The best Hacker News stories from Show from the past day
Latest posts:
Show HN: Qwen-2.5-32B is now the best open source OCR model
Last week was big for open source LLMs. We got:<p>- Qwen 2.5 VL (72b and 32b)<p>- Gemma-3 (27b)<p>- DeepSeek-v3-0324<p>And a couple weeks ago we got the new mistral-ocr model. We updated our OCR benchmark to include the new models.<p>We evaluated 1,000 documents for JSON extraction accuracy. Major takeaways:<p>- Qwen 2.5 VL (72b and 32b) are by far the most impressive. Both landed right around 75% accuracy (equivalent to GPT-4o’s performance). Qwen 72b was only 0.4% above 32b. Within the margin of error.<p>- Both Qwen models passed mistral-ocr (72.2%), which is specifically trained for OCR.<p>- Gemma-3 (27B) only scored 42.9%. Particularly surprising given that it's architecture is based on Gemini 2.0 which still tops the accuracy chart.<p>The data set and benchmark runner is fully open source. You can check out the code and reproduction steps here:<p>- <a href="https://getomni.ai/blog/benchmarking-open-source-models-for-ocr">https://getomni.ai/blog/benchmarking-open-source-models-for-...</a><p>- <a href="https://github.com/getomni-ai/benchmark" rel="nofollow">https://github.com/getomni-ai/benchmark</a><p>- <a href="https://huggingface.co/datasets/getomni-ai/ocr-benchmark" rel="nofollow">https://huggingface.co/datasets/getomni-ai/ocr-benchmark</a>
Show HN: Duolingo-style exercises but with real-world content like the news
I've been working on a little side project that combines Duolingo-like listening comprehension exercises with real content .<p>Every video is transcribed to get much better transcripts than the closed captions. I filter on high quality transcripts, and afterwards a LLM selects only plausible segments for the exercises. This seems to work well for quality control and seems to be reliable enough for these short exercises.<p>Would love your thoughts!
Show HN: Duolingo-style exercises but with real-world content like the news
I've been working on a little side project that combines Duolingo-like listening comprehension exercises with real content .<p>Every video is transcribed to get much better transcripts than the closed captions. I filter on high quality transcripts, and afterwards a LLM selects only plausible segments for the exercises. This seems to work well for quality control and seems to be reliable enough for these short exercises.<p>Would love your thoughts!
Show HN: Nue – Apps lighter than a React button
Show HN: Nue – Apps lighter than a React button
Show HN: Nue – Apps lighter than a React button
Show HN: I made a C program to create a vanity SHA-1 hash for a text file
Show HN: JavaScript PubSub in 163 Bytes
Show HN: GuMCP – Open-source MCP servers, hosted for free
Hello! We open sourced all our current MCP servers to platforms like Slack, Google sheets, Linear, Perplexity and will be contributing a few more integrations every day to the project.<p>problems we're hoping to solve:<p>- Many people are creating MCP servers for the same apps. They're scattered across different repos but flavors of the same thing. We're making one standardized mono project for all MCP servers.<p>- Startups are charging for hosting MCP servers. This is blocking tons of people from being able to play around with MCP casually. We're hosting them for free.<p>- Non-technical people should be able to use MCP without needing to learn how to clone a repo and set up a venv. We're trying to enable a one click integration if people want to use the free hosted service.<p>The plan is to keep contributing until we have an MCP server for basically every useful app anyone could want.
Show HN: NoteUX – Fast and minimalist note-taking app
Show HN: NoteUX – Fast and minimalist note-taking app
Show HN: I made a little puzzle game about a rogue chess knight
Show HN: I made a little puzzle game about a rogue chess knight
Show HN: I made a little puzzle game about a rogue chess knight
Show HN: WhatsApp MCP Server
Hi HN – I built an open-source, self-hosted Model Context Protocol (MCP) server for WhatsApp: <a href="https://github.com/lharries/whatsapp-mcp" rel="nofollow">https://github.com/lharries/whatsapp-mcp</a><p>It connects to your personal WhatsApp account via the WhatsApp Web multi-device API (using whatsmeow from the Beeper team), and doesn't rely on third-party APIs. All messages are stored locally in SQLite. Nothing is sent to the cloud unless you explicitly allow your LLM to access the data via tools – so you maintain full control and privacy.<p>The MCP server can:<p>- Search your messages, contacts, and groups<p>- Send WhatsApp messages to individuals or groups<p>Why build this?<p>99% of your life is stored in WhatsApp, by connecting an LLM to WhatsApp you get all this context. And your AI agent can execute tasks on your behalf by sending messages.
Show HN: WhatsApp MCP Server
Hi HN – I built an open-source, self-hosted Model Context Protocol (MCP) server for WhatsApp: <a href="https://github.com/lharries/whatsapp-mcp" rel="nofollow">https://github.com/lharries/whatsapp-mcp</a><p>It connects to your personal WhatsApp account via the WhatsApp Web multi-device API (using whatsmeow from the Beeper team), and doesn't rely on third-party APIs. All messages are stored locally in SQLite. Nothing is sent to the cloud unless you explicitly allow your LLM to access the data via tools – so you maintain full control and privacy.<p>The MCP server can:<p>- Search your messages, contacts, and groups<p>- Send WhatsApp messages to individuals or groups<p>Why build this?<p>99% of your life is stored in WhatsApp, by connecting an LLM to WhatsApp you get all this context. And your AI agent can execute tasks on your behalf by sending messages.
Show HN: Multi UPS SNMP based shutdown
Show HN: OmniTools – Self-Hosted Open-Source Swiss Army Knife for Everyday Tasks
Hey HN!<p>I built OmniTools, a self-hosted web app that offers a collection of useful tools for everyday tasks. Think PDF manipulation,image, text utilities, and more—all in one place.<p>Open-source & self-hosted<p>Works fully in the browser (privacy-first)<p>Built with React, deployed via Docker<p>I made this because I wanted a free, self-hosted alternative to online tools that often require subscriptions. Would love feedback & ideas on what to add next!<p>GitHub link: <a href="https://github.com/iib0011/omni-tools" rel="nofollow">https://github.com/iib0011/omni-tools</a>
Demo link: <a href="https://omnitools.app" rel="nofollow">https://omnitools.app</a><p>Let me know what tools you want next.
Show HN: Bknd – Firebase alternative that embeds into any React stack
I built bknd because I was tired of stitching together auth, file storage, and database APIs every time I started a new project. Existing solutions were either too hard to extend, too basic, or required a separate deployment.<p>bknd runs directly inside your frontend app — no separate backend required. It works with Next.js, Remix, Astro, React Router. It also runs standalone on Cloudflare Workers, AWS Lambda, Bun, or Node.<p>It supports Postgres, LibSQL (Turso), D1, SQLite and has adapter-based storage. You get instant APIs, multi-strategy auth, media handling and a built-in Admin UI.<p>Curious what you’d build with it, feedback welcome!
Show HN: I built a tool to add noise texture to your images
I'm excited to introduce Noise Tools – a simple yet powerful tool that lets you effortlessly add noise textures to your images. Whether you're a designer, artist, or just experimenting with aesthetics, Noise Tools helps you enhance your visuals with just a few clicks.<p>Why I built this?
I often found myself needing high-quality noise textures for design projects but struggled to find a quick and easy solution. So, I built Noise Tools to make the process easy for everyone!<p>Features:
Generate noise textures instantly
Adjust intensity & styles
No downloads or complicated settings<p>Would love to hear your thoughts! Try it out and let me know what you think.<p>Check it out here: noisetools.vercel.app