2026

• Rather Than Publishing Papers, We Should Publish Agents

  Why a static PDF is no longer enough, and how the Agentic Publication Protocol (APP) turns each paper into a versioned open-source repository with its own dedicated AI agent.

  X

  Xiao-Liang Qi

  

  S

  Sirui Lu

  Sep 20, 2026

  11 min read

  ai agents scientific-publishing reproducibility open-science app

• FormalFlow: Long-Horizon Autoformalization with AI Agents

  How AI agents and human supervision produced a 126k-line Lean proof of the low individual-degree test, and what we learned about statement drift, review gates, and long-horizon proof composition.

  

  S

  Sirui Lu

  Y

  Yanqiao Zhu

  Sep 20, 2026

  ai lean formalization multi-agent research texra

• A Few Personal Thoughts on AI and Science

  Watching the Millennium Prize Problems fall one by one to AI, from a seat fairly near the front — on test-time scaling, formal proof, and what AlphaGo taught us about our own field.

  

  S

  Sirui Lu

  Sep 14, 2026

  3 min read

  ai science go lean formalization multi-agent research texra