Xyra

good idea, I emailed them. And no search engine, only Scry or direct web fetch of an URL in a Scry result or from the agent's memory.

thank you!

let me know if it was faster and more compositionally expressive than you expected

thanks, I try

Good point, fixed

Hi, yes. New users right now get free credit. When the server isn't burdened, queries are currently ~free (under $.01 per second, when most are well under a second). We're just trying to learn here how to use this thing productively together.

As load increases, the price goes up. There's also exponential egress pricing because Scry is a place where computation happens, not a metaphorical torrent. I'm still tweaking things, balancing between multi-user server load, giving normal active human users lots of priority and no fear in using it hard and deliberately, and fair pricing for agents and automated heavy workflows people set up in the background.

that's a good idea. it's very doable

thanks. I'm still sorting out pricing. I don't have experience with B2B and am interested in empowering out-of-distribution people, so will likely maintain some free tier + congestion-based auction pricing for non-commercial use.

I've been working on grep.it.com. can currently grep over 200 TB of the internet

hi, i've been working on this, grep.it.com. (i found your comment with it)

making the internet sql queryable, crawling, cleaning, indexing, embedding many sources into source-aware schemas, solving contention problem with free-floating pricing.

Currently crawling over 1M records/sec. software is still in alpha.

scry.io.

goal is a 10PB NVMe cluster online by November (need funding champions) as a public benefit project, so prosocial researchers and builders and their agents can have low-friction access to running analytical queries over the public internet.

Hetzner, Postgres, Rust, SvelteKit

What did you think?

emailed you, and it's https://venmo.com/u/XyraSinclair.

We can iterate fast with understanding useful paradigms of vector manipulation. Yesterday I added `debias_vector(axis, topic)` and l2_normalization guidance.

Thank you! I got the idea December 3, and initially released it December 19.

I'm raising at least $175k and doing a serious startup.

Thanks, that's very interesting.

~300 token chunks right now. Have other exciting embedding strategies in the works.

The scale is there. I'm scraping, cleaning, token efficientizing dozens of sources every single hour. The lack of monies for embedding everything was a temporary problem.

in the direction of "empowering the public with new capabilities they didn't have before", Scry offers, with the copy and paste of a prompt and talking with an agent:

1) Full readonly-SQL + vector manipulation in a live public database. Most vector DB products expose a much narrower search API. Basically only a few enterprise level services let you run arbitrary SQL on remote machines. Google BigQuery gives users SQL power, but it mostly doesn't have embeddings, connect public corpora, have as good of indexes, and doesn't have support an agentic research experience. Beyond object-level research, Scry a good tool for exploring and acquiring intuitions about embedding-space.

2) An agent-native text-to-SQL + lexical + semantic deep research workflow. We have a prompt that's been heavily optimized for taking full advantage of our machine and Claude Code for exploration and answering nuanced questions. Claude fires off many exploratory queries and builds towards really big queries that lean on the SQL query planner. You can interrupt at any time. You have the compute limits to do lots of exhaustive exploration--often more epistemically powerful than finding a document often, is being confident than one doesn't exist.

3) dozens of public commons in one database, with embeddings.

Thank you! I'll be getting millions more quality, embedded documents, it'll be here just getting more useful.

Thank you!

You submit a SQL query to periodically run, we run it and store the results. As we ingest more documents (dozens of sources are being ingested every day), we run it again. If there's different outputs, you get an email.

Maybe more actually, server costs and API credits for my agent-coordination research are expensive.

Exactly, people want precision and control sometimes. Also it's very hard to beat SQL query planners when you have lots of material views and indexes. Like this is a lot more powerful for most use cases for exploring these documents than if you just had all these documents as json on your local machine and could write whatever python you wanted.

Yeah I've out a lot of care into rate-limiting and security. We do AST parsing and block certain joins, and Hacker News has not bricked or overloaded my machine yet--there's actually a lot more bandwidth for people to run expensive queries.

As for getting good semantic queries for different domains, one thing Claude can do besides use our embed endpoint to embed arbitrary text as a search vector, is use compositions of centroids (averages) of vectors in our database, as search vectors. Like it can effortlessly average every lesswrong chunk embedding over text mentioning "optimization" and search with that. You can actually ask Claude to run an experiment averaging the "optimization" vectors from different sources, and see what kind of different queries you get when using them on different sources. Then the fun challenge would be figuring out legible vectors that bridge the gap between these different platform's vectors. Maybe there's half the cosine distance when you average the lesswrong "optimization" vector with embed("convex/nonconvex optimization, SGD, loss landscapes, constrained optimization.")

Thank you, I've started ingestion operations of pubmed.

What is hyperbole? We are collectively experiencing a software intelligence explosion (people are shipping good software at prolific rates now due to Opus 4.5 and GPT-5.2-Codex-xhigh). With Scry, you can run arbitrary SELECT SQL statements over a large corpus and have an easier time composing embedding vectors in whatever mathematical ways you want, than any other tool I've seen.

I've since improved it, and also discovered a new method of vector composition I have added as a first-class primitive:

debias_vector(axis, topic) removes the projection of axis onto topic: axis − topic * (dot(axis, topic) / dot(topic, topic))

That preserves the signal in axis while subtracting only the overlap with topic (not the whole topic). It’s strictly better than naive subtraction for “about X but not Y.”

Yes, thanks for explaining it.