frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: I applied Markowitz port. theory to agent teams / proved it in a zkVM

https://www.mnemom.ai/showcase
2•alexgarden•1h ago
I run multi-agent teams in high-consequence scenarios. Read: fuckups at 3 AM = I'm awake.

I kept hitting the same issue. I couldn't get a rules-based system to enforce behavior and I had no real way to prove that agents really did what they said they did. I can log and monitor them - set up (a million) Slack alerts but none of these things are PROOF. Logs are mutable. And that matters more every day as agents get more powerful (take THAT, @meta)

So I went down the rabbit hole.

The obvious answer is zero-knowledge proofs. Prove behavior cryptographically. Except proving an LLM inference in a zkVM is computationally Star Trek. Lagrange proved GPT-2 E2E, Polyhedra can do Llama-3 at 150 seconds per token — production-scale is still hours, not seconds.

The a-ha: I don't need to prove the model is correct. I need to prove the auditor is honest.

My system intercepts agent thinking blocks (Claude, OpenAI, Gemini), analyzes them against a behavioral contract, and produces a verdict: clear, review needed, or boundary violation. That derivation is deterministic — ~10,000 RISC-V cycles. Provable today.

So I built a guest program inside SP1's zkVM (on Modal) that re-derives the verdict from scratch, ignoring what the auditor claimed, and generates a STARK proof. If the auditor said "clear" but the evidence warranted "violation," the proof fails. The auditor cannot lie.

Quis custodiet ipsos custodes (Who watches the watchmen?) — answered with math. Sub-second on GPU.

OK... pretty cool, but what ELSE can you do with it?

Great question! Once I had provable individual verdicts: what about teams? Can I prove the group is safe?

I ended up applying financial risk theory to AI agent fleets (things I never expected to be doing with my life). CoVaR for tail risk — one bad agent in a group of four good ones doesn't average out to "fine." Markowitz portfolio theory for coherence — treating value alignment like diversification. DebtRank for contagion — if Agent A fails, who's exposed? Originally designed for bank failures. Works disturbingly well for agents.

Then I needed Shapley attribution for individual risk. Except real Shapley is exponential (2^n subsets), and Monte Carlo introduces randomness. Randomness = non-determinism = unprovable in a zkVM. Leave-One-Out approximation: deterministic, O(n²), the only Shapley variant that works inside a prover.

Oh, and all of it runs in Q16.16 fixed-point arithmetic (i32) because floating-point produces different results on different architectures, and "different results" inside a zkVM = worthless proof. I implemented exp, sqrt, and clamp from scratch in integer math. Casting spells at 2 AM in the dark again.

The whole stack — CoVaR, Markowitz, DebtRank, Shapley, circuit breakers — computes in TypeScript on Cloudflare Workers (instant), then re-derives in Rust inside the zkVM (provable). Both produce identical results. If they don't, something is very wrong.

So what?

Every agent accumulates cryptographically attested checkpoints — Ed25519 signatures, SHA-256 hash chains, Merkle trees, STARK proofs — and earns a Trust Score. Credit rating for AI agents, AAA to CCC. The score isn't an opinion. It's a computation over evidence anyone can independently verify. FICO computes scores from data you can't inspect. This computes scores from data anyone can cryptographically verify.

Everything I described here is live code. Four agents handling a production incident — coherence matrix, trust topology, Merkle visualization, drift detection: https://mnemom.ai/showcase

Apache-licensed. Zero-code gateway: npm install -g @mnemom/smoltbot && smoltbot register

GitHub: github.com/mnemom | Docs: docs.mnemom.ai

Show HN: OpenPawz – Open-source desktop AI agent platform (Rust/Tauri, 75 tools)

https://github.com/OpenPawz/openpawz
1•openpawz•1m ago•0 comments

'Birdbrain' benefits: How being an expert birdwatcher may boost cognition

https://www.nbcnews.com/health/health-news/birdwatching-birding-brain-boost-cognition-research-rc...
1•mikhael•2m ago•0 comments

Tell HN: GPT-5.3-codex is now available on the API

1•bigwheels•3m ago•0 comments

Show HN: I proved AI Model Collapse is a topological inevitability

https://www.amazon.co.uk/dp/B0GHT81F1K
2•Mhh1430•4m ago•0 comments

MatX inference chip 500M raise

https://seekingalpha.com/news/4555907-ai-chip-startup-matx-raises-500m-in-race-to-compete-with-nv...
1•Rogbeer•4m ago•0 comments

The AI-Augmented Scientist

https://www.theclimatebrink.com/p/the-ai-augmented-scientist
1•DamonHD•5m ago•0 comments

I built an engine to migrate Oracle PL/SQL to Java

https://blog.thalesouza.com.br/en/blog/logic-migration/
1•trsthales•5m ago•1 comments

Show HN: Intellegix HN Daily – AI podcast that reads HN's top stories aloud

https://open.spotify.com/show/7LZzKOKjYjYFrc4eI9AMpy
1•intellegix•6m ago•0 comments

Ezra and Jack Clark on Agents

https://www.nytimes.com/2026/02/24/opinion/ezra-klein-podcast-jack-clark.html
1•paulpauper•8m ago•0 comments

How the NBA can fix tanking

https://www.nytimes.com/athletic/7058361/2026/02/20/nba-tanking-how-to-fix-rule-changes/
1•paulpauper•8m ago•0 comments

State of Clojure Surveys

https://vlaaad.github.io/clojure-survey-trends
1•vlaaad•9m ago•0 comments

MCPs just got a front end, and it's a bigger deal than it sounds

https://wonderwhy-er.medium.com/mcp-apps-ai-just-got-a-frontend-and-we-built-one-cd1f6fc762f6
1•wonderwhyer•9m ago•1 comments

Crunchbase Data: AI Boom Has Changed Who's Funding Companies in 2025 vs. 2021

https://news.crunchbase.com/venture/data-2025-vs-2021-funding-hottest-companies-ai/
1•nadis•9m ago•0 comments

Show HN: Free AI-Powered Tools (writing, SEO, marketing, dev tools)

https://ai-tools-woad-six.vercel.app
1•oteo_builds•10m ago•0 comments

Robert Carradine Dies at 71

https://www.usatoday.com/story/entertainment/celebrities/2026/02/24/robert-carradine-death-lizzie...
1•toomuchtodo•10m ago•1 comments

AI can help startups define their ICP

https://mercury.com/blog/startups-ai-icp
1•nadis•10m ago•0 comments

The True Face of Prompt Injection

https://terallite.substack.com/p/the-true-face-of-prompt-injection
1•soletta•11m ago•0 comments

Show HN: A simple, free web app to track my portfolio across brokers

https://erincayaz.github.io/portfolio-tracker/
1•erinc951•11m ago•0 comments

Show HN: Brainstorm-MCP – Let GPT, DeepSeek, and Groq Brainstorm Together

https://github.com/spranab/brainstorm-mcp
1•spranab•13m ago•1 comments

Looks Like it is Happening

https://www.math.columbia.edu/~woit/wordpress/?p=15500
11•jjgreen•14m ago•0 comments

Ever wondered how Commodore 64 pixel art is still evolving today?

https://c64graphicsexplorer.com
1•angelday•14m ago•0 comments

Hosting the Olympics: the most expensive participation trophy

https://www.not-ship.com/hosting-the-olympics-the-worlds-most-expensive-participation-trophy/
1•speckx•15m ago•0 comments

Show HN: Disk Inventory X updated for Apple Silicon

https://diskinv.github.io/
2•lababidi•17m ago•2 comments

Anthropic's Responsible Scaling Policy: Version 3.0

https://www.anthropic.com/news/responsible-scaling-policy-v3
2•soheilpro•17m ago•0 comments

The End of Baseload Power as We Know It

https://finance.yahoo.com/news/end-baseload-power-know-200000114.html
1•epistasis•18m ago•0 comments

Apple's Touch-Screen MacBook Pro to Have Dynamic Island, New Interface

https://www.bloomberg.com/news/articles/2026-02-24/apple-s-touch-screen-macbook-pro-to-have-dynam...
2•jbredeche•18m ago•0 comments

Porting TeX from Pascal to TypeScript Using LLMs

https://hublog.hubmed.org/archives/002032
1•chl•18m ago•0 comments

Apple Accelerates US Manufacturing

https://www.apple.com/newsroom/2026/02/apple-accelerates-us-manufacturing-with-mac-mini-production/
12•haunter•19m ago•1 comments

Show HN: Bookie – Conquer the bookkeeping and accounting chaos of freelancing

https://www.bookie.tax/
1•nxnze•19m ago•0 comments

I built The Murderer's Lock – a security vault with a unique approach

https://murderers-lock-demo.vercel.app/
1•jennaleighwest•19m ago•0 comments