Show Levi Run Alphaevolve Claude Code/codex

Evidence

• Show HN: Levi – run AlphaEvolve on your Claude Code/Codex for dirt cheap — Hi HN,<p>Wanted to share something I'm excited about.<p>I’ve been fascinated by AlphaEvolve and its results for more than a year now, but using open source frameworks seems overwhelming because of the high costs. I can’t really afford hundreds of Claude Opus calls every time I want to run it. I want to be able to try it out many times and all sorts of unique domains. What if it was possible for AlphaEvolve to be much more affordable <i>while</i> getting a better performance? .<p>Over the last six months or so, I’ve been working on LEVI, an open source AlphaEvolve-like system that can outperform existing open source frameworks at a fraction of the cost (upto 35x cheaper!). It can also run on Claude Code or Codex, making it even more accessible (I've mostly been using it with a QWEN-30B). LEVI comes in two flavors where I felt it’ll make the most difference: Code Optimization, and Prompt Optimization (sorry math, you got a less direct path; workable through the code route).<p>The core thesis behind LEVI is that with the right search architecture, smaller models can substitute for or outperform larger ones. This means it’s much more economical to rely on smaller models for most of the work. That’s the entire takeaway. Making this work in practice is a different problem, but if you forget everything else from this post this is the only message I think I’m really trying to convey here.<p>LEVI does it in three ways: 1) Invest in solution diversity from the start and ensure its maintained. We don’t want to converge to the same solution, especially with smaller models in the mix, and rely on large models to pull us out of the basin. 2) Use smarter routing across larger and smaller models (i.e. most mutations don’t require a Claude Opus X) 3) For prompt optimization not every rollout is as important. Build a proxy subset to approximate.<p>I’ve tried LEVI on systems problems (like MoE scheduling or database transaction scheduling) and found that LEVI outperforms existing frameworks on almost every problem I threw at it while consistently using a smaller budget (unto 7x cheaper). For prompt optimization, across problems like IFBench and HotSpotQA, LEVI reaches a similar or better score as GEPA while using less than half the rollouts!<p>Happy to answer any questions or take any suggestions! If there are unexpected or niche domains where this can be applied, I would love to hear.<p>Technical Blog: <a href="https://ttanv.github.io/levi/" rel="nofollow">https://ttanv.github.io/levi/</a> GitHub: <a href="https://github.com/ttanv/levi" rel="nofollow">https://github.com/ttanv/levi</a>

  HACKER_NEWS

• Show HN: Modern RSS reader with social capabilities based on AT Proto — Hey HN!<p>The web is a bit broken and looking at social medias like X makes me a bit depressed. I feel like the original content that we had before on a multitude of website has been replaced by ton of click-bait content that just want impressions and reactions.<p>I would like us to go back to simpler times, where everyone had their publishing space (their website, or an hosted solution) and everyone could subscribe to it in their own software and do what they want with their feed.<p>I still think that social is important, like sharing, because afterall it's what made social networks interesting in the first place, but it should happen outside of where the content is.<p>That's why I wanted to build an RSS reader that is respectful of this principle of federation: no in-app reader, just RSS subscriptions. And I've added social capabilities based on the AT Protocol, like sharing / liking and following people.<p>The project was coded with AI, but it's my convictions that I've put it in. It's self-hostable and I hope that more and more people will be convinced to publish on their own website instead of X / Bluesky.

  HACKER_NEWS

• Show HN: Pluely – open-source alternative to Cluely (~10MB, Always On Display) — Hey HN,<p>A bit of backstory: I was job hunting for months, applying everywhere and racking up a ton of rejections. It got frustrating, so I decided to flip the script and build something useful instead: an open-source take on Cluely, that popular stealth AI tool from the $15M company. I figured why not recreate something that's already out there and buzzing, but make it free, transparent, and way better for privacy. Plus, it felt like a fun challenge.<p>This is actually my first project ever using Tauri, and I'm hooked. I started out trying Electron, but it just wasn't cutting it for something that needs to be fast and reliable - Electron apps end up bloated with all that browser overhead, sucking up RAM and starting slow, which kills the whole "instant access" vibe I was going for. Tauri, with its Rust core, keeps things lean and snappy without any of that extra weight. Anyway, Pluely is a super lightweight desktop app, just ~10MB (that's 27 times smaller than Cluely), designed to give you real-time AI help during video calls, interviews, sales talks, or pretty much any chat, all without anyone spotting it and without real-time meeting transcription(can add it based on the response).<p>Here's what makes it tick: An undetectable overlay: It's a see-through window that floats over any app like Zoom or Teams, and it stays hidden in screen shares, recordings, or screenshots. Support for multiple LLMs: Hooks up to OpenAI models like GPT-4, Anthropic's Claude, Google Gemini, or xAI Grok. It pulls the models dynamically, and everything runs off your own API keys stored right on your device - no servers or middlemen. Flexible inputs: Type stuff in, speak naturally (using Whisper for speech-to-text and voice detection so it just picks up when you talk), or toss in images and screenshots for the AI to break down.<p>Privacy and speed first: All processing is local, it uses half the CPU and RAM of similar tools, built on Tauri plus React. No tracking, no data leaving your machine. Conversation history: Kept locally so you can pick up where you left off. Works everywhere: Grab it as a .dmg for macOS, .msi for Windows, or .deb for Linux.<p>It's killer for stuff like nailing job interviews with discreet tips, debugging code on the spot in tech meetings, or getting strategy nudges during negotiations. Unlike the original Cluely, this one's totally free, you can tweak it however you want, and it's all out in the open.<p>Download now on GitHub releases: <a href="https://github.com/iamsrikanthnani/pluely/releases" rel="nofollow">https://github.com/iamsrikanthnani/pluely/releases</a><p>I'd love your thoughts, especially on adding custom LLM providers or tweaking the voice detection. What else would you want in a stealth AI sidekick?

  HACKER_NEWS