1. Show HN: Serendipity – rediscover the web one interesting page at a time (serendipity.surf)
Worthwhile webpages, one at a time.
2. Is MCP Good Yet? (ismcpgoodyet.com)
Spoiler: no. MCP features across Codex, Cursor, Claude Code, Grok, and OpenCode.
3. Show HN: Semantic Overlays – an NX bit for LLM prompt injection (live demo) (semantic-overlays.vercel.app)
Tiny adapters added to a frozen language model. They only fire at marked token positions, and they change how the model perceives those spans.
6. Show HN: MQLens – A Native MongoDB GUI (SSH, X.509, Kerberos, No Electron) (mqlens.com)
MQLens is a free, native, cross-platform MongoDB GUI with the power of paid tools: every auth mode (SCRAM, X.509, AWS, Kerberos, LDAP), TLS/SSH/proxy, aggregation pipelines with explain plans, bulk edit, index/view management, schema analysis, GridFS, an embedded mongosh, and an AI query assistant. Credentials are encrypted locally with zero telemetry. Apache-2.0.
7. Show HN: A Bluesky bot that shares Wikipedia links from the HN front page (bsky.app)
Bot that shares Wikipedia articles posted to HackerNews (created by @jackm.in)
8. Show HN: A Curated Library of Hero Sections from Fast-Growing Startups (www.herotxt.page)
Real hero-section headlines and sub-headlines from 118 live product websites, with screenshots and version history. See how competitors reposition over time.
10. Show HN: Saccade – Live semantic browser truth for AI agents (github.com)
Closed-loop browser control runtime for AI agents. Contribute to nanlogic/saccade development by creating an account on GitHub.
11. Show HN: Nos4, new nostalgic internet toy. A complete IOS4 right on the browser (nos4.fun)
Everything you remember. Right in your browser.
12. Show HN: I measured when impossible-to-book restaurants release their tables (tablefriend.app)
One famous restaurant drops 52 to 65 tables at 5:00pm Pacific, every single day. Measured, not folklore.
13. Show HN: OSS, K8s-native AI platform for distributed multi-model inference (github.com)
Distributed, multi-model LLM inference on Kubernetes you own - axem-solutions/shaide
14. Nemotron 3 Ultra Explained (miraflow.ai)
NVIDIA's Nemotron 3 Ultra is a 550B open-weight hybrid Mamba-Attention MoE model built for long agentic runs. Here is how its architecture, training, and throughput numbers actually work.
NVIDIA’s Nemotron 3 Ultra is a highly advanced AI model designed for complex tasks, featuring a unique architecture and training process. The article breaks down its impressive specifications and performance metrics, making it a must-read for those interested in cutting-edge AI technology.
15. Show HN: Keeplea – photograph everything you want to remember (keeplea.app)
Your camera is for moments. Keeplea is for everything else — the paint, the plate, the wifi password. Photograph it once and ask for it later in your own words.
16. We Let AI Agents Rewrite a 92M-Message-a-Day Service in Go (www.checklyhq.com)
How Checkly rewrote a Node.js service handling 92M messages a day into Go using AI agents, and the test harness that made it safe to ship.
Checkly transformed a high-traffic Node.js service that processes 92 million messages daily into Go with the help of AI agents. The article details the innovative test harness they developed to ensure the transition was safe and effective.
17. Show HN: Unitpost/email – open-source Email Components and Layouts (www.unitpost.com)
Email component library that works in Outlook, Gmail, and Apple Mail. MIT. Install with npm or a Skill. No account.
18. Show HN: Seatlr – We rebuilt our traveler chat into an AI-powered travel network (www.seatlr.com)
Experience seamless communication with cutting-edge offline-first technology, real-time sync, and multilingual support that breaks all barriers.
20. Show HN: PearPie – Private AI chat that syncs peer-to-peer, no accounts needed (www.pearpie.ai)
PearPie is out on Mac, Windows, Linux and mobile. No account, no setup, nothing to learn: your conversations stay on your own devices, and premium models run in Europe.
21. Show HN: Openheim – a multi-provider LLM agent runtime, written in Rust (openheim.io)
Open-source Rust LLM agent. TUI, ACP stdio agent for editors, MCP integration, and self-hosted WebSocket server. OpenAI, Anthropic, Gemini, Ollama — no lock-in.
22. Show HN: I Built the AI command center that runs all local models in one place (generativeide.com)
Your AI Coding Command Center. Run any model, local or cloud. Plan with frontier AI, loop locally. Privacy-first, model-agnostic, air-gap ready.
24. Show HN: I rebuilt tank game with new style (www.battle-tank.com)
Play Battle Tank free in your browser — a LEGO-style arcade shooter inspired by Battle City. Destroy enemies, collect power-ups, and defend your base across 10 stages. No download needed!
25. Show HN: Claude Code Arcade (github.com)
Vibe code with an 80s arcade vibe — games in your Claude Code status line - jystervinou/claude-code-arcade
26. Show HN: Turn repeated coding-agent corrections into rules/skills (blume.codes)
A narrow always-on desktop companion that watches every harness, surfaces what each agent is doing, and asks you to tidy up the rulebook when chat contradicts it.
27. Optimal transport is art: images to collection of optimally placed particles (blog.wolfram.com)
Is it possible to generate a stippling drawing from any image programmatically? An investigation into stippling properties, spacing, writing a program.
Learn how to create stippling drawings from images through a program that analyzes spacing and placement of particles. This investigation reveals the techniques behind transforming visual art into a structured computational process.
29. Show HN: Deslop, a skill that turns legalese and AI slop into plain English (github.com)
A professional plain English style guide and agent skill that translates legalese and purges AI slop. - fayerman-source/deslop
30. Show HN: Chaosword – A Crossword? (chaosword.com)
Play three fresh crosswords every day, free: a 5x5 Morning warmup, a 7x7 Evening grid, and an 11x11 Midnight capstone. Fair grids, sharp and occasionally feral clues, the same puzzles for everyone.