Show HN: Kevin-32B – how to do multi-turn RL on writing CUDA kernels https://ift.tt/kXTPjpD

Show HN: Kevin-32B – how to do multi-turn RL on writing CUDA kernels Hey – we just published a blog post about Kevin-32B = K(ernel D)evin. It's to our knowledge the first open-source model that's RL-trained on CUDA kernels. Our goal was to demonstrate multi-turn RL using GRPO. We used 180 Python->CUDA conversion tasks from the KernelBench dataset. The results were surprisingly strong! We were able to outperform top reasoning model like o3 & o4-mini. We're sharing our training setup and learnings in the blogpost. Also the model is on HuggingFace: https://ift.tt/aqge680 https://ift.tt/8m2LWQ0 May 7, 2025 at 01:18AM

Show HN: X402 – an open standard for internet native payments https://ift.tt/IRjze1x

Show HN: X402 – an open standard for internet native payments Hi HN – excited to announce x402, initially developed by Coinbase (YC 12) x402 lets any HTTP API charge per request without issuing API keys or storing credit cards. Buyers (humans or AI agents) keep funds in their own wallet and dynamically discover compatible endpoints, call them as usual, and automatically pay a microtransaction in USDC or other tokens to settle. 90 second demo: https://www.youtube.com/watch?v=PV-L2AfLhJg Problem: Every time we want to use a new API we have to: find the service online create a developer account, copy a secret key into env vars, pre-fund or hand over a credit card This flow blocks agents even more. They can’t solve CAPTCHAs or enter credit cards. It also hurts sellers: fraud, chargebacks, onboarding friction, and marketing to humans are huge pain points. Why buyers care Zero setup – Hit a new endpoint immediately. Runtime discovery – Because every x402 service exists in a common registry, an agent can search, compare, and invoke in one loop. Self-assembling agents become practical. Easily create proxy servers – Want an endpoint that isn’t supported? You can use our proxy server template to spin up an x402-compatible instance yourself using traditional API keys, and monetize it for others wanting access. Why sellers care Reach incremental demand – Long-tail bots, side projects, one-off scripts, all of which too small for an account/signup flow, can now pay you. Micropayments without fraud – All payments settle onchain, nothing for stolen credit cards or chargebacks to reverse. Embedded distribution – instead of marketing to humans, create a compelling service meeting demand for agents and watch the requests roll in. How we got here Last year we launched AgentKit (wallets for AI agents). Tens of thousands of agents now hold onchain balances, but they can’t pay for most web services. We revived the long-unused HTTP 402 (“Payment Required”) status code and wrote a spec to make it real. Marc Andresseen calls the lack of native value transfer “the original sin of the internet,” and we see x402 as the absolution of this sin. How it works: x402 specifies a standard response body to accompany a 402 status code. This response body contains machine understandable instructions for how to pay. Payments are signature based an included as an `X-PAYMENT` header in a subsequent request to the same API endpoint. The accepting server can verify and settle payment themselves, or delegate the onchain settlement to what we call a facilitator. This means you don't have to touch crypto as a developer, you can just integrate a middleware and start receiving stablecoin payments in as little as 1 line of code. Because x402 natively traverses your existing client / server requests, it can be implemented in any language, and doesn't require webhooks, or any other complex integration. Its literally this simple: `paymentMiddleware("0xYourAddress", {"/your-endpoint": "$0.01"})` Ask HN API providers – does the one-line integration fit your stack? What’s missing? Agent / infra builders – if a service isn’t available is the proxy server template sufficient? File issues, PRs welcome Everyone – poke holes in the trust and fee model; we’d love to iterate with your feedback Curious to learn more? Check out our documentation and repo for more information, and please don’t hesitate to reach out to get onboard. https://ift.tt/wSJf9vB https://x402.org https://ift.tt/DhpJZYe... https://www.x402.org/ May 6, 2025 at 11:54PM

Show HN: Feedsmith — Fast parser & generator for RSS, Atom, OPML feed namespaces https://ift.tt/fvhw4sb

Show HN: Feedsmith — Fast parser & generator for RSS, Atom, OPML feed namespaces Hi HN! While working on a project that involves frequently parsing a lot of feeds, I needed a fast JavaScript-based parser to extract specific fields from feed namespaces. Existing Node packages were either too slow or merged all feed formats, losing namespace information. So I decided to write it myself and created this NPM package with a simple API. Feedsmith supports all feed formats and many popular namespaces, including: Podcast, Media, iTunes, Dublin Core, and more. It can also parse and generate OPML files. I am currently adding support for more namespaces and feed generation for RSS, Atom and RDF. The library grew into something bigger than I initially anticipated, so I also started creating a dedicated documentation website to describe all the features. https://ift.tt/J0zYVO1 May 6, 2025 at 11:33PM

Show HN: Real-time AI Voice Chat at ~500ms Latency https://ift.tt/DF2Mph6

Show HN: Real-time AI Voice Chat at ~500ms Latency https://ift.tt/qP2EmRA May 6, 2025 at 01:47AM

Show HN: Tkintergalactic - Declarative Tcl/Tk UI Library for Python https://ift.tt/Su1edos

Show HN: Tkintergalactic - Declarative Tcl/Tk UI Library for Python https://ift.tt/2tLJsGa May 5, 2025 at 11:32PM

Celebrate Small Business Week with Muni: Explore the Bayview, Chinatown, Castro and Excelsior

Celebrate Small Business Week with Muni: Explore the Bayview, Chinatown, Castro and Excelsior
By Glennis Markison

Artist credit: Dan Bransfield. This week in San Francisco, we’re celebrating and lifting up our small business community, the heart of San Francisco! We’re thrilled to help you visit your favorites – and discover plenty of new ones – this Small Business Week (May 5 to 9). You can always count on us when you Go Local. We’ll showcase a dozen neighborhoods throughout the week and cover how you can get to them on Muni. Plus, we'll share a little slice of what we do behind the scenes to make your trips to small businesses smooth and reliable. Today, we’ll help you explore the Bayview, Chinatown...

Published May 05, 2025 at 05:30AM
https://ift.tt/s3RVG6h

Show HN: I built a mini macOS app to reveal my yearly subscription spending https://ift.tt/bIjyUL5

Show HN: I built a mini macOS app to reveal my yearly subscription spending I built a macOS app to track my subscriptions after realizing I was spending over $XXXX annually on services I barely used. I wanted a simple, privacy-focused tool to help me stay on top of recurring charges without relying on third-party integrations or sharing financial data. Key Features: – Visual Calendar: See all upcoming charges in a monthly view. – Custom Notifications: Set reminders for upcoming payments. – Highlights: Flag subscriptions as annual, trial, or one-time. – Statistics: View projected yearly spending, average monthly costs, and peak spending months. – Custom Categories: Organize subscriptions with user-defined categories. – Multi-Currency Support: Convert prices on the fly to your preferred currency. – Status Management: Mark subscriptions as active or canceled, with accurate updates in your stats. – Quick Addition: Start typing a service name, and the app auto-suggests logos, categories, and colors. – Import and export data The app is free to use with some limitations. I’m currently working on additional features and would appreciate any feedback or suggestions from the community. https://ift.tt/by8cC4K May 5, 2025 at 11:31PM