AutoResearchClaw turns ideas into papers

This is the kind of project that pushes agentic research from demo to workflow: not just summarizing papers, but generating evidence, running code, and pruning fake citations. If it works as advertised, it is less a writing assistant and more a research operations stack.

• –The repo claims a 23-stage pipeline with OpenAlex, Semantic Scholar, and arXiv lookups, plus sandboxed experiments and citation verification.
• –It is opinionated about output quality: multi-agent peer review, statistical charts, and NeurIPS/ICML/ICLR-ready LaTeX all ship in the loop.
• –The self-evolving angle matters more than the paper generation itself; lessons from failed runs are fed back into later runs, which is where real compounding could happen.
• –The open-source release lowers the barrier for labs and hobbyists to try autonomous research on their own topics.