Show HN: We post-trained a model that pen tests instead of refusing https://ift.tt/MoCkSfW

Show HN: We post-trained a model that pen tests instead of refusing 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? https://ift.tt/JEgxUrb June 20, 2026 at 07:19PM

Show HN: Jumpjet – a WASM runtime for game developers https://ift.tt/NLsyDSh

Show HN: Jumpjet – a WASM runtime for game developers I built Jumpjet because I realized that engine and indie game developers are always repeating the same work: building the core infrastructure that touches the OS. Webassembly solves this in the Component Model by enabling interop between packages written in different languages. And in my opinion it's sort of the perfect fit for Jumpjet's model: providing a chassis without an engine. Jumpjet works by defining a very close mapping of WebGPU (and a few other WebIDL features) to WIT so that they can be used in any language that can target the wasm Component Model. Your game then runs as a guest application in Jumpjet's host runtime (powered by wasmtime), which shrinks final bundle size considerably versus something like Electron. Right now a bare bones game in Jumpjet is about 40mb. Right now the project is in an alpha or possibly pre-alpha state, it's not production ready. On the commercial side, I think there's an opportunity for cloud storage, game server hosting, a package manager and/or marketplace, distribution, and more. Right now you can target macOS, Windows, Linux, Android and iOS. (I haven't done any real testing on mobile so good luck.) The languages you can use will depend on their support for generating bindings from .wit files. There are a few templates available, I recommend one of the Rust ones. If you are a game developer or just like tinkering, I'd love for you to try the project out and tell me what you think! https://jumpjet.dev June 20, 2026 at 01:22AM

Show HN: Continuous Nvidia CUDA PC Sampling Profiler https://ift.tt/Jn9zL1T

Show HN: Continuous Nvidia CUDA PC Sampling Profiler Blog post about how we extended our open source profiler to include support for continuous production PC sampling. https://ift.tt/eFj7AQz June 15, 2026 at 09:19PM

Show HN: NanoEuler – GPT-2 scale model in pure C/CUDA from scratch https://ift.tt/lgkIsdy

Show HN: NanoEuler – GPT-2 scale model in pure C/CUDA from scratch https://ift.tt/7TiU0gX June 19, 2026 at 11:48PM

SHOW HN: I built a "living proof-of-work" profile for builders https://ift.tt/dxzhQ7T

SHOW HN: I built a "living proof-of-work" profile for builders https://kritive.com June 19, 2026 at 01:50AM

Show HN: I built a daily flag quiz in honor of the World Cup https://ift.tt/Ys14ql8

Show HN: I built a daily flag quiz in honor of the World Cup https://orbisearth.web.app/ June 19, 2026 at 01:45AM

Show HN: Run Agent Skills with mistral.rs v0.8.10: /v1/skills support and more https://ift.tt/yLNsz86

Show HN: Run Agent Skills with mistral.rs v0.8.10: /v1/skills support and more Hey all! I'm the maintainer of mistral.rs. I just landed support for OpenAI-compatible Agent Skills via a /v1/skills endpoint, and it works with local open models. Until now Skills have basically been locked to closed models, and with the ability to have private, local intelligence becoming increasingly important, but this feature allows you to do XYZ with local models. It's fully compatible with OpenAI's /v1/skills API, so you can drop mistral.rs into your existing code with minimal difficulty. We support the accompanying tools too: /v1/files or input_file for attaching files to your prompts, and mistral.rs also allows models to send generated files back using the OpenAI-compatible method. It's also easier than ever to try mistral.rs: we are including prebuilt binaries for NVIDIA CUDA, Apple Silicon, and CPU! # Linux/Mac > curl --proto '=https' --tlsv1.2 -sSf https://ift.tt/97azeRy... | sh # Windows > irm https://ift.tt/97azeRy... | iex Then: mistralrs serve --agent --isq 4 -m google/gemma-4-E4B-it Super excited for you to try this out and any feedback! Do you have any suggestions for what you would like to see in the next releases? Check out the GitHub: https://ift.tt/R1mGktC Docs & Quickstart: https://ericlbuehler.github.io/mistral.rs/ June 18, 2026 at 12:33PM