skyllwt/AutoSci

Karpathy's LLM-Wiki vision, fully realized — wiki-centric full-lifecycle AI research platform powered by Claude Code

What it solves

AutoSci is designed to automate the traditionally human-intensive scientific research lifecycle. It reduces the manual effort required to coordinate literature reviews, idea generation, experimental design, execution, and manuscript writing across long-term projects.

How it works

AutoSci operates as a memory-centric agentic system. It uses structured persistent memory to maintain context across projects and improves its own procedures over time. The system integrates with agent runtimes like Claude Code, Codex, and OpenCode to provide a suite of specialized research skills. These skills allow the agent to ingest papers, ideate research directions, design and run experiments (both locally and remotely), evaluate results, and even generate conference posters from LaTeX source.

Who it’s for

It is built for researchers and scientists who want to AI-assisted automation for the full research pipeline, from initial literature discovery to the final publication process.

Highlights

  • Full Lifecycle Automation: Covers everything from paper ingestion and daily arXiv recommendations to manuscript writing and rebuttal.
  • Memory-Centric Design: Maintains a compounding research memory across multiple projects.
  • Experimental Workflow: Includes structured paths for ideation, pilot runs, formal experiment design, and guided method iteration (Hiera).
  • Multi-Runtime Support: Compatible with Claude Code, Codex, and OpenCode.
  • Visualization Tools: Provides a Web UI and Obsidian integration for exploring the research knowledge graph.
  • Automated Poster Generation: Converts LaTeX drafts into print-ready conference posters.

Related

  • Project
    synthetic-sciences/openscience

    OpenScience is an open‑source AI research workbench that lets users describe scientific tasks in plain language. The built‑in agent plans, searches literature, writes and runs code (Python/R/shell), executes experiments on local or remote compute, and produces a reproducible report, with every step logged. Available as a desktop app, web workspace, or CLI, it supports custom LLMs, a large library of domain‑specific skills, and extensible plugins.

  • Project
    EvoMap/AutoResearch

    An open-source agent workflow for AI/ML research that automates idea generation and experiment execution to produce paper-ready evidence.

  • Project
    EvoScientist/EvoSkills

    A repository of installable skill packs that extend AI agents with specialized workflows for the full academic research lifecycle, from ideation and experimentation to paper writing.

  • Project
    OpenLAIR/dr-claw

    A full-stack AI research workspace that automates the end-to-end academic lifecycle from literature surveys and ideation to experimentation and paper writing.

  • Project
    microsoft/ResearchStudio

    An AI co-author toolkit that automates the research lifecycle, from generating reviewer-defensible research ideas to creating posters, videos, and blogs from finished papers.