Run Nix Based Environments in Kubernetes
Hacker News (score: 32)Description
More from Hacker
Production-Grade Container Deployment with Podman Quadlets – Larvitz Blog
Production-Grade Container Deployment with Podman Quadlets – Larvitz Blog
Show HN: Built an MCP server using Cloudflare's Code Mode pattern
Show HN: Built an MCP server using Cloudflare's Code Mode pattern Read this article by Cloudflare this morning <a href="https://blog.cloudflare.com/code-mode/" rel="nofollow">https://blog.cloudflare.com/code-mode/</a> the main argument being that LLMs are much better at writing typescript code than tool calls because they've seen typescript code many more times.<p>HN Discussion: <a href="https://news.ycombinator.com/item?id=45399204">https://news.ycombinator.com/item?id=45399204</a> <a href="https://news.ycombinator.com/item?id=45386248">https://news.ycombinator.com/item?id=45386248</a><p>Deno provides a great sandbox environment for Typescript code execution because of its permissions system which made it easy to spin up code that only has access to fetch and network calls.<p>Stick an MCP proxy on top of that and you've got "CodeMode" (code intermixed with MCP tool calls) for more advanced workflow orchestration.<p><a href="https://github.com/jx-codes/codemode-mcp" rel="nofollow">https://github.com/jx-codes/codemode-mcp</a><p>There's a lot of things that can be improved here. Like a virtual file system for the agent to actually build up its solution instead of being forced to one shot the solution but the bones are there.
Optimizing ClickHouse for Intel's 280 core processors
Optimizing ClickHouse for Intel's 280 core processors
Launch HN: RunRL (YC X25) – Reinforcement learning as a service
Launch HN: RunRL (YC X25) – Reinforcement learning as a service Hey HN, we’re Andrew and Derik at RunRL (<a href="https://runrl.com/">https://runrl.com/</a>). We've built a platform to improve models and agents with reinforcement learning. If you can define a metric, we'll make your model or agent better, without you having to think about managing GPU clusters.<p>Here's a demo video: <a href="https://youtu.be/EtiBjs4jfCg" rel="nofollow">https://youtu.be/EtiBjs4jfCg</a><p>I (Andrew) was doing a PhD in reinforcement learning on language models, and everyone kept...not using RL because it was too hard to get running. At some point I realized that someone's got to sit down and actually write a good platform for running RL experiments.<p>Once this happened, people started using it for antiviral design, formal verification, browser agents, and a bunch of other cool applications, so we decided to make a startup out of it.<p>How it works:<p>- Choose an open-weight base model (weights are necessary for RL updates; Qwen3-4B-Instruct-2507 is a good starting point)<p>- Upload a set of initial prompts ("Generate an antiviral targeting Sars-CoV-2 protease", "Prove this theorem", "What's the average summer high in Windhoek?")<p>- Define a reward function, using Python, an LLM-as-a-judge, or both<p>- For complex settings, you can define an entire multi-turn environment<p>- Watch the reward go up!<p>For most well-defined problems, a small open model + RunRL outperforms frontier models. (For instance, we've seen Qwen-3B do better than Claude 4.1 Opus on antiviral design.) This is because LLM intelligence is notoriously "spiky"; often models are decent-but-not-great at common-sense knowledge, are randomly good at a few domains, but make mistakes on lots of other tasks. RunRL creates spikes precisely on the tasks where you need them.<p>Pricing: $80/node-hour. Most models up to 14B parameters fit on one node (0.6-1.2 TB of VRAM). We do full fine-tuning, at the cost of parameter-efficiency (with RL, people seem to care a lot about the last few percent gains in e.g. agent reliability).<p>Next up: continuous learning; tool use. Tool use is currently in private beta, which you can join here: <a href="https://forms.gle/D2mSmeQDVCDraPQg8" rel="nofollow">https://forms.gle/D2mSmeQDVCDraPQg8</a><p>We'd love to hear any thoughts, questions, or positive or negative reinforcement!
No other tools from this source yet.