nowfound

Dev tools · April 1, 2026

IM

I made a Mario Galaxy game with Claude Code and Three.js in 53 days

I'm the developer. Happy to share the process and learnings. Architecture: - Custom SoA ECS for cache-friendly iteration, zero allocations in hot loops. Benchmarked against bitECS/miniplex. - Process manager inspired by Game Code Complete's but data-oriented so processes are pure data, systems contain the logic. - Multi-field gravity with 5 types (point, parallel, segment, cylinder, convex hull) with spring-damped transitions between fields. - Rapier3D WASM with SIMD. Gravity-aware fake shadow system for playability. - WebGL 2 shared context between Three.js and Pixi.js. - Debug editor…

What it does

In the maker’s words, at launch

I'm the developer. Happy to share the process and learnings. Architecture: - Custom SoA ECS for cache-friendly iteration, zero allocations in hot loops. Benchmarked against bitECS/miniplex. - Process manager inspired by Game Code Complete's but data-oriented so processes are pure data, systems contain the logic. - Multi-field gravity with 5 types (point, parallel, segment, cylinder, convex hull) with spring-damped transitions between fields. - Rapier3D WASM with SIMD. Gravity-aware fake shadow system for playability. - WebGL 2 shared context between Three.js and Pixi.js. - Debug editor for placing objects, configuring settings, visualizing colliders, etc (Tweakpane + Three.js helpers) 76K lines of TypeScript. 735 commits. Runs on mobile. Plays best with gamepads. The AI Process: Claude Code (Opus) wrote ~95% of the code. I provided architecture, constraints, and direction; looked at some but not much of what it wrote. My process for every feature: braindump what I want, relevant technical details, and "does this make sense?" into the chat. Largely unorganized. I built custom Claude Code skills like /lets-build:plan that spawns sub-agents to research the codebase first, then asks me clarifying questions. We go back and forth until it sounds right, then Claude writes plan documents split into phases so the app stays runnable after each one. As we got closer to ship date, I started having Claude review its own plans. This mostly catches the main issues. Built dedicated skills for that too: /review-plan, /code-review, /retrospective, /ecs-review. Where Claude struggled: - It defaulted to OOP in TypeScript even though the project is data-oriented ECS. Took a lot of steering to overcome. Built /ecs-plan and /ecs-review skills specifically to catch and fix this. - Level design failed completely. I tried making a CLI tool so Claude could help place objects in 3D space where "down" could be anywhere. Tried elevation maps, architectural diagrams; didn't help. - At one point it had an index.ts that was thousands of lines long so ensuring a fast first paint was a disaster. I started refactoring manually but then felt it was a giant waste of my time to do it. So I had Claude map the dependency tree and do the refactor which was a piece of cake for it. The hardest problems were the camera and the gravity shadows. Both work but still have edge cases. Galaxy's camera also has problems; part of that might be inherent to arbitrary gravity directions. The shadows are all fake, cast based on gravity field type, and the letter shapes (M, R) create concave, irregular geometry that Galaxy didn't have for planetoids. 87 plans total, averaging 2-3 pages each. The CLAUDE.md project file is 164 lines of hard constraints learned from debugging sessions. Every constraint has a token massacre behind it. One of the biggest "wins" is having Claude create tools that you would probably never do for a project with a deadline in ~50 days. That time tradeoff (or even impossibility) is no longer valid and this likely compounds even more for longer projects.

Does the same job

all alternatives →
  • MA
    made a html5 game, polished it many many times2011 · hotbazooka.com · ▲211

    hi all, I've always wanted to make my own game, but couldn't afford a thousand-dollar adobe (flash+photoshop) license. I'm not a talented programmer, have no degree in CS. But I understand some elements of design. So I got me a copy of ImpactJS, and after about 5 months of steep learning curve (at least for me), I managed to come up with a game. I also integrated it with a highscore and badge system, similar to what you get at Kongregate. You need to login with Google/OpenID to earn badges/scores. link is: http://www.hotbazooka.com/privatejoe hope to hear some feedback/questions from you guys!

  • IM
  • 3M
    3D multiplayer space shooter made with Three.js2019 · urk-3d-multi-space-game.herokuapp.com · ▲115
  • IS
    I spent 6 years making a cool free game2015 · ▲151

    Hey everyone! Long time lurker, first time poster. I've just reached beta in my game BoxByte Lyte. -What is it- It's a free 3d arcade game for hardcore players that uses color matching in interesting ways. Also it's a one button game, well movement and one button. Cool huh? The final release will be free too. -What I need- This is a very challenging game, so I really need to balance the difficulty so that it's still hard but not impossible. It would be great just to hear your experiences of the game and how far in it you got. -Where can I report bugs?-…

  • AA
  • GJ

More dev tools this month

the category →
  • Dograh592

    The open source VAPI alternative

    Dev tools · 25d ago · dograh.com

  • Meridian530

    Don't let your work go unnoticed. Get promoted!

    Dev tools · 20d ago · meridiona.com

  • x1516

    Lovable for iPhone apps go from idea to App Store

    Dev tools · 11d ago · x1.new

  • Open-source GTM skills for technical founders

    Dev tools · 29d ago · gtmcofounder.com

  • OpenTrailPaper is open-source bike computer firmware for the LilyGO T5S3 4.7" E-Paper PRO. It supports offline maps, GPX routes, FIT recording and Bluetooth sensors.

    Dev tools · 2d ago · opentrailpaper.com

  • Nuphos380

    The AI-Native DevOps Workspace.

    Dev tools · 24d ago · nuphos.ai

Launched alongside, April 2026

the whole month →
  • Brila1,367

    One-page websites from real Google Maps reviews

    AI · Apr 2026 · brila.ai

  • AG

    Thought the resources for GPU arch were lacking, so here we are

    Life & fun · Apr 2026 · jaso1024.com

  • IB

    Built a ~9M param LLM from scratch to understand how they actually work. Vanilla transformer, 60K synthetic conversations, ~130 lines of PyTorch. Trains in 5 min on a free Colab T4. The fish thinks the meaning of life is food. Fork it and swap the personality for your own character.

    AI · Apr 2026 · github.com

  • AI meeting notes: now bot-free, in ChatGPT & Claude + more

    AI · Apr 2026 · fathom.ai

  • BC

    Life & fun · Apr 2026 · sam-burns.com

  • IB

    With social media and now AI, its important to keep the indie web alive. There are many people who write frequently. Blogosphere tries to highlight them by fetching the recent posts from personal blogs across many categories. There are two versions: Minimal (HN-inspired, fast, static): https://text.blogosphere.app/ Non-minimal: https://blogosphere.app/ If you don't find your blog (or your favorite ones), please add them. I will review and approve it.

    AI · Apr 2026 · text.blogosphere.app