The pitch was refreshingly blunt. "Thought the resources for GPU arch were lacking, so here we are."
That's how Jason—known on Hacker News by the handle Jaso1024—introduced Mvidia, a browser-based puzzle game where players wire together a graphics processing unit from scratch, starting with individual transistors. Posted to the site's "Show HN" forum earlier this spring, the project rocketed to the front page within hours, collecting 907 points and 179 comments. The game has since been compared to cult favorites like Turing Complete and NANDGame, beloved among programmers and hardware hobbyists for their uncompromising approach to teaching computer architecture through interactive play.
But there's a twist. Jason built much of Mvidia with significant assistance from Claude, an AI coding assistant—a detail he volunteered early in the discussion thread and one that set off its own flurry of side debates about what it means to develop software in 2026.
Transistors, Then Memory, Then… What Exactly?
Mvidia lives at jaso1024.com/mvidia, free to play and requiring nothing but a modern browser. The game starts players at the atomic level of digital logic: wiring up NMOS and PMOS transistors to form gates. From there, the curriculum ascends through NAND, NOR, and NOT configurations before landing on memory fundamentals—DRAM wordlines, bitlines, the headache-inducing reality of refresh cycles.
Interspersed are what Jason calls "minigames": quick drills on binary-to-hex conversions, truth table memorization, the kind of rote exercises that feel tedious until you realize you've internalized them. There's an endless mode that awards stars for speed and accuracy. And after early feedback that the difficulty curve was too punishing, Jason added seven introductory levels based on that feedback.
The architecture hints at grander ambitions. In one comment thread, Jason mentioned he'd previously built an "act 0" covering the physics of transistor behavior, then scrapped it for being too granular. He plans to resurrect it. What's less clear is where the game goes after memory—shader cores, SIMT execution models, the parallel-processing magic that separates GPUs from CPUs. That's the part the community seems most eager to see, though Jason has been noncommittal about timelines.
AI as Silent Partner
Jason's candor about using Claude "quite a bit"—his words—was unusual enough to spark discussion. He credited the AI particularly with wiring and simulation systems, the kind of fiddly work that can bog down solo developers for weeks. Some commenters bristled at the idea; others shrugged and asked to see the code. (No public repository exists yet.)
What's harder to dismiss is the pace of iteration. When users griped about timed challenges causing frustration, Jason shipped difficulty toggles the same day. Complaints that the leap from transistors to NAND gates was too abrupt? He added those seven intro levels and made the timed exercises optional. A capacitor modeling bug got squashed. Wire routing that was hard to distinguish against the grid background was swapped from lines to dots.
That kind of responsiveness suggests either a developer with unusual bandwidth or one who's figured out how to delegate the grunt work. Perhaps both.
The Shadow of Turing Complete
Hardware education games occupy a narrow but devoted niche. Turing Complete, a Steam title where you design a CPU from logic gates and write your own instruction set, came up again and again in the Hacker News thread. So did NANDGame, a free web simulator that builds computers from NAND gates, and SHENZHEN I/O, Zachtronics' brutally difficult assembly-puzzle game. Several commenters mentioned Nand2Tetris, the MOOC curriculum that's become canonical for anyone trying to understand computation from first principles.
Mvidia's gamble is to do for GPUs what those games did for CPUs. One commenter summed up the appetite: "I want to skip the beginning and get faster into what makes GPUs different"—shader cores, rasterization pipelines, the architectural quirks that make modern graphics cards capable of training neural networks and rendering photorealistic worlds at 60 frames per second.
Those features aren't in the game yet. Whether they ever will be is an open question. Jason has been active in the comments, enthusiastic but vague about future plans. There's no commercial backing that's been disclosed, no venture funding, no educational institution partnership. The game is free, the updates are ad hoc, and the developer seems to be shipping in public while holding down what one assumes is a day job.
When Gamification Works (and When It Doesn't)
Educational simulators walk a tightrope. Too forgiving and they feel like clickable PowerPoints; too punishing and players quit before the lesson lands. The games that succeed—Kerbal Space Program teaching orbital mechanics, Zachtronic's entire catalog drilling algorithmic thinking—nail the balance between rigor and reward.
Mvidia isn't there yet, but the iteration loop is promising. Jason's willingness to adjust difficulty on the fly, to add onboarding without dumbing down the core challenges, suggests he understands the genre. The comparisons to Turing Complete and NANDGame aren't just flattery; they're signals that a small but vocal community sees potential for Mvidia to become a reference tool, something CS educators bookmark and self-taught hardware nerds work through on weekends.
A podcast recap of Hacker News highlights recently featured Mvidia as a standout submission, describing it as "hands-on GPU design learning" and noting the flurry of UX feedback. The HN Arcade listing categorizes it alongside other browser-based learning games—no paywall, no install friction, just a URL and curiosity.
Whether Jason intends to turn this into a full-fledged curriculum or let it remain a side project is anyone's guess. He hasn't said. But the reception—enthusiastic, critical in the constructive sense, surprisingly engaged for a niche technical topic—suggests there's an audience hungry for exactly this kind of thing. In a moment when AI tools are lowering the barrier to building software, maybe the real story is what gets built, and whether it teaches something worth knowing.
For now, Mvidia is free, functional, and iterating in the open. That alone might be enough.
