The open source AI research agent.
Daily AI ecosystem digest from 10 sources (GitHub, ArXiv, HN, HuggingFace, Product Hunt, Dev.to, Lobste.rs). Bilingual ZH/EN reports via GitHub Actions.
Mimir — 一站式科研工作台插件:LaTeX 论文边写边编译、arXiv 文献管理、实验追踪、指标图表、GPU 服务器 SSH 任务编排,管理科研全周期。An open-source research workbench plugin for the whole research cycle.
Runtime policy enforcement for AI agents. Cryptographic audit trail, human-in-the-loop approvals, kill switch. Zero code changes.
Talk to any ArXiv paper using ChatGPT
A skill to kill from-scratch coding — Claude checks real arXiv prior art before it designs a new architecture.
Open-source research workspace: search academic papers, upload PDFs, inline citations with on-PDF highlights, reader-mode Q&A, and get citation export in four formats.
A zero-dependency TypeScript memory kernel that gives any agent long-term memory. No training or backpropagation: each output reads a few tag-compatible neighbors, updated by a manual Hebbian rule. Links used often are reinforced; idle ones evaporate by half-life. Based on the paper Local Pheromone Network (arXiv:2606.30669).
Zero-Mem: Zero-Token Memory Operations for the pi coding agent (faithful reimplementation of arXiv:2607.29377)
AI4Scholar for DeepSeek Harness (dsh): 38 native academic tools — Semantic Scholar, PubMed, Google Scholar, arXiv, bioRxiv/medRxiv, DOI, full text, auto-cite, figures, unified search. Powered by ai4scholar.net