Status: wip/braindump, but also feel free to comment!

A 3-month incubator for software people, to hack on EA and AI safety projects. In the spirit of original YC Summer Fellows Program.

Why?

Vibecoding is easy, who even needs software engineers anymore?

Well: It turns out that building apps takes more than coding. Product taste, design, distribution & marketing are all things that 2026 LLMs still fail at. (Not to mention long-term coherence and maintainability within a codebase.)

We believe that there’s a bunch of concrete pain points within the EA & AI safety ecosystem, that good software could solve. And a whole bunch of EAs come from software backgrounds! Unfortunately, too many folks have been psyoped into upskilling into research, or earning-to-give, or grantmaking; we aim to change this.

Also! LLMs are now pretty smart, but they’re still not used that much. What can we now build? What can we now automate?

Timeline

• May 9: launch
  • Submit applications as a Manifund project
    • … or as a microsite?
    • … or as a Manifund PR?
• May 16: early applications due (for review by May 23)
• May 23: all applications due (for review by May 29)
• Jun 12-14: Manifest
• Jun 15: Kickoff (?)
  • <10 weeks of programming>
• Aug 21: Demo Day

Example projects

take these more as inspiration, than literal projects to immediately build

• microsites like AI 2027, CAIS letter
• single-purpose tools to help communicate info
  • microcovid
  • “what is agi .com”
  • “all about datacenters .com”
• better versions of various marketplaces
  • jobs, like 80k job board
  • funding, like Manifund
  • housing, group house directory like directorysf
  • events, like partiful, EA forum opportunities
• tools for better decisions, AI for epistemics
  • AI powered prediction markets, forecasting
  • new forms of media platforms (like moltbook of forecasting or debate)
  • squiggle