{"id":801,"date":"2025-12-01T14:31:28","date_gmt":"2025-12-01T06:31:28","guid":{"rendered":"https:\/\/www.itsmetoo.com\/?p=801"},"modified":"2025-12-01T14:31:28","modified_gmt":"2025-12-01T06:31:28","slug":"from-high-school-dropout-to-sora-researcher-how-gabriel-petersson-mastered-ai-with-projects-and-prompts","status":"publish","type":"post","link":"https:\/\/www.itsmetoo.com\/index.php\/2025\/12\/01\/from-high-school-dropout-to-sora-researcher-how-gabriel-petersson-mastered-ai-with-projects-and-prompts\/","title":{"rendered":"From High School Dropout to Sora Researcher: How Gabriel Petersson Mastered AI with Projects and Prompts"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

23, never finished high school.<\/strong>\u200bNo university diploma on his resume.But in OpenAI\u2019s internal system, his title reads: Gabriel Petersson, Research Scientist (Sora Team).<\/strong>\u200bAfter dropping out of high school, Gabriel worked at Depict.ai, Dataland, and Midjourney before officially joining OpenAI in December 2024 to research video generation.No degree. But a string of projects.On November 28, 2025, he shared his learning method on the podcast Extraordinary<\/em>: When facing a new field, he never starts with thick textbooks. Instead, he tackles real problems, lets ChatGPT break them into steps, writes code, debugs, then learns backward\u2014from the problem to math, algorithms, and papers.<\/strong>\u200bDo first, learn later.<\/strong>\u200b His approach: Learning AI with AI.<\/em>And he\u2019s not alone.Last month, OpenAI CEO Sam Altman (also a dropout) said he envies today\u2019s dropout generation. With AI tools, they can build useful products faster, while traditional diplomas are losing their edge.A Swedish high school dropout walked from a classroom to Sora\u2019s lab by seizing this era\u2019s biggest opportunity:Diplomas are depreciating. Skills are being revalued. Learning is being rewritten by AI.<\/strong>\u200bSo how did he do it?<\/p>\n\n\n\n

\n\n\n\n

Part 1 | Selling Pok\u00e9mon Cards at 14, Joining OpenAI at 23<\/h2>\n\n\n\n

Many assume he\u2019s a coding prodigy, a straight-A student with competition medals.Reality? No degree, no famous mentors, no connections.But he\u2019s always done one thing: built projects, hit roadblocks, then iterated.<\/strong>\u200bFirst pivot: Age 14<\/strong>\u200bLiving in rural Sweden, he earned his first $20K by reselling Pok\u00e9mon cards. Wanting a price-comparison website, he Googled \u201chow to build a webpage\u201d on YouTube. No teachers, no pressure\u2014just him teaching himself to build his first tool.\u201cI wasn\u2019t a programmer. I just wanted something, so I built it.\u201dSecond pivot: COVID-19 year<\/strong>\u200bAt 18, he built a hand sanitizer price-comparison site in a week, earning $22K in its first week. Pandemic demand surged, and his simple scraper + frontend solved a real problem.Months later, he was recruited as CTO of Curb Food (Sweden\u2019s largest cloud kitchen, 80 employees). He built a 7-person engineering team from scratch, developed a kitchen management system, coded and debugged on a sofa.The project launched. He left. But he learned one thing: Success or failure doesn\u2019t matter. What matters is the product you leave behind.<\/strong>\u200bThird pivot: Discovering Midjourney<\/strong>\u200bIn 2023, he stumbled upon AI-generated art and was blown away. That day, he dove headfirst into AI.Not a design expert, he created one of Midjourney\u2019s best UI tools: fast-grid<\/em>. All self-taught: How to arrange the interface? How to structure prompts? How to tweak model parameters? Trial after trial.The tool got copied, recommended, and shared. Some called him a hacker, but his goal was simple: Make Midjourney easier to use.<\/strong>\u200bNo credits, no certificates\u2014but real outputs: Products built, models fixed, bugs squashed.OpenAI wanted exactly that:<\/p>\n\n\n\n

    \n
  • \u2022A working GitHub project.<\/li>\n\n\n\n
  • \u2022A demo solving real problems.<\/li>\n\n\n\n
  • \u2022Someone who actually\u00a0does<\/em>things.<\/li>\n<\/ul>\n\n\n\n

    He never stepped into a university. But he walked into the world\u2019s top AI lab, step by step.<\/p>\n\n\n\n

    \n\n\n\n

    Part 2 | Learn by Doing, Not Learn Then Do<\/h2>\n\n\n\n

    Traditional AI video research demands a PhD.But Gabriel relied on real, executable GitHub projects\u2014not diplomas.<\/strong>\u200bHe admits he initially didn\u2019t understand diffusion models or even Transformer attention mechanisms.But instead of opening textbooks, he opened ChatGPT.<\/strong>\u200bHe asked specific questions: \u201cIf I want to build a small video-generation model, what\u2019s the first thing to learn?\u201d<\/em>ChatGPT answered. He dug deeper: \u201cCan you give an example? Code? Explain each line? Pretend I\u2019m 12 and explain again?\u201d<\/em>Then he copied the code into his project, tested it. Hit bugs? Screenshotted errors, iterated until fixed.No fixed steps. No standard answers. Just one goal: Fix this bug in front of me.<\/strong>\u200bGabriel calls this:Recursive learning from the task.<\/strong>\u200bStart with the problem, reverse-engineer the knowledge chain, fill gaps layer by layer.Traditional path (bottom-up):Learn math \u2192 probability \u2192 neural networks \u2192 then understand diffusion models.Gabriel\u2019s path (top-down):Diffusion model bug \u2192 trace to attention \u2192 need math \u2192 learn concepts \u2192 fix code.Result? \u201cWith this method, I mastered diffusion models in three days. The traditional path takes six years.\u201d<\/em>Not theoretical understanding\u2014but hands-on parameter tuning, practical application.This ChatGPT-powered reverse learning gave him PhD-level skills\u2014in days, not years.\u201cMost people ask ChatGPT once and quit. I keep asking until I truly understand.\u201d<\/em>From tasks, intuition, and errors\u2014he traced back knowledge.AI is the best teacher. But you have to chase the answers.<\/p>\n\n\n\n

    \n\n\n\n

    Part 3 | Not Just Using ChatGPT, but Researching with It<\/h2>\n\n\n\n

    Sora is one of OpenAI\u2019s most complex multimodal models.Generating each video frame requires orchestrating billions of parameters, multi-module coordination, and spatiotemporal reasoning.This isn\u2019t about chatting for an image. It\u2019s real AI R&D: training models, tuning parameters, coding, debugging. Gabriel does this daily.His typical workflow:<\/p>\n\n\n\n

      \n
    1. 1.Observe Sora\u2019s generated videos, spot inconsistencies.<\/li>\n\n\n\n
    2. 2.Hypothesize the flawed architecture layer, prompt GPT-4 to analyze.<\/li>\n\n\n\n
    3. 3.Review\/edit core module code based on suggestions.<\/li>\n\n\n\n
    4. 4.Debug with ChatGPT or cross-check papers\u2014retrain, evaluate visual results.<\/li>\n\n\n\n
    5. 5.If wrong, ask again, retry, retrain.<\/li>\n<\/ol>\n\n\n\n

      Often, his chats with GPT aren\u2019t for answers\u2014but to refine ideas.He briefs GPT like a human colleague: \u201cI adjusted these parameters in this module. The generated video has issues\u2014I think the model overlooks details, maybe needs more temporal info\u2026\u201d<\/em>GPT responds with two paths:<\/p>\n\n\n\n

        \n
      • \u2022Quick-fix:<\/strong>\u200b Tweak weights, retrain directly.<\/li>\n\n\n\n
      • \u2022Structural-change:<\/strong>\u200b Add mechanisms (e.g., depth-wise conv).<\/li>\n<\/ul>\n\n\n\n

        Gabriel decides which to try.He makes the final call. He writes the code.But GPT is his second brain<\/strong>: Fearless in trials, tireless, unfrustrated, broad in coverage, always responsive.In traditional research, a PhD student waits for advisors, reference code, or lab meetings to proceed.Gabriel outsources these to GPT\u2014a 24\/7 research partner.\u201cI\u2019m not beating PhDs with ChatGPT. I\u2019m using it as a researcher.\u201d<\/em>The difference isn\u2019t the model\u2019s intelligence\u2014but how you use it.<\/strong>\u200b<\/p>\n\n\n\n

        \n\n\n\n

        Part 4 | Diplomas Don\u2019t Matter. These 3 Things Do<\/h2>\n\n\n\n

        Gabriel downplays the \u201cdropout success\u201d label. In interviews, he clarifies:\u201cI didn\u2019t get into OpenAI by proving I\u2019m special. I showed them I\u2019m already doing what they need.\u201d<\/em>So what did he do right? Three things:1. Start with projects, not books<\/strong>\u200bNever asks \u201cWhich concept to learn first?\u201d<\/em>Instead: \u201cWhat problem should I solve now?\u201d<\/em>Midjourney frustrating? Build a UI tool.Don\u2019t get diffusion models? Start with a video demo.Want to research Sora? Build a simplified version and tune it.Learn by doing. Every project fills a knowledge gap.2. Use AI to accelerate understanding, not skip it<\/strong>\u200bNot using GPT to write code\u2014but to understand it.<\/strong>\u200bAsk until concepts click. Debug with GPT analyzing each line. Let GPT summarize key paper changes.Most use AI to avoid thinking. He uses AI to enhance thinking.<\/strong>\u200b3. Let projects speak, not diplomas<\/strong>\u200bNo degree, no awards, no recommendation letters for OpenAI. But he submitted:<\/p>\n\n\n\n

          \n
        • \u2022A working GitHub project (fast-grid<\/em>).<\/li>\n\n\n\n
        • \u2022Engineering experience at Midjourney.<\/li>\n\n\n\n
        • \u2022A self-built video-generation pipeline.<\/li>\n<\/ul>\n\n\n\n

          Not what he learned<\/em>\u2014but what he built.<\/em>OpenAI hired him not for reciting \u201ctransformer\u201d terms\u2014but for building a working transformer.<\/strong>\u200bThey want problem-solvers, experiment-drivers, result-producers\u2014not noun-memorizers.Many call it an \u201cunderdog success.\u201d But it reflects a bigger trend:Diplomas prove you attended class. Projects prove you did the work.<\/strong>\u200bFollowing a syllabus is step-by-step. Following problems is real skill.AI breaks knowledge barriers. With AI, you can learn anytime. Whether you learn<\/em>depends on how you ask.In the AI era, scarcity isn\u2019t knowledge\u2014but questioning ability and self-driven learning.<\/strong>\u200bFuture hiring won\u2019t ask \u201cWhere did you graduate?\u201d<\/em>but \u201cCan you use AI to create real results?\u201d<\/em><\/p>\n\n\n\n

          \n\n\n\n

          Epilogue | Works Speak Louder Than Exams<\/h2>\n\n\n\n

          Gabriel\u2019s story isn\u2019t about rebels beating the system.It\u2019s a replicable AI-era learning path:Start with projects. Accelerate with AI. Prove with works.<\/strong>\u200bNot an exception\u2014but a paradigm shift:<\/p>\n\n\n\n

            \n
          • \u2022Not\u00a0learn then do<\/em>\u2014but\u00a0do while learning.<\/em><\/li>\n\n\n\n
          • \u2022Not\u00a0wait for diplomas<\/em>\u2014but\u00a0let projects speak.<\/em><\/li>\n\n\n\n
          • \u2022Not\u00a0use AI to write<\/em>\u2014but\u00a0use AI to deeply understand.<\/em><\/li>\n<\/ul>\n\n\n\n

            He says: \u201cI\u2019m not a genius. I just changed how I learn\u2014using AI to learn AI.\u201d<\/em>Now, works matter more than diplomas.<\/strong><\/p>\n","protected":false},"excerpt":{"rendered":"

            23, never finished high school…<\/p>\n","protected":false},"author":2,"featured_media":802,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[9,8,6],"tags":[217,32,218,216],"class_list":["post-801","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-companies","category-deep-tech","category-people","tag-high-school","tag-openai","tag-scientist","tag-sora"],"_links":{"self":[{"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/posts\/801","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/comments?post=801"}],"version-history":[{"count":1,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/posts\/801\/revisions"}],"predecessor-version":[{"id":803,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/posts\/801\/revisions\/803"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/media\/802"}],"wp:attachment":[{"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/media?parent=801"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/categories?post=801"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.itsmetoo.com\/index.php\/wp-json\/wp\/v2\/tags?post=801"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}