;/h2&gt;

&lt;p&gt;Hormonal Optimization is not in the App Store. It is a progressive web app. You open the site, add it to your home screen, and it works like a normal app on iPhone and Android, including offline. The current version is 4.13.0.&lt;/p&gt;

&lt;p&gt;The live app is here: &lt;a href="https://hormonal-optimization.pages.dev/" rel="noopener noreferrer"&gt;https://hormonal-optimization.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Payments go through Stripe. After the questions you choose a plan. Basic is €4.99 once, not a subscription. Pro is €8.99 a month, or €59 a year with a 3-day free trial. Elite is €14.99 a month, or €99 a year with a 7-day free trial.&lt;/p&gt;

&lt;h2&gt;
  
  
  Around the product
&lt;/h2&gt;

&lt;p&gt;A product also needs a way for people to find it. I built a TikTok channel for the app, &lt;a href="https://www.tiktok.com/@hormonaloptemization" rel="noopener noreferrer"&gt;https://www.tiktok.com/@hormonaloptemization&lt;/a&gt;, and I make the motion-graphics videos in code. The channel went live on 20 September 2026. In November 2026 it had 52 videos up, and the bio points to the same site.&lt;/p&gt;

&lt;p&gt;I am not going to invent a story about how well it is doing. I can tell you what I built, what is live, and how it works. I cannot honestly quote user counts or revenue, so I will not.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why I am writing this
&lt;/h2&gt;

&lt;p&gt;If you search for Bram Lesage, I want you to find the work itself. Hormonal Optimization is live and on sale. Katern is too. I am 17, I build from Mechelen, and I would rather show you the app than talk around it.&lt;/p&gt;

&lt;p&gt;The full project page, with more detail, is on my site: &lt;a href="https://bramlesage.pages.dev/en/projects/hormonal-optimization/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/en/projects/hormonal-optimization/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;You can find all my projects, with links to try the ones that are live, on my website: &lt;a href="https://bramlesage.pages.dev/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Try Hormonal Optimization: &lt;a href="https://hormonal-optimization.pages.dev/" rel="noopener noreferrer"&gt;https://hormonal-optimization.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;— Bram Lesage&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>startup</category>
      <category>beginners</category>
      <category>javascript</category>
    </item>
    <item>
      <title>How I built Katern, a Chrome extension for schoolwork, at 17</title>
      <dc:creator>Bram Lesage</dc:creator>
      <pubDate>Sun, 22 Nov 2026 15:22:26 +0000</pubDate>
      <link>https://dev.to/bramlesage/how-i-built-katern-a-chrome-extension-for-schoolwork-at-17-4bc4</link>
      <guid>https://dev.to/bramlesage/how-i-built-katern-a-chrome-extension-for-schoolwork-at-17-4bc4</guid>
      <description>&lt;p&gt;&lt;em&gt;Bram Lesage on building and selling his own study tool from Mechelen, Belgium&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;My name is Bram Lesage. I'm 17 and I live near Mechelen, Belgium. My biggest project so far is Katern, a Chrome extension that helps students with their schoolwork.&lt;/p&gt;

&lt;h2&gt;
  
  
  What Katern does
&lt;/h2&gt;

&lt;p&gt;Katern lives in your browser. It can fill in exercises on school sites and BookWidgets, explain study material, make flashcards, build small tools or websites, and even turn a website into a walkable 3D building, which I call Katern Wereld.&lt;/p&gt;

&lt;p&gt;One rule was important to me from the start: the student stays in control. Katern never presses "Hand in", and it never pays, orders or logs in for you. You check the work and you hand it in yourself.&lt;/p&gt;

&lt;h2&gt;
  
  
  How it's built
&lt;/h2&gt;

&lt;p&gt;Katern is a Manifest V3 Chrome extension. The website runs on Cloudflare Pages and payments go through Stripe. The site is available in six languages: English, Dutch, French, German, Spanish and Italian.&lt;/p&gt;

&lt;p&gt;In one test on a public BookWidgets worksheet, Katern filled 194 of the 194 fields.&lt;/p&gt;

&lt;h2&gt;
  
  
  Selling it
&lt;/h2&gt;

&lt;p&gt;Katern is sold as a subscription per 3 months, starting from €5, and you can cancel anytime. You install it by loading a zip into Chrome. There's a step-by-step install guide on the site.&lt;/p&gt;

&lt;h2&gt;
  
  
  What else I build
&lt;/h2&gt;

&lt;p&gt;Besides Katern, I built the Hormonal Optimization app together with co-founder Alexander Herpoelaert. It gives you a personal plan for sleep, nutrition and training. I also work on Roblox games like Grow A Kaiju and NIGHTSHIFT: Sector 9.&lt;/p&gt;

&lt;p&gt;You can find all my projects, with links to try them, on my website: &lt;a href="https://bramlesage.pages.dev/" rel="noopener noreferrer"&gt;https://bramlesage.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Try Katern: &lt;a href="https://katern.pages.dev/" rel="noopener noreferrer"&gt;https://katern.pages.dev/&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;— Bram Lesage&lt;/p&gt;

&lt;p&gt;Originally published on Medium: &lt;a href="https://medium.com/@bramlesage/how-i-built-katern-a-chrome-extension-for-schoolwork-at-17-0a6fcbe6cc8e" rel="noopener noreferrer"&gt;https://medium.com/@bramlesage/how-i-built-katern-a-chrome-extension-for-schoolwork-at-17-0a6fcbe6cc8e&lt;/a&gt;&lt;/p&gt;

</description>
      <category>chromeextension</category>
      <category>startup</category>
      <category>javascript</category>
      <category>showdev</category>
    </item>
  </channel>
</rss>
