Use Claude to build a Lean runtime for web targets

A developer had Claude help create Lasm, a Lean 4 runtime that can run across Node, browser, serverless, edge, and WASI setups.

Done with Claudeยท Advanced

What they did
They directed Claude to work inside the Lasm repository, which already included agent-focused files like AGENTS.md, llms.txt, and an AI rules document. The project compiles Lean 4 ahead of time into WebAssembly so it can run in Node, edge, serverless, WASI, and browser environments.
How it went
The Conduit sample app reached 117 proved theorems and passed the official API suite, with all 13 files succeeding, plus the Playwright end-to-end tests. The tradeoffs were that Lean on Node ran slower than a standard Node HTTP setup, and the edge and browser versions stayed single-threaded.
Tips
Set a clear target environment list before the agent starts, and have it pause for review before any destructive or irreversible change. Expect a large first-time tool download, and pin Lean 4.35.0 with Node 24 or newer.

Try it yourself with Claude

I want to build [runtime/library/tool] for [browser, Node, serverless, edge, WASI]. First, outline a compact architecture and test plan, then pause and wait for my approval. After I approve, make the changes in small steps, run the relevant tests after each step, and stop before any irreversible action such as deleting files, submitting changes, or altering published releases. At the end, report what works, what is still missing, and any performance limits you found.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on github.com โ†—

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.