edouard-claude

Longe — AI skill for Claude Code

AI community

A self-improving harness for any LLM, in one Rust binary.

How to install Longe

This entry records only its repository, not the path inside it, so there is no exact command to give. Open edouard-claude/longe and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Longe does

A self-improving harness for any LLM, in one Rust binary. Persistent Lua REPL as the single tool, three-level memory, persistent sub-agents, enforced budgets, external verification, per-process sandbox, and post-run reflection over its own store. Built from the YC Harness Club talks.

Alternatives in AI

Gtm Engineer Skills — Claude Code skill for improving website AEO (AI Engine Optimization) and GEO (Generative Engine Optimization) 761 ★
Claudish CLI — for external AI models) 679 ★
Agentty — AI pair programming in your terminal — one static binary, sub-ms startup, any model 597 ★

README

Longe

A self-improving harness for any LLM, in one Rust binary.
The model is the engine; the runtime is the car.

Thesis · Credits · Architecture · Quick start · Lua surface · Sandbox · Benchmark

A *longe* is the long rein used to let a horse run in a circle: it lets the animal go, and it sets the radius and the duration. Longe does the same for a language model. It lets the model run, and it imposes the budget, the verifier and the sandbox.

Longe is a from-scratch implementation of the ideas presented at the **YC Harness Club** evening (Y Combinator, 2026), where four projects made the same argument from four directions: Prime Agent and Continual Harness (Prime Intellect, Princeton), Open Jarvis (Stanford, Hazy Research) and QM (Y Combinator). This README credits those authors, states their thesis, and shows where each idea lives in the code.