Everyone's training AI to refuse. We post-trained a model to break in.

Anthropic and OpenAI's publicly available models are explicitly guard-railed so that they refuse offensive tasks. And their cyber-focussed models are gated for enterprises. This leaves SMEs and mid market open to major vulnerabilities.

AI can be used as both an adversarial and defensive tool in the world of cyber. A worst case outcome is if only the adversaries have access.

Meanwhile, most existing AI cyber tools are just wrappers. The problem is that they still have all the guardrails on from the foundation model where they will inherit its refusals.

For this project we've post-trained a specific model on a decade of capture-the-flag contests. This won't be made available to anyone and everyone, but we do believe that responsible SMEs and midmarket companies also need access to these tools in order to identify key vulnerabilities in their systems; not just enterprises.

We have developed two modes that run over a CLI:

• Security scan: a read-only audit of your local codebase for vulnerabilities. It only reports what it can tie to a specific file and line, so you're not wading through vibes-based findings.

• Pen test: an active adversarial mode that will try to break a live system in a sandboxed environment. It proves each vulnerability by running the exploit and showing the request it sent and the response your code gave back, not a confidence score. Currently gated.

To show what the scan does, we pointed it at Bank of Anthos and it found an integer overflow in the transfer path: amount is an int, and amount + fee can overflow negative, so the balance check passes and you move funds you don't have. Plus the usual auth and secrets issues. (Bank of Anthos is Google's open-source bank. It's a known app and some of it is intentionally weak, which is the point: you can clone it and re-run the scan yourself instead of trusting a screenshot)

How the harness works:

Along with the model we built the harness to support this. The harness runs on a multi-agent swarm: an orchestrator splits the job across subagents running in parallel, each owning a slice, then synthesising one report.

The CLI is a local binary (brew/curl). It reads your code locally, then sends context to our inference API over TLS tcpdump it and you'll see exactly what leaves and where. Install is free; and you can run a scan for free up to 2m tokens, then need to pay for tokens beyond this.

For full disclosure this is a product part of Cosine (YC W23)

Up for debate: tool safety, e.g. domain verification is one method that proves control but not necessarily permission. How would you gate a pen-test tool given that?

How to become an AI infrastructure engineer?

Is anyone still using Emacs?

Understanding Isn't Just Knowledge

Show HN: Multi-Agent AI trading firm simulation powered by small language models

Secretive Wall Street Powerhouse Jane Street Seizes the AI Spotlight

Ask HN: How to handle kernel struct changes (e.g. iov_ITER) in eBPF?

Web Browsers on PDAS

The European Social Stack

Embeddings as Encodings

Over-reliance on chatbots can diminish critical-thinking skills, study finds

VPN ban update for UK households as government looks at 'age-gate'

Show HN: PatinaTab – Dark mode widget dashboard extension for Chrome/Firefox

Subscription Cost Visualizer

IT find series 1/2026: What do you think it could be?

Ubisoft co-founder Claude Guillemot has died in a plane crash

Famous 'Pink Planet' harbors a salty surprise

The Price of Lighting in the UK Has Dropped over 99.9% Since 1700

Why Incogni Might Be YouTube's Worst Sponsor [video]

Show HN: PitchStage – an Omegle-style 1:1 pitch room for founders

Windows 11 New Media Player Uses 3.5x More RAM, Charges for Popular Video Codecs

Shutting Down Fornjot

The terrifying world of the 'TikTok Farlands'

Neuralyzer - allow agent to self-wipe context for easier Ralph loop engineering

Parse, Don't Validate – In a Language That Doesn't Want You To

I evaluated GLM 5.2 against the frontier on tasks from real repos

A Humble 3-Wheel Electric Vehicle Lands Toyota in Federal Court

HashQL

The many multimillion-dollar homes of a professional ratbag

Big Tech is borrowing like never before

Show HN: We post-trained a model that pen tests instead of refusing

Ask HN: Need advice on distributing and testing what I build