Skip to content
@autor3search

autor3search

Autoresearch for your codebase — your coding agent proposes optimizations, a frozen harness measures them, KEEP or DISCARD.

autor3search

Autoresearch for your codebase. Point your coding agent at your repo, come back to faster code.
autor3search.dev


Your agent proposes an optimization. A frozen harness measures it. The verdict is KEEP or DISCARD — every experiment measured, nothing taken on faith.

Tests are frozen and restored before every run. The metric lives in a compiled binary the agent cannot reach. Noise is discarded, regressions are rejected. Accepted changes land as one commit each; every experiment, including the failures, lands as one row in results.tsv. Stop the run whenever you like — nothing already accepted is thrown away.

Pick your language

repository install status
Go autor3search/go go install …@latest early but working
TypeScript autor3search/typescript npm install --save-dev autor3search-typescript on npm · measures your *.bench.ts benchmarks
JavaScript autor3search/javascript npm install -g autor3search-javascript on npm · standalone, measures vitest benchmarks
Python autor3search/python uv tool install autor3search-python early but working — validated on humanize
Java autor3search/java release jar + launcher early but working — validated on org.json
C# autor3search/csharp dotnet tool install -g autor3search-csharp on nuget · early but working
Rust autor3search/rust cargo install --locked autor3search-rust on crates.io · measures with criterion
Swift autor3search/swift brew install autor3search/tap/autor3search-swift v0.1.0 · 10.98× on apple/swift-asn1 — a local clone, not upstreamed

Same loop, same rules in every port. They differ only in how the harness installs and what it must not touch.

Start a run

Copy the agent prompt for your language — the ## Start here block of that repo's README, or the whole set at autor3search.dev/llms-full.txt — and paste it into your coding agent from inside the repository you want to optimize.

Does it work?

One overnight run, p<0.0001. Real measurements, never illustrations. Validated on go-humanize, mapstructure and google/uuid.


MIT. Takes after karpathy/autoresearch.

Popular repositories Loading

  1. go go Public

    Autonomous AI-driven performance optimization for any Go repository

    Go

  2. typescript typescript Public

    Autonomous AI-driven performance optimization for any TypeScript repository.

    TypeScript

  3. python python Public

    Autonomous AI-driven performance optimization for any Python repository

    Python

  4. java java Public

    Autonomous AI-driven performance optimization for any Java repository

    Java

  5. javascript javascript Public

    Autonomous AI-driven performance optimization for any JavaScript repository

    JavaScript

  6. csharp csharp Public

    Autonomous AI-driven performance optimization for any .NET repository

    C#

Repositories

Showing 10 of 10 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…