March 29, 2026
Docs or drama?
Lat.md: Agent Lattice: a knowledge graph for your codebase, written in Markdown
Map your code with wiki notes—brilliant upgrade or just “more docs”
TLDR: Lat.md proposes a Markdown-based “map” of your code with wiki links and AI‑powered search, aiming to keep knowledge tidy and in sync. Commenters are split between “finally, organized notes the bots can read” and “it’s just the same old docs that will go stale and won’t scale,” with jokes about missing security files.
Lat.md wants you to map your codebase with plain Markdown files linked like a mini‑Wikipedia, complete with a command that checks everything stays in sync and a search that can use AI if you add a key. And the community? They’re split, loud, and very online. Supporters like nimonian say they’re already doing this with VitePress (a docs site generator) and that “agents” (AI helpers) are happy reading raw .md. Skeptics like iddan clap back that this is just reinventing the dusty “docs” folder with an extra logo. The biggest anxiety comes from ssyhape: these knowledge maps “go stale immediately,” but they admit keeping it inside the repo means pull‑request reviews and “git blame” (who changed what, when) could help.
Then there’s the scale drama: thousands of pages of notes could become their own maze. jatins doubts throwing the whole Obsidian‑style notebook at an AI solves anything if the bot can’t sift it smartly. Meanwhile, Yokohiii side‑eyes the package list and jokes that the project’s own security docs are missing.
Through it all, fans argue Lat.md’s wiki links, “lat check,” and searchable sections make code knowledge usable for both humans and bots. Critics say it’s lipstick on docs. The popcorn is hot, the comments are hotter.
Key Points
- •Lat.md (“Agent Lattice”) organizes codebase knowledge into interconnected markdown files under a lat.md/ directory.
- •Sections link via wiki-style syntax and to source code symbols; source files can annotate links back to documentation with @lat comments.
- •The lat CLI validates links and references (lat check) and provides commands for locating, viewing, referencing, searching, and expanding content.
- •Semantic search uses embeddings and requires an OpenAI or Vercel AI Gateway API key configured via environment variables, file, or helper command.
- •Installation is via npm; development requires Node.js 22+ and pnpm, with standard install, build, and test commands.