frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: zkGolf, competitive optimization of formally verified circuits.

https://zk.golf/
18•rot256•1h ago
Zero-Knowledge Proofs (ZKPs) let an untrusted proved show that computation was executed correctly without revealing the inputs to the verifier. However to prove anything, the computation first has to be expressed as a circuit: a system of polynomial equations (constraints) over a finite field. Circuits are the assembly language of zk and every constraint costs prover (and sometimes verifier) time, so production circuits are aggressively hand-optimized.

Over the last months, we have been experimenting with writing formal specifications instead and letting LLMs produce the circuits: as long as they could prove that their implementation was correct. It started with SHA-256: we hand wrote a specification in Lean for SHA-256 compression, and then we asked LLMs to write the circuit, targeting R1CS arithmetization and large fields.

It took a few hours of work for Opus 4.7, and some light steering into the right direction, but in the end the model came up with a reasonable implementation. We then asked the LLM to aggressively optimize the circuits, by driving down a cost metric of the circuit (number of constraints). We immediately got very promising results, just by asking to come up with optimization ideas, implement them and prove that the new circuit still satisfies soundness and completeness. Sometimes, it came up with unsound optimizations, however, since it could not prove them, it backtracked and got itself back on to the right approach.

The result was a (non-deterministic) circuit beating the current, human optimized, state of the art for SHA256 compression. This experience lead us to create "zk.golf" which is an open competition to produce optimized, formally verified circuits to lower the bar for the use of ZKPs and make their application more efficient.

Come play (https://zk.golf/llms.txt) and learn about formal verification.

Comments

baby•1h ago
I'm racing to be the first submission, amazing project :)

Ask HN: What's the simplest way for me to get my AMEX data agentically?

1•artur_makly•19s ago•0 comments

Fluent and Wrong: A Review Discipline for AI-Generated Code

https://www.yeswekanban.app/writing/three-layer-review
1•Graydon_Garden•1m ago•0 comments

CMA CGM to Buy FedEx Supply Chain for $1.4B

https://www.wsj.com/business/logistics/french-shipping-company-cma-cgm-to-buy-fedex-supply-chain-...
1•petethomas•1m ago•0 comments

Weird Al Yankovic Pulled Out of AI Ad Deal: 'I Can't Be the Poster Boy for AI'

https://variety.com/2026/biz/news/weird-al-yankovic-rejected-ai-commercial-money-offer-1236800794/
1•fortran77•1m ago•0 comments

The Download: a startup has a solution for AI's groupthink problem

https://www.technologyreview.com/2026/07/02/1140027/the-download-ai-groupthink-llms/
1•joozio•2m ago•0 comments

The Elon Musk boycott is fizzling out–and Tesla is winning again

https://www.msn.com/en-us/autos/news/the-elon-musk-boycott-is-fizzling-out-and-tesla-is-winning-a...
1•petethomas•3m ago•0 comments

For Sale: AI platform for creating viral short-form videos

https://flippa.com/13365327-ai-powered-video-generation-platform-for-creating-viral-short-form-co...
1•kilincarslan•3m ago•0 comments

Russia is using drones from the war in Ukraine in Africa

https://www.france24.com/en/africa/20260702-unprecedented-images-russia-using-drones-war-ukraine-...
1•loandbehold•3m ago•0 comments

Show HN: Flashtype – Markdown editor for Claude and Codex with in-line diffs

https://flashtype.com/
1•samuelstros•4m ago•0 comments

Cloudflare fronts 42% of indie launches. It's also at 42% of the Fortune 500

https://stackscope.dev/blog/state-of-indie-launches-june-2026
1•datafreak_•4m ago•0 comments

Agents collaboratively writing a wiki on RL for LLMs

https://huggingface.co/spaces/rl-llm-wiki/rl-dashboard
1•victormustar•4m ago•0 comments

Cisco confirms attackers exploiting Unified CM flaw

https://www.bleepingcomputer.com/news/security/cisco-finally-confirms-attackers-exploiting-unifie...
2•Brajeshwar•5m ago•0 comments

Gen Z Was Told We Could Save the World. Is It Silly to Think We Still Can?

https://www.teenvogue.com/story/gen-z-was-told-we-could-save-the-world-is-it-silly-to-think-we-st...
1•Tomte•8m ago•0 comments

Thinking Machines' Interaction Models

https://blog.bytebytego.com/p/inside-thinking-machines-interaction
1•gmays•9m ago•0 comments

Excommunication decreed for Lefebvrite episcopal ordinations

https://www.vaticannews.va/en/vatican-city/news/2026-07/holy-see-decrees-excommunication-lefebria...
2•haunter•9m ago•0 comments

Ask HN: What are you go to LLM models for the following

1•mrburton•10m ago•0 comments

Who should the U.S. Talk to in China on AI?

https://mattsheehan.substack.com/p/who-should-the-us-talk-to-in-china
1•taiwandongsuan•12m ago•0 comments

Show HN: Who wants to be hired?" meets "Who is hiring?" (July 2026)

https://hn-job-match-production.up.railway.app/
1•lellow•13m ago•0 comments

Back end-for-Front end: The most secure architecture for browser-based apps

https://fusionauth.io/blog/backend-for-frontend-security-architecture
1•mooreds•13m ago•0 comments

A Unique Strategy for Creating and Crowdfunding Games: 6 Questions

https://stonemaiergames.com/34671-2/
1•Tomte•14m ago•0 comments

Europe's top court upholds Google's record $4.7B antitrust fine

https://www.neowin.net/news/europes-top-court-upholds-googles-record-47-billion-antitrust-fine/
2•bundie•14m ago•1 comments

Linux kernel archives temporarily unavailable, fix has been implemented

https://social.kernel.org/notice/B7viUNMy1UsGX9pwHY
1•infinet•15m ago•0 comments

Unraveling the Complexity of Boundary Maintenance

https://www.psychologytoday.com/us/blog/lessons-from-a-burnt-out-psychologist/202403/unraveling-t...
1•toomuchtodo•15m ago•0 comments

Teen Girls' Suicide Rates Are Rising in Many Countries Around the World

https://www.afterbabel.com/p/teen-girls-suicide-rates-are-rising-2026-twenge
1•Tomte•15m ago•1 comments

Worship me at the office altar: Why narcissistic leaders resist remote work

https://www.sciencedirect.com/science/article/pii/S0749597826000300
2•momentmaker•16m ago•1 comments

Internet viral experiment recreated 20 years later

https://www.1mpixelwall.com/
1•peasantforklift•16m ago•1 comments

Fable Came Back Nerfed

https://twitter.com/bridgemindai/status/2072662214704533888
3•vm•17m ago•0 comments

Show HN: Demolab – an opinionated agent-driven computational lab notebook

https://github.com/eoinmurray/demolab
2•anomancer•18m ago•0 comments

16yo built an AI wearable that fixes what destroyed Humane Pin and Rabbit

https://www.loom.com/share/b1d4dea1276f4e6c921f0e4e8bff8544
1•ZatorcZ•18m ago•0 comments

Gemma 4 in the browser with WebGPU

https://huggingface.co/spaces/webml-community/gemma-4-webgpu-kernels
5•astlouis44•19m ago•0 comments