grandamenium

Autoresearch Anything — Data skill for Claude Code

Data community

Claude Code skill: autonomous experimentation pipeline for any business metric using Karpathy's autoresearch pattern.

How to install Autoresearch Anything

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

What Autoresearch Anything does

Claude Code skill: autonomous experimentation pipeline for any business metric using Karpathy's autoresearch pattern.

Alternatives in Data

HTML Anything — ✨ The agentic HTML editor — your local AI agent writes the HTML, you ship it 8.5k ★
Skill Production Pipeline — Claude-skills — Effective: 2026-03-07 Applies to ALL new skills, improvements, and deployments 5.3k ★
Pipeline Feature — Add a feature to the FFmpeg transcode pipeline following Tunarr's pipeline builder pattern 2.5k ★

README

Autoresearch-Anything

A Claude Code skill that applies Andrej Karpathy's autoresearch pattern to ANY business metric with an objectively verifiable success metric. Build an autonomous experimentation pipeline that modifies, measures, keeps or discards, and compounds improvements - all while you sleep.

Quick Start

# Clone into your Claude Code skills directory
git clone https://github.com/jamesgoldbach/autoresearch-anything.git ~/.claude/skills/autoresearch-anything

# Invoke the skill in Claude Code
/autoresearch-anything

What It Does

This skill guides you through building a custom autonomous experiment pipeline for any measurable business outcome. Inspired by Karpathy's autoresearch repo (where an AI agent optimizes ML models overnight), this brings the same pattern to cold email reply rates, landing page conversions, ad CTR, newsletter open rates, and anything else you can measure with an API.

The skill handles:

Education - Explains the autoresearch pattern and how it applies to business
Structured Q&A - Scopes your specific pipeline (metric, platform, constraints, execution)
Deep Research - Researches your domain before the first experiment
Scaffold Generation - Creates a complete project with program.md, config, and connector stubs
Guided Build - Helps you build the real measurement and deployment connectors
Persistence Setup - Configures local (launchd + tmux) or cloud (GitHub Actions) execution
Launch - Gets the autonomous loop running

How It Works

The core loop follows Karpathy's exact pattern:

LOOP FOREVER:
1. Read context (results history, accumulated learnings, knowledge files)
2. Hypothesize (what change might improve the metric?)
3. Modify the experiment file
4. Git commit
5. Deploy
6. Wait (measurement window)
7. Measure (query the metric via API)
8. Keep if improved, discard if not (git reset + redeploy previous)
9. Log learnings (both positive and negative)
10. Repeat