Sibyl-Research-Team

Sibyl-Research-Team / sibyl-research-system

Public

Fully Autonomous AI Research System with Self-Evolution, built natively on Claude Code

151
18
80% credibility
Found Mar 10, 2026 at 113 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Sibyl Research System is an open-source tool that automates end-to-end machine learning research using teams of AI agents to generate ideas, run experiments, write, and review papers.

How It Works

1
🔍 Discover Sibyl

You learn about Sibyl, the smart helper that handles full research projects on its own, turning ideas into complete papers without any coding.

2
đź“Ą Bring it home

Download the free tool and place it on your powerful computer ready for research adventures.

3
🤖 AI sets it all up

Chat with the friendly AI inside to connect your computer's brainpower and online helpers automatically, making everything ready in moments.

4
đź’ˇ Share your research dream

Simply describe the science question you want explored, like better ways to teach computers to see.

5
▶️ Hit go and relax

Press the start button, step back, and let Sibyl's team of clever thinkers search, test, debate, and create.

6
🔄 It learns and gets better

Watch amazed as Sibyl spots issues, fixes them, learns from each step, and keeps improving the work automatically.

đź“„ Enjoy your new paper

Receive a professional research paper, complete with experiments and insights, ready to share with the world.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 113 to 151 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is sibyl-research-system?

Sibyl is a fully autonomous AI research system that handles end-to-end ML projects: from literature surveys and multi-agent idea debates to GPU-parallel experiments, paper writing, and NeurIPS-level reviews—all without human input. Built natively in Python on Claude Code with MCP servers for SSH GPU access and arXiv searches, it iterates ideas, runs pilots/full experiments on remote servers, and outputs compiled LaTeX papers. Users just configure a GPU server, start via CLI commands like `/sibyl-research:start`, and let it self-evolve prompts and strategies across projects.

Why is it gaining traction?

Unlike static tools like AI-Scientist or single-agent scripts, Sibyl's dual-loop design—inner research iteration plus outer self-evolution—automatically refines everything from hypotheses to GPU scheduling based on past failures. Its self-healing fixes runtime errors autonomously, and multi-model reviews (Claude + optional GPT) ensure quality gates hit publication standards. Developers dig the true hands-off autonomy on remote GPUs, with tmux persistence and no permission prompts via Claude flags.

Who should use this?

ML researchers prototyping ideas on shared GPU clusters, like those benchmarking models (e.g., GPT-2 vs. Qwen) or exploring reinforcement learning for mobile manipulation. Academic labs wanting fully autonomous paper drafting for NeurIPS submissions, or indie devs in fully remote setups tired of manual experiment loops. Skip if you lack SSH GPU access or prefer supervised workflows.

Verdict

Worth starring for Claude Code fans—solid docs and setup scripts make it approachable despite 58 stars signaling early maturity (credibility score: 0.800000011920929%). Test on a demo project first; self-evolution shines after 3-5 runs, but monitor costs on heavy models.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.