It's important to mention that this project has received $0 in funding from EAIF, LTFF, Manifund, and other EA/EA-adjacent orgs. That tells you A LOT about the epistemic inclination and technological orientation of funders in the community. If it were not for a random angel investor that just ASKED to invest $5k after this went viral on Hacker News, I would be in a bleak financial situation and plausibly unable to even maintain this tool.

Update: The service has been improving a lot. Infra is more robust, higher query limits, more sources, a further optimized Claude Prompt with much improved guidance for meaningful embedding vector manipulation.  Someone could write some interesting quicktakes from the research you're easily able to do with this.

It's hard to fathom what's straightforwardly possible already.

You're so right, I should have caught that. The database data is good, but Opus 4.5 did a rare error. I've since updated the response.

I fed your message to Scry! Seems the forum is significantly more longtermist.

I just built something that can help donors explore the EA/alignment ecosystem: alignment-explorer.exopriors.com.

It's a read-only SQL + vector search API over a continuously growing alignment-adjacent corpus, that Claude Code w/ Opus 4.5 is exceptionally good at using. The compositional vibe search (vector mixing) functionality combined with long query time limits is quite something.

I am new to EA, but it seems that a true effective altruist would not be interested in retiring. When just a $1000 can avert decades of disability-adjusted life years (years of suffering), I do not think it is fair to sit back and relax (even in your 70's) when you could still be earning to give.