Curated index of autonomous AI research frameworks, self-improving agents, and experiment loops inspired by karpathy/autoresearch. Use when looking for autoresearch tools, autonomous ML experiment runners, self-improving agent architectures, or research automation frameworks.
A curated, high-signal index of autonomous improvement loops, research agents, and descendants inspired by karpathy/autoresearch.
Frameworks that generalize autoresearch to any measurable metric:
| Tool | Description |
|---|---|
| kayba-ai/recursive-improve | Recursive self-improvement with execution traces and keep-or-revert evaluation |
| uditgoenka/autoresearch |
| Claude Code skill generalizing autoresearch for software, docs, security, debugging |
| supratikpm/gemini-autoresearch | Gemini CLI skill with Google Search grounding and 1M token context |
| zkarimi22/autoresearch-anything | Any measurable metric: prompts, API perf, landing pages, test suites, SQL |
| jmilinovich/goal-md | GOAL.md pattern where agents construct fitness functions before optimizing |
| james-s-tayler/lazy-developer | Orchestrates autoresearch across prioritized optimization goals |
| Entrpi/autoresearch-everywhere | Cross-platform auto-detection and loop startup |
| greyhaven-ai/autocontext | Closed-loop control plane with evaluation, persistent knowledge, staged validation |
End-to-end autonomous research pipelines:
| Tool | Description |
|---|---|
| SakanaAI/AI-Scientist | First comprehensive system for fully automatic scientific discovery |
| SakanaAI/AI-Scientist-v2 | Workshop-level discovery via agentic tree search |
| HKUDS/AI-Researcher | NeurIPS 2025 - Full automation: hypothesis to manuscript to peer review |
| OpenRaiser/NanoResearch | Plans experiments, generates code, runs jobs, writes papers |
| SamuelSchmidgall/AgentLaboratory | End-to-end: idea to literature review to experiments to report |
Hardware-specific forks of autoresearch:
| Tool | Description |
|---|---|
| ShengranHu/ADAS | ICLR 2025 - Meta-agents that invent novel agent architectures |
| MaximeRobeyns/self_improving_coding_agent | SICA - Agent that edits its own codebase |
| gepa-ai/gepa | ICLR 2026 Oral - Genetic-Pareto reflective prompt evolution |
| MrTsepa/autoevolve | GEPA-inspired self-play with Elo/Bradley-Terry rating |
| metauto-ai/HGM | Huxley-Godel Machine for SWE-bench self-improvement |