SOTA Public Research Database + Search Tool

Project summary

Recently I released https://exopriors.com/scry. It's a novel search tool over arXiv, LessWrong, Hacker News, community-archive.org, and other sources that seem worth prioritizing for operators in the software intelligence explosion. Basically, it's a highly refined Claude Code prompt, for having your agent copilot utilize our hardened, well-indexed, embedded, high-performant readonly-SQL + vector database, to answer your exceptionally nuanced questions.

It's such a powerful paradigm of search, you can quickly use it to genuinely update your worldview. It's not just about finding really nuanced documents, but often, not finding them, because you can just do exhaustive search, and that negative evidence can be even more valuable. It's also extremely powerful for navigating and exploring embedding space, going far deeper than simple nearest-neighbor lookups of common RAG.

What are this project's goals? How will you achieve them?

This project's goal is to uplift people operating in the software intelligence explosion by giving them access to a SOTA intelligence and research tool. I will further achieve that by expanding operations including:

getting a bigger server;
scraping, cleaning, embedding, indexing a large fraction of all publicly available academic papers and dozens of additional sources;
increasing the ergonomics over real-world use cases;
better advertising the tool to the extended community, perhaps experimenting with small bounties for the most interesting queries;
and talking with users to understand their uncertainty about how to benefit from this incredible paradigm of search.

In further building out and giving more people access to this tool, an additional benefit is I will have a concrete artifact that will help my foundational epistemic infrastructure startup compete with more short-termist companies for VC funding that I will absolutely need to stay competitive in the software intelligence explosion in 2026.

How will this funding be used?

I'm considering this to be a roughly 5 month operation, and can steward it longer if it's getting good use.

I need 5 months of a large Hetzner server (extreme price efficiency, supporting world-class search and experiments for up to ~40 concurrent heavy users). $1547/month * 5.

I need competitively priced embeddings over ~250 billion tokens. Voyage-3.5-lite at $.02/M means roughly $5000.

I will have a stipend of $3000 for the first month, then $1000 for maintenance.

And then I will have a $1k advertising budget.

If only partially funded, we can trim costs up to about 50% (20% less RAM, 50% less embeddings, run for a month or two less).

Who is on your team? What's your track record on similar projects?

Just me. I shipped https://exopriors.com/scry within 2 weeks of being unlocked by Opus 4.5, and the project has just been continuously improved for the community, while on an extremely limited budget.

What are the most likely causes and outcomes if this project fails?

People are going to have a useful, ergonomic, SOTA tool. I am going to advertise it and show off some of the very interesting questions it's able to answer. If the project fails to answer thousands of nuanced user questions and really highlight this powerful paradigm (spawning other versions in the community), it'll probably be because we didn't advertise it smart enough to counter all the other shiny creations being created in the software intelligence explosion.