2026年5月22日 的 Show HN

Pablo is a Chrome extension that copies the HTML and CSS behind any element you hover.

It captures computed styles, fonts (with @font-face and Google Fonts links), CSS keyframes, and animation props from GSAP and Framer Motion. The output is structured so it pastes cleanly into Claude Code, Cursor, or Codex when you want to rebuild a component in your own stack.

Manifest v3, no host permissions, no backend. Free.

https://usepablo.dev

Happy to answer questions about how the extraction works, and would love to hear about sites where the output breaks.

Spec Driven Development approach allows to squeeze more from coding agents thanks to few strong concepts: - decomposition across two dimensions. first you generate specs in multiple steps (requirements, code analysis, design), than you split task into multiple subtasks and implement them one by one - you clear context between every step - after spec generation and after subtask implementation. this helps keep cost low and context clear and focused which boost performance - specs written to disk help with information persistency - delivering specs layer by layer help to catch early when agent got you wrong

Repo with claude plugin for spec driven development: https://github.com/sermakarevich/sddw

The Darwin kernel (XNU) has been open source since 1999. But "open source" and "you can actually build and boot it" are very different things. XNU depends on proprietary tooling, kexts, and a bunch of open source projects published without a build system.

Darnix builds the whole thing with Nix. The kernel, the filesystem, the boot image, all the way to a running QEMU instance. We patched XNU to run on QEMU without kexts, ported HFS+ from a kernel extension into the kernel itself, fixed GRUB's Mach-O loader, and wired it all together as a flake. The build is fully sandboxed. No root, no volume mounting, no device access, no network.

The bigger idea is a revival of PureDarwin (https://www.puredarwin.org/), a standalone OS on Apple's open source Darwin layer, with Nix managing everything above the kernel. Sort of like a Nix OS on XNU instead of Linux.

Right now the kernel boots, mounts a ramdisk, and runs a single static binary. Next step is a shell.

https://github.com/jonhermansen/darnix

Technical details and the full list of patches are in the README. I would love to hear from anyone who’s thought about this space!

---

Darnix is not affiliated with or endorsed by Apple Inc. This is not macOS. Apple, macOS, and related trademarks are the property of Apple Inc. Darwin is licensed under the APSL.

Hello! This project is a TLS/X.509 certificate manager and PKI. Designed to provision and install your TLS certificates across public and private endpoints. It's something I created for my home lab, but has quickly expanded as I have been experimenting more and more with AI development tooling.

You can see the product overview: https://www.zaita.com It's architecture and usage is nicely documented: https://docs.zaita.com

Unfortunately, you do need to sign up to have a play, but there is a demo instance that doesn't require a credit card. Just enter an email (I won't send you any emails) and password.

Keen for feedback. I've found the major players in the industry are only interested in Enterprise arrangements. I wanted something I could use in my home lab across my many docker containers etc, but still had cool features like network discovery and certificate transparency log scanning.

Hello HN! I'm Claire, founder of nao Labs (YC X25). Two months ago I started building a company brain in a git repo. Now I fully run my company with 8 AI agents, 20+ skills, and a self-learning context repo.

I built it in a git repo because I refused to lock into any tool - not a context layer, not a specific agent harness. Today I'm open sourcing it so others can build their own.

Sylph is the open source version of the company brain I use. It gives you the structure to host your own company context, build skills, create AI agents, and already has a self-learning loop scaffolded. It is a Git repo, with no lock-in on any tool, that can work with any agent: Claude Code, Codex, Cursor.

Sylph is made for you to make your own: fork it, run /sylph-setup, and it will build your own company brain according to your own context.

Repo: github.com/getnao/sylph

How I built it: https://thenewaiorder.substack.com/p/i-built-a-company-brain...

Tell me what you think, and if you've got some tips if you built an AI brain for your company too!

Made this video on our project which might be of interest to folks here! Happy to answer any questions. First time making a long-form video.

Previous HN thread: https://news.ycombinator.com/item?id=41235789

Video is also on PeerTube: https://spectra.video/w/4wTobcS16Ww5UAZVMWrC9E

Store memories, auto-extract entities and relationships, search semantically. MCP server + REST API + SDKs. Self-hostable, cloud option, MIT license.

It's a clone of 7 Little Words, but with topics from computer science and programming. No sign-up, no app install, no tracking. It's a PWA and works offline, also as a home screen app.

Tech: js, no libs, Canvas API, Web Audio, AI-aided but not vibe coded, puzzles curated by hand.

Hey everyone,

I'm back with some exciting updates. I previously introduced an open source private home security camera in 2024, which uses OpenMLS for end-to-end encryption. It was called Privastead then and it's now renamed to Secluso.

John Kaczman found my project from here and has been working on it with me over the last year and half. We've made a lot of improvements to the software, which we would like to share with you:

- You can now set this up on your Raspberry Pi in less than 5 minutes with no technical expertise using our easy-to-use GUI deploy tool. We've put together a comprehensive build-your-own guide that walks you through the required steps (you can find a link at the top of the repository README).

- We use a customized, minimal OS based on the Yocto project for the camera.

- Every part of our stack except for the iOS app has reproducible builds. This includes our Android app, camera/server binaries, deploy tool, and the aforementioned OS.

- We've re-designed our mobile app, which is now on the iOS App Store and Google Play store.

- We now support UnifiedPush for more privacy-preserving push notifications.

Looking forward to seeing what you all think!

Hi HN, I’m Mike, the founder of OpenRig.

I built this because my Claude Code + Codex setup kept forming little "topologies" of long-lived agents that worked well together, but the terminal sprawl was intense. So I built a primitive the agents could intuitively reach for to save and recreate these setups on the fly. This then led to more agent-first primitives like coordination, declarative workflow patterns, workspaces, etc.

Several months in and these "rigs" I manage with openrig require a lot less babysitting and I can manage more projects at once without getting overwhelmed.

The short version: OpenRig is a way to save and operate that topology instead of rebuilding it by hand every time.

Demo video: https://youtu.be/yCFSRnPDFqY?si=n5e627d0CU3X3bmE

GitHub: https://github.com/mvschwarz/openrig

CoreMem lets you build collections of context, called a mem, and share it with any AI agent via URL, a Chrome extension, MCP, Cursor/VS Code plugins, a skill, and more. Instead of re-explaining your project or goal when you switch agents or start new sessions, CoreMem keeps your context centrally organized so that any AI tool can read it.

This originally started as a CLI I built that kept pieces of context (Project A/B/C details, my writing style, preferred tech stacks, coding style, etc) in a SQLite database. I could instruct various agents to “use my `coremem` CLI to retrieve details about [project A] before we get started.” It solved a problem for me b/c I am continually bouncing around between different projects and chat agents, and having to re-explain myself every time became an exercise in either repeating myself or copy/pasting summaries I’d saved from previous sessions.

I decided to make this a little more robust and portable, so I turned that original CLI into a SaaS. Tl;dr: You can create a “mem”, which is a collection of 1 or more pieces of related context, and share that mem with any agent to quickly get them up to speed.

Right now I’ve got integrations in the form of revokable share links, a Chrome Plugin, Cursor Plugin, Cursor/VS Code extension, Claude Code plugin, ChatGPT/Claude/Gemini/et al via MCP. Since I mostly work from the CLI, I use the Claude Code plugin or create 5-min share links I can drop into a chat, but I’ve tried to make this useful to people who mainly work from a browser or an IDE.

I’ve been coding for 30+ years, and I vibed most of this. I was able to use CoreMem to help it built itself as I jumped between various coding agents, having them grab context then start a new task. I’m sure my architecture and engineering experience helped, but building this in a few weeks confirmed for me that the barrier for someone to build a tool they need to solve a problem is incredibly low.

The rush I used to get from coding has mostly faded, but I’m getting similar rushes managing different agents to build things now.

I built a browser app for recording calls and turning them into transcripts/summaries without inviting a meeting bot.

The current version is BYOK: you provide an OpenAI API key, stored locally in your browser. Recordings/projects stay in browser storage; selected audio/transcript text is sent to OpenAI only when you run transcription/summary.

I am testing whether the wedge is real: no bot in the meeting, no SaaS workspace for recordings, summaries/exports afterward.

I would appreciate feedback on the trust boundary and whether BYOK is acceptable for a first version.

I used it once to record a meeting I didn't want to spend 1 hour on, and just needed the summary - the tool captured the meeting audio even after I turned off the laptop audio.

One unexpected benefit of LLMs is I can work on projects I otherwise wouldn't have taken on. I made a web-based autoshooter (with multiplayer support) heavily using AI / LLMs.

This is something I'd consider "alpha" quality so don't expect a super polished experience but it's hopefully fun

Hey HN, I'm sharing a tool I made to create yardage books for golf courses.

I'd originally written a Python tool a few years ago that pulled data from OpenStreetMap, auto-calculated relevant distances, and then created imagery for a yardage book. If you want to see the original, it's here: https://github.com/npilk/hacker-yardage

But Python was a pretty big technical barrier for most casual golfers. Then recently I realized most of the same functionality could be bundled into a JS app. Claude did most of the port for me.

The result is OpenYardage. You can search for a golf course and the tool will auto-generate a yardage book for you.

I'd welcome any feedback or ideas! The Python version still has some features I haven't been able to port yet, the main one being topographical data to visualize slopes.

The source is available here: https://github.com/npilk/openyardage-web

Hey folks, I've been working on this atomic orbital for the last few months. It's specifically designed to help you learn about quantum mechanics. Would love some feedback! <3 <3

HEVCut is a photo and video compression app for Apple platforms I have been working on. Today's update adds AVIF as an output format for photos, which I believe makes it the first app on the App Store to ship AVIF encoding on iOS.

Apple added AVIF decode in 2022 (iOS 16, macOS Ventura) but never shipped an encoder, public or private. Photos cannot export AVIF, ImageIO cannot write it, and no other third party app on the store does it either. So if you wanted AVIF files coming off an iPhone, your options were "send the original somewhere else and re-encode it" or "wait."

On typical iPhone photos, AVIF lands around half the size of HEIC at matched visual quality. A 5 MB HEIC compresses to roughly 700 KB. HDR (including gain map HDR content from newer iPhones) is preserved, SDR works correctly across readers, everything runs on device.

The rest of the app does the usual library cleanup things: HEIC and HEVC recompression, duplicate detection, surfacing space hogs, swipe to delete, a private vault. AVIF is the new piece.

I've spent the last year building Charm, a native macOS menu bar app that corrects spelling, fixes grammar, and predicts your next word.

Three features:

- Spells: NSSpellChecker plus a local LLM for context-aware corrections (catches "definately" -> "definitely" without the false positives macOS autocorrect is known for). - Polish: sentence-level grammar fixes triggered at punctuation boundaries. - Oracle: next-word prediction, Tab to accept.

100% local. Choose Gemma 2 2B or Qwen 2.5 3B; the model downloads once from HuggingFace on first use, then everything runs on your Mac. No cloud option exists, no accounts, no API keys, no telemetry. The model auto-unloads after 5 minutes idle so the RAM cost is bounded.

Built in Swift / SwiftUI. Needs Accessibility and Input Monitoring permissions because text replacement requires reading keystrokes globally. Nothing leaves your machine.

One-time purchase.

I'd love feedback, especially edge cases where corrections go wrong.

I built Quit All, an iPhone app for breaking bad habits.

  The main idea is that most habit trackers help after relapse, but cravings happen before that. Quit All has an SOS mode with a timer, GIFs/prompts, streak tracking, relapse logging,
  savings, milestones, danger-time stats, and iOS widgets.

  YouTube demo:
  https://www.youtube.com/watch?v=qwNK4rqOY88

  App Store:
  https://apps.apple.com/us/app/quit-all-break-every-habit/id6760978934

  Website:
  https://quit-all.com

Hey HN, if you do use Safari, try my little ReTab extension. It makes jumping back and forth between tabs way smoother and honestly improves my productivity & enjoyment a ton.

Tap Cmd+E to jump back & forth between your last visited tabs Hold Cmd+E to show a list of your tab history to select (just like Cmd+Tab on macOS)

Tab history is tracked per-window so cycling stays scoped to the window with focus. No analytics, no remote calls, nothing leaves the Mac — a single Safari Web Extension on MV3 spec + small Swift container app.

22s demo: https://youtu.be/oysvE06Ys-0

Any feedback or ideas are welcome!

Hi everyone,

I wanted to share an open source Telegram-based personal AI assistant I built. It’s a model-agnostic agent with memory, skills, tools (like web search, browser user, etc.) operating in a persistent workspace. It also has support for scheduled tasks, and can build powerful HTML-based apps that live in the workspace.

Here are some of my favorite use cases:

* Send Lilo photos of food, and it tracks your calories.

* Leave a voice note on your run to pause your supplements, and Lilo adds a TODO.

* Have Lilo remind you when the Knicks game starts and even send you score updates every 5 minutes.

* Have Lilo read an article out loud. Or give you a summary of the top stories on Hacker News.

* Forward a Uber receipt, and pull it up later to file for a reimbursement at work.

* Schedule a meeting with Jess next week, ask for suggestions on location, and next week, remind you to leave for the meeting on time.

While Telegram is my most frequently used channel, Lilo can also be accessed by email, WhatsApp, a website and a mobile app. Email is particularly useful: I often forward receipts, invites, etc for Lilo to handle.

How is this different from OpenClaw and Hermes Agent? Here are some reasons:

- Runs on a remote machine/in the cloud rather than your local machine - your local data is safe, and the assistant is available 24/7.

- More visual/ more GUI - Lilo comes with a default set of apps like a TODO list that you can interact with not just by text, but also with a GUI on the mobile and web app.

- The Telegram integration is very comprehensive (handles replies, voice notes, reactions, etc.).

I use Lilo a ton to manage my life. Would love to hear your feedback!

Github: https://github.com/abi/lilo

We’ve been building an open-source FPGA-based 400G RDMA-like SmartNIC called OpenRDMA.

The project includes RTL, drivers, and the FPGA data path implementation. Current hardware throughput has reached around 200Gbps so far.

One interesting aspect has been verification: we’ve been using LLM-generated cocotb testbenches quite extensively. AI is still weak at SystemVerilog, but surprisingly useful for Python-based cocotb verification workflows.

The entire stack is open-source: github.com/open-rdma/open-rdma

Would love feedback from people working on FPGA networking, RDMA, SmartNICs, or high-performance infrastructure.

Glimpse is a native macOS markdown viewer. The thing I built it around: the AI features run entirely on-device through Apple's FoundationModels (the same model behind Apple Intelligence). Summaries, key insights, an auto-glossary of jargon in whatever you're reading. Nothing leaves your machine, no API key, no account.

I read long markdown all day, mostly engineering docs and RFCs, and the existing options frustrated me. Electron apps felt slow. The fast native ones had no AI. The ones with AI wanted me to send a private spec to someone's cloud to get a summary. I wanted to open a 40-page doc and ask "what's actually new here" without it leaving my Mac.

Other things it does: double-click any rendered block to edit just that part of the source, file watching that doesn't fight you while you edit, presentation mode, KaTeX and Mermaid offline, and a QuickLook extension so .md files preview the same way in Finder.

Worth being upfront. AI needs macOS 26 because FoundationModels doesn't exist before it. The viewer and editor still work on older macOS, AI is just gated. Mac App Store only right now. 14 day free trial, then monthly or yearly subscription. I went with subscription because I want runway to keep shipping, but I'm honestly not sure it's the right model for a tool like this, so push back if it bugs you.

I built AIMirror because I got tired of analytics tools showing me what users did, but never why they struggled.

AIMirror is a lightweight UI/UX analysis toolkit for web apps that detects frustration, hesitation, dead clicks, rage clicks, and confusing user flows in real time — then explains what’s likely going wrong so you can actually fix the experience for the users instead of guessing.

I am a creative technologist with a design degree majoring in digital design, with a heavy programming background I found combining the two exctremely complementary yet somewhat confusing for a lot of people who only either do the UX/UI side, or the back end, systems side. UIMirror helps people who want to help with that middle ground guess work that stops an app from being useful, or accessible.

a cool extension to make your AI chats better