CodeLayers – See your codebase's dependency layers in 3D
Hey HN, I'm Long. I started building CodeLayers in November — a 3D code visualization app that started on Apple Vision Pro and is now on iPhone and iPad. Why I built this: AI agents are writing more code than ever, and I realized I had no idea what my codebase actually looked like anymore. I wanted a way to see the architecture at a glance — what depends on what, where changes ripple, where the complexity is hiding. And I wanted it on my phone, not buried in some CI dashboard. But getting the visualization right was harder than I expected. Force-directed graphs were the obvious first…
What it does
In the maker’s words, at launch
Hey HN, I'm Long. I started building CodeLayers in November — a 3D code visualization app that started on Apple Vision Pro and is now on iPhone and iPad. Why I built this: AI agents are writing more code than ever, and I realized I had no idea what my codebase actually looked like anymore. I wanted a way to see the architecture at a glance — what depends on what, where changes ripple, where the complexity is hiding. And I wanted it on my phone, not buried in some CI dashboard. But getting the visualization right was harder than I expected. Force-directed graphs were the obvious first attempt. They look cool for 20 files. At 500+ it's an unreadable hairball — positions are random and it looks different every reload. No meaning to where anything sits. City view was next — files as buildings, metrics as height. Impressive in screenshots but useless for understanding how code actually connects. You see size, not relationships. BFS from foundations is what finally clicked. I run BFS upward from depth-0 files (files that import nothing — your utils, types, constants). Each file's depth is the longest path from any foundation. Cycles get collapsed via Kosaraju's SCC detection. The result: layers emerge naturally. Position has meaning. Bottom = bedrock everyone depends on. Top = entry points. You're looking at real architecture, not a physics simulation. And it tells you something useful: depth predicts blast radius. A change at depth 0 can ripple through hundreds of files. A change at depth 5 probably touches nothing else. Two features I'm most excited about: Remote AI agents from your phone. You can point Claude, Gemini, or Codex at your repo and ask questions about your code while looking at the 3D graph. Waiting for a build? Pull out your phone, ask the agent "what calls this function?" and see it highlighted spatially. Dead time becomes time understanding your codebase. Watching your graph grow. The CLI syncs in real-time as you code. Watch your architecture evolve — see when a new dependency adds a layer, catch when a refactor creates a cycle, notice when a file is becoming a god object that everything imports. In a world where AI agents are making PRs, being able to see the effect of changes on your codebase at a glance is the thing I care about most. Privacy was a hard requirement — all source code is encrypted on-device with zero-knowledge encryption before it leaves your phone. I can't read your code. Free to explore public repos. Pro unlocks working with your own repositories — private repo support, CLI watch mode, git time travel, and AI agents. App Store: https://apps.apple.com/app/codelayers/id6756067177 How the layering algorithm works: https://codelayers.ai/blog/the-hidden-hierarchy-in-your-code... Built this solo in about 3 months. Would love feedback — especially on what you'd want from a tool like this.
Does the same job
all alternatives →- VAVisualizing a Codebase2021 · octo.github.com · ▲283
I explored an alternative way to view codebases to the typical folder/file list, showing a bird's-eye-view of its structure. https://octo.github.com/projects/repo-visualization
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
