2025年12月10日 の Show HN

Built this over the last few days, based on a Rust codebase that parses the latest ALPR reports from OpenStreetMaps, calculates navigation statistics from every tagged residential building to nearby amenities, and tests each route for intersection with those ALPR cameras (Flock being the most widespread).

These have gotten more controversial in recent months, due to their indiscriminate large scale data collection, with 404 Media publishing many original pieces (https://www.404media.co/tag/flock/) about their adoption and (ab)use across the country. I wanted to use open source datasets to track the rapid expansion, especially per-county, as this data can be crucial for 'deflock' movements to petition counties and city governments to ban and remove them.

In some counties, the tracking becomes so widespread that most people can't go anywhere without being photographed. This includes possibly sensitive areas, like places of worship and medical facilities.

The argument for their legality rests upon the notion that these cameras are equivalent to 'mere observation', but the enormous scope and data sharing agreements in place to share and access millions of records without warrants blurs the lines of the fourth amendment.

Mobile keyboards today are almost entirely based on the 26-key, 3-row QWERTY layout. Here’s a new 2-row, 16-key alternative designed specifically for smartphones.

Hi HN!

I'm building a JavaScript debugger called Wirebrowser. It combines network inspection, request rewriting, heap snapshots, and live object search.

The main experimental feature is BDHS (Breakpoint-Driven Heap Search): it hooks into the JavaScript debugger and automatically captures a heap snapshot at every pause and performs a targeted search for the value or structure of interest. This reveals the moment a value appears in memory and the user-land function responsible for creating it.

Another interesting feature is the Live Object Search: it inspects runtime objects (not just snapshots), supports regex and object similarity, and lets you patch objects directly at runtime.

Whitepaper: https://fcavallarin.github.io/wirebrowser/BDHS-Origin-Trace

Feedback very welcome, especially on whether BDHS would help your debugging workflow.

https://github.com/mcp-shark/mcp-shark Site: https://mcpshark.sh/

I built MCPShark, a traffic inspector for the Model Context Protocol (MCP).

It sits between your editor/LLM client and MCP servers so you can: • See all MCP traffic (requests, responses, tools, resources) in one place • Debug sessions when tools don’t behave as expected • Optionally run “Smart Scan” checks to flag risky tools / configs

Browser-based fighting game where you can play as Elon, Zuck, Jensen, Sam, or Sundar.

Built it over a couple weekends using Claude Code and Cursor, deployed on Vercel.

For years I’ve wanted to start a podcast to interview curious and passionate makers in the depths of their creative pursuits.

I would love any feedback, a rating, and if you know anyone would would make a great guest, please let me know!

We're releasing early efforts on coding agent governance with Cupcake [1] - an open-source policy enforcement layer with native integrations. You write rules in policy-as-code (OPA/Rego), and Cupcake integrates them into the agent runtime via Hooks.

See it in action (Desktop only): https://cupcake-policy-studio.vercel.app/example-policies/se...

Help us build: https://github.com/eqtylab/cupcake

We are EQTY Lab, our mission is verifiable AI (identity, provenance, and governance). With the rise of capable agents like Claude Code, it became immediately clear that those deploying these agents need the ability to conduct their own alignment and safety controls. We can’t rely solely on the frontier labs.

This is why we created the feature request for Hooks in Claude Code [2], and pivoted away from filesystem and OS-level monitoring once those hooks were implemented. Hooks provide the critical points we need:

* Evaluation: Checking agent intent and actions.

* Prevention: Stopping unsafe or unwanted actions.

* Modification: Adjusting the agent's output before execution.

Policy-as-Code with OPA/Rego - While many agent security papers suggest similar policy architectures using invented DSLs, Cupcake is fundamentally built on Open Policy Agent (OPA) and its policy language, Rego [3].

We chose Rego because it is:

* Industry-Robust: Widely adopted across enterprise DevSecOps and cloud-native environments.

* Purpose-Built: Offers unique, mature advantages for defining, managing, and enforcing policy as code.

* Enterprise-Oriented: This makes Cupcake compatible with existing enterprise governance frameworks.

Cupcake is released under the Apache-2.0 license. We will formalize a path to v1.0.0 in Q1 of 2026. This is an early preview version. The goal with Cupcake is not suppression, but to ensure an agent is able to drive fast without crashing. To collaborate, or join forces: ramos at eqtylab dot io.

[1] https://github.com/eqtylab/cupcake

[2] https://github.com/anthropics/claude-code/issues/712

[3] https://www.openpolicyagent.org/

DskDitto is a small CLI utility that rapidly scans your disk for duplicates wasting disk space. It’s written in Golang and provides a sleek TUI to interactively manage file dups. It’s very fast and capable but the project is still somewhat in its infancy. If any one wishes to become a contributor please reach out. Feedback welcome. Thanks :-)

Qyavix is a tiny DOM runtime with a React-like state hook (`u`) and a simple render function (`r`). The entire framework is only 155 bytes (minified, no gzip).

I built it for benchmarks, experiments, and micro-UI systems. The official website (https://qyavix.pages.dev) is fully rendered using Qyavix itself.

Features: 155 bytes total size Zero dependencies Works with plain DOM nodes Hooks-like state system Fast re-rendering (full refresh) Pure JavaScript, no build tools needed

Would love feedback from the community!

I love the Wordle game, but one puzzle a day feels way too slow if you're actually trying to get better. So I built wordles.org - a simple unlimited Wordle practice site.

It also has one extra thing I always wished the original had: you can create your own 4–10 letter word and share a link so friends can guess it. I'm not sure if people actually want a practice tool like this or if I'm just the kind of person who plays 10+ rounds a day.

Curious what others think: Is unlimited Wordle useful or just noise? Are custom puzzles fun or annoying?

If something feels off, tell me straight. I'd rather get blunt feedback than polite silence.

Hey HN,

I'm a solo developer and kept running into the same problem: I'd want Claude CLI to work on multiple tasks at once, but locally I could only run one instance at a time. So I built The Box.

It lets you spin up isolated sandbox environments in the cloud and run Claude CLI agents in parallel. You write the prompt, pick a GitHub repo, and it handles the rest – creates a branch, does the work, and opens a PR.

The core workflow: - Connect your GitHub repos - Create tasks with prompts - Claude CLI runs in isolated sandboxes - Get notified when done (browser push or Telegram) - Review and merge PRs

Tech stack: Go (Fiber) for backend services, Next.js for frontend, PostgreSQL, Redis for signaling, RabbitMQ for task queue. Workers run Claude CLI in Docker containers.

Currently free to use (with limits) while I finish payment integration. Would love feedback from anyone who's tried to scale their AI coding workflow.

https://the-box.dev

Hey!

It always annoyed me to manually try to figure out the best winning ads from competitors, so I created my own ads library that automatically calculates it

I got tired of needing to copy my AI coding files and configs across multiple projects...

So I created OpenPackage, the open source CLI tool that packs your specs, rules, commands, subagents, skills, plugins, and more, into reusable, composable, and shareable packages.

It's basically a hybrid of npm, GitHub, and Claude Code Plugins for AI coding files.

As an example, here's one of my favorite commands to use for code cleanup: https://openpackage.dev/@awesome/f/.openpackage/commands/cle... And my NextJS project scaffolding package: https://openpackage.dev/@hyericlee/nextjs

Would love to see the packages you all come up with to power your AI coding workflows!

P.S. Will be launching a TUI for OpenPackage soon, so drop a star on GitHub and stay tuned.

Hey HN! We built an open source context layer to power AI agents and apps even in air-gapped infra setups.

TL;DR: Use our API (or one of the 6 SDKs) to push context into the Skald platform, and get semantic search and AI chat out of the box.

We’ve seen companies spend months building a context layer system internally, only for it to have subpar performance and require active maintenance.

Skald gives you the plumbing to get started really fast when building context-aware agents and AI apps (customers have gone to prod in a day with us) but is still highly extensible and configurable to fit individual needs.

Our core is entirely MIT-licensed and you can even run it with a fully self-hosted stack (document parsing, chunking, vector search, LLM, etc) in your VPC. We think solutions in this space must be open source, and not just “open source but bring 6 API keys to get started”.

Our customers have used Skald to quickly build features for their own users, as well as for internal tooling.

The self-hosted MIT version is completely fully-featured, so much so that it’s what we run on our Cloud offering today. We do also have an enterprise license for larger customers that runs on their infra and has a few enterprise-focused features on top of the MIT version.

Keen to hear your feedback and I’ll be around to answer questions!

GitHub: https://github.com/skaldlabs/skald

Hi HN,

I built steem.dev, a new AI vibe-coding tool that generates fully functional WordPress plugins from natural language prompts. Instead of searching for the right plugin, hiring a developer, or fighting with overbuilt tools like Elementor add-ons, Yoast extensions, WooCommerce tweaks, image compressors, custom dashboards, or random code snippets copied from forums, you just tell steem what you want and it builds the plugin for you.

The idea came from my own workflow. Every time I needed a small tweak or a one-off feature, I found myself opening Gemini or Claude, prompting for code, pasting it into a plugin skeleton, fixing the hooks, debugging naming collisions, adding sanitization, reorganizing files, and trying again. The LLMs were powerful, but the process around them was messy and repetitive. I realized I was still doing all the boring engineering work by hand.

So I built steem.dev. You describe the feature. Steem generates the plugin with proper WordPress structure, correct hooks, security escapes, activation logic, admin pages if needed, AJAX endpoints, REST endpoints, blocks, widgets, or whatever the feature requires. It validates the code, packages it, and you download a ready-to-install zip. No manual setup, no boilerplate, no debugging hallucinated APIs.

Examples of things people have already built with it: a custom post type dashboard widget with analytics, a WooCommerce checkout fee rule, a lightweight SEO field manager, a role-based content lock plugin, a cron-driven RSS importer, a keyword highlighter, and even a full image offloading plugin. All from prompts.

Steem is not a marketplace of prebuilt plugins. It is closer to a personal developer that instantly builds exactly what you want, every time, without any technical knowledge. It is opinionated about correct WordPress architecture, avoids insecure patterns, and prevents the usual mistakes that come from copy-pasting random code.

My goal is to make WordPress development feel like talking to an engineer instead of writing PHP. The tool is free to try, and I would love feedback from people who work with WP, LLM coding tools, or plugin ecosystems. Does this model make sense? What kinds of plugins should I showcase as examples? How deep should the generated architecture go? And if you have thoughts on safety, sandboxing, or plugin validation, I am all ears.

Happy to answer any questions.

I built a one-page tool to speed up my stock research workflow.

You type a ticker, and it gives you quick links to 20+ research tools, like FinViz, TradingView, SEC EDGAR, Yahoo Finance, Seeking Alpha, Reddit etc. And each deep link is pre-loaded with that ticker.

Nothing fancy. Just got tired of typing the same ticker into 6 different sites every time I wanted to dig into a company.

Built with Next.js, Prisma, and Tailwind. Would love feedback on what tools to add or how to make it more useful.

Example:

https://finstack.pro/t/GOOGL

Hey HN,

I built this because the "matching to meeting" flow in modern dating is broken. People chat for weeks, exchange numbers, meet up, and realize there is zero chemistry in 10 seconds.

Meet Zero is a burner video link tool.

The Tech: - Video: Built on Stream's Video SDK (WebRTC) for low latency. - Auth: Supabase (Anonymous guest logins via UUIDs in sessionStorage). - Privacy: No phone numbers stored. Rooms are ephemeral and have a hard 10-minute expiry (enforced via edge functions). - AI: I integrated OpenAI (GPT-4o) as an on-demand "Wingman" (it generates topic cards based on a vibe, rather than listening to the audio, to preserve privacy).

It’s free for guests (no install needed). Would love feedback on the WebRTC stability and the "Vibe" prompt latency.

Live demo here: https://meet-zero.co

I built this because I was tired of guessing chunk_size=1000 and overlap=200 in my RAG pipelines and hoping for the best.

RAG-TUI is an open-source terminal tool that lets you visualize exactly how your text is being split before you index it. It helps you tune parameters in real-time and spot issues like sentences getting cut in half.

The Stack: Built with Textual (Python) for the TUI. Chonkie for token-based chunking. Usearch for local vector search. Integrates with Ollama for entirely local/private debugging (also supports OpenAI/Groq).

Key Features: Visual Feedback: See color-coded chunks and their overlap regions instantly. Quality Indicators: The UI flags "bad chunks" (e.g., for cuts mid-sentence, for clean breaks). Batch Testing: Run a set of queries against your current settings to calculate a "hit rate" before you deploy.

Export: Generates the Python code (LangChain/LlamaIndex) for your tuned config.

It’s currently in beta (v0.0.1). I’d love feedback on the chunking visualizations and what other metrics would be useful for debugging retrieval.

Repo: https://github.com/rasinmuhammed?tab=repositories PyPI: https://pypi.org/project/rag-tui/

Hi HN,

I'm a solo dev. I built this because I found manual keyword matching for ATS systems to be a huge time sink.

It's a Next.js app that takes a PDF resume + Job Description text, analyzes the semantic gap, and rewrites the bullet points to match the target keywords.

Tech stack: Next.js, Vercel AI SDK, Tailwind.

There is a free interactive demo (no signup/login required) to test the parsing speed.

Link: https://cvora.net

Feedback on the parsing accuracy is welcome.

Distilled is my first solo, end-to-end project. Designed and built by me, to solve a problem I had myself: I always wanted a calmer way to read.

Distilled shows one HN article at a time. Each article has a few clear insights, just enough to spark curiosity and invite deeper reading. And is clearly categorised.

Built this app to replace mindless scrolling of tiktok, slowing down and reading without any noise.

I’d love your thoughts about this idea.

Used Google Gemini 2.0 Flash for a recommendation system, extracting articles, generating insights, and classifying. System is not perfect, but I am working on it.

Will be really happy to receive any feedback, recommendations or suggestions.

A year ago I wouldn’t have believed this project is possible for a designer to build solo.

I saw users trying to create gameplay style videos with voice over, captions and hence created a simple free tool. Trying to add more features. Please let me know if you want any specific feature.