wanshuiyin/Auto-claude-code-research-in-sleep
11,118 stars · Last commit 2026-06-01
ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works with Claude Code, Codex, OpenClaw, or any LLM agent.
README preview
# Auto-claude-code-research-in-sleep (ARIS ⚔️🌙)
<p align="center">
<a href="https://huggingface.co/papers/2605.03042">
<img src="docs/hf_daily_paper_1.svg" alt="Hugging Face Daily Paper · #1 Paper of the Day" width="360">
</a>
</p>
[](https://huggingface.co/papers/2605.03042) · [](https://wanshuiyin.github.io/Auto-claude-code-research-in-sleep/ARIS_INTRO.html) · [](docs/aris_intro_slides.pdf) · [](AGENT_GUIDE.md) · [](https://mp.weixin.qq.com/s/tDniVryVGjDkkkWl-5sTkQ) · [](https://github.com/VoltAgent/awesome-agent-skills) · [-orange?style=flat)](https://aidigitalcrew.com) · [](https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep/stargazers) · [💬 Join Community](#community) · [](#citation)
💡 *Use ARIS as a skill-based workflow in [Claude Code](https://docs.anthropic.com/en/docs/claude-code) / [Codex CLI](skills/skills-codex/) / [Cursor](docs/CURSOR_ADAPTATION.md) / [Trae](docs/TRAE_ARIS_RUNBOOK_EN.md) / [Antigravity](docs/ANTIGRAVITY_ADAPTATION.md) / [GitHub Copilot CLI](docs/COPILOT_CLI_ADAPTATION.md) / [OpenClaw](docs/OPENCLAW_ADAPTATION.md), or get the full experience with the standalone CLI — enjoy any way you like!*
🌱 *ARIS is a methodology, not a platform. What matters is the research workflow — take it wherever you go.*
🔥 *ARIS natively fits — and already supports — any agent's ultracode-style deep mode: its **breadth** pass adapts to whatever a runtime exposes — Claude Code's ultracode / dynamic workflows on Opus 4.8 (xhigh, or max effort if budget allows), Codex `spawn_agent` / equivalents, or another model entirely — degrading cleanly **parallel fan-out** → **agent spawn** → **plain sequential**.*
*ultracode supplies firepower to the breadth half ARIS always had, giving three clean roles: **depth → breadth**, **cross-model review → accuracy**, **research wiki → memory**.*
*However a loop is driven — ultracode breadth or goal-mode persistence — every loop reports to the same cross-model jury + research wiki: it can drive, never acquit.*
🤖 **AI agents:** Read [`AGENT_GUIDE.md`](AGENT_GUIDE.md) instead — structured for LLM consumption, not human browsing.