Lurk Lab

Building a verifiable foundation for AI

[Porçu Quine @porcuquine2]

The revised Lurk paper has been published in the Journal of Functional Programming: https://t.co/U83OHojaXs

We have done so much since then… looking forward to making it public soon.

[Porçu Quine @porcuquine2]

Here's a snapshot of what we're working on at @LurkLab. This also serves as a snapshot of the kind things coding agents can plausibly accomplish these days.

[Nicolas Ramsrud @NicolasRamsrud]

we are quietly shipping, anon

sneak peek at LurkOS, our provable operating system sending 100k transactions between actors

yes, this is all intra-actor and without proofs, but this is just the start.

[Lurk Lab @LurkLab]

👀

[Nicolas Ramsrud @NicolasRamsrud]

is that JS fib running in LurkScript REPL?

Why yes it is.

[Lurk Lab @LurkLab]

LurkScript now works well enough to implement the Lurk standard library so far! Our modular design allows the lurk CLI REPL and prover to use the LurkScript CLI frontend to seamlessly add LurkScript support to Lurk.

https://t.co/HFw6ZYl3yX

[porter @portport255]

@NicolasRamsrud @0xLita Wow, I haven't kept up with Lurk. That's amazing. Okay @LurkLab is going on my study list 🤝

[Nicolas Ramsrud @NicolasRamsrud]

Lurk does something similar with similar results.

All data and code is commitment based, including proof I/O (which is then opened and committed to at each step of computation), so with content-addressed data storage you can generate proofs of computation over arbitrarily sized datasets stored in arbitrary locations. Even private data.

Although we use Poseidon to get const sized commitments, this makes me wonder if there is some PCS method to use as a commitment scheme for proofs over large datasets that would perform better than Poseidon.

[Nicolas Ramsrud @NicolasRamsrud]

A little view behind the curtain, anon.

You can now write provable JavaScript with Lurk.

This shows why Lurk is so powerful. It took just 5 weeks to make it possible to compile JS into a JS representation of Lurk that can either be interpreted in place as JS or passed through Lurk for proving.

Onwards to on-boarding millions of devs to ZK.

[Lurk Lab @LurkLab]

Provable JavaScript? Check.

Users can now write programs in JavaScript, compile to LurkScript, evaluate in place or pass through Lurk to generate a proof.

Getting ready for that Web2 scale ZK.

[Porçu Quine @porcuquine2]

My LurkScript implementation is now minimally feature-complete. This lets you author arbitrary Lurk programs in a (functional, immutable) subset of JavaScript. It's implemented in TypeScript, and will allow tight integration with the JS ecosystem, toolchain, and web application.