2026년 4월 10일의 Show HN

Hello HN users,

This is a CAD by code project I have been working on on my free time for more than year now. I built it with 3 goals in mind: - It should be familiar to CAD designers who have used other programs. Same workflow, same terminology. - Reduce the mental effort required to create models as much as possible. This is achieved by: - Provide live rendering and visual guidance as you type. - Allow the user to reference existing edges/faces on the scene instead of having to calculate everything. - Provide interactive mouse helpers for features that are hard to write by code: Only 3 interactive modes for now: Edge trimming, Sketch region extrude, Bezier curve drawing. - Implicit coding whenever possible: e.g: There are sensible defaults for most parameters. The program will automatically fuse intersecting objects together so you do not have to worry about what object needs to be fused with what. - It should be reasonably fast: The scene objects are cached and only the updated objects are re-computed.

I think I have achieved these goals to a good extent. The program is still in early stages and there are many features I want to add, rewrite but I think it is already usable for simple models.

Hi all,

Finding a good data structure for a word processor is a difficult problem. My notebook diaries on the problem go back 25 years when I was frustrated with using Word for my diploma thesis - it was slow and unstable at that time. I ended up getting pretty hooked on the problem.

Right now I’m taking a professional break and decided to finally use the time to push these ideas further, and build MiniWord — a WYSIWYG word processor in Python.

My goal is to have a native, non-HTML-based editor that stays simple, fast, and is hackable. So far I am focusing on getting the fundamentals right. What is working yet is:

- Real WYSIWYG editing (no HTML layer, no embedded browser) with styles, images and tables.

- Clean, simple file format (human-readable, diff-friendly, git-friendly, AI-friendly)

- Markdown support

- Support for Python-plugins

Things that I found:

- B-tree structures are perfect for holding rich text data

- A simple text-based file format is incredibly useful — you can diff documents, version them, and even process them with AI tools quite naturally

What I’d love feedback on:

- Where do you see real use cases for something like this?

- What would be missing for you to take it seriously as a tool or platform?

- What kinds of plugins or extensions would actually be worth building?

Happy about any thoughts — positive or critical. Greetings

Eve is an AI agent harness that runs in an isolated Linux sandbox (2 vCPUs, 4GB RAM, 10GB disk) with a real filesystem, headless Chromium, code execution, and connectors to 1000+ services.

You give it a task and it works in the background until it's done.

I built this because I wanted OpenClaw without the self-hosting, pointed at actual day-to-day work. I’m thinking less personal assistant and more helpful colleague.

Here’s a short demo video: https://www.loom.com/share/00d11bdbe804478e8817710f5f53ac61

The main interface is a web app where you can watch work happen in real time (agents spawning, files being written, use of the CLI). There's also an iMessage integration so you can fire a task asynchronously, put your phone down, and get a reply when it's finished.

Under the hood, there's an orchestrator (Claude Opus 4.6) that routes to the right domain-specific model for each subtask: browsing, coding, research, and media generation.

For complex tasks it spins up parallel sub-agents that coordinate through the shared filesystem. They have persistent memory across sessions so context compounds over time.

I’ve packaged it with a bunch of pre-installed skills so it can execute in a variety of job roles (sales, marketing, finance) at runtime.

Here are a few things Eve has helped me with in the last couple days:

- Edit this demo video with a voice over of Garry: https://www.youtube.com/watch?v=S4oD7H3cAQ0

- Do my tax returns

- To build HN as if it was the year 2030: https://api.eve.new/api/sites/hackernews-2030/#/

AMA on the architecture and lmk your thoughts :)

P.S. I've given every new user $100 worth of credits to try it.

Keeper is an embeddable secret store (Argon2id, XChaCha20-Poly1305 by default). Four security levels, audit chains, crash-safe rotation. Vault is overkill for most use cases. This is for when you ge paranoid about env and need encrypted local storage that doesn't suck. No security through obscurity, hence, It's still early, so now's the best time to find weird edge cases, race conditions, memory leaks, crypto misuse, anything that breaks. The README has a full security model breakdown if you want to get adversarial.

Hi there, Zero Stars here.

I recently published some new work to Hokusai Pocket, which is a cross-platform binary made on top of raylib and MRuby that runs GUIs from ruby scripts.

License?

MIT!

How does it work?

The binary is available on the GitHub releases page: https://github.com/skinnyjames/hokusai-pocket/releases/tag/0...

You can download the binary on x86 Windows, OSX, or Linux, and run your GUI application with

hokusai-pocket run:target="<your_hokusai_app.rb>"

For a little bit of a hello world, I started a photoshop clone

https://github.com/skinnyjames/hokusai_demo_paint

Also a little game

https://github.com/skinnyjames/pocket-squares

Docs / Help?

The docs are in progress, but the old docs for the CRuby version express some of the basic ideas around the project. https://hokusai.skinnyjames.net/docs/intro

(I'm also available to answer questions in between slinging pizza)

Deps?

Hokusai pocket currently uses

* libuv for offloading cpu intensive tasks to a worker pool to prevent blocking the UI thread, and I plan to integrate some libuv networking as well.

* raylib for backend graphics / I've also built with SDL on arm64 to run applications on my pinephone

* NativeFileDialog for the lovely integration into filesystem.

* MRuby for running or embedding the scripts

* tree-sitter for the custom template grammar (Although templates can be built with ruby)

Anyway, I hope you get a chance to try it. If you make something cool AND have docker installed, you can also publish your work as single binary

`hokusai-pocket publish:target=<your cool program.rb>`

Would love feedback, apps, and help with documentation and more build targets.

urs truly,

@ ᴗ @

Feels a bit like Figma, but for coding agents.

Instead of going back and forth with prompts, you give the agent a DESIGN.md that defines the design system up front, and it generally sticks to it when generating UI.

Google Stitch seems to be moving in this direction as a standard, so we put together a small collection of DESIGN.md files based on popular web sites.

I’ve been cautiously (and nervously) playing with openclaw and a number of other claw and code agents for a while now, but trying out different ones was tricky so I wanted a simple way to switch out channel ownership… then I wanted more. Security is hard, and I wanted to make it easier. This is FAR from polished, and no claims that I’m a “security expert” but I tried to think and research a bit on different threat models (I think of 2 broad ones for agents, external adversaries and internal agentic failures) and try and offer best in class protection on both, while also not having any special opinion on what a good agent may look like today or in the future… this is just a gateway, and hopefully one that can work for nearly any agent now or in the future, but trying to come with batteries included for some of the more popular options today like openclaw, zeroclaw, claw-code, clause and opencode, not all there yet but contribution and critiques welcome.

This is an interactive version of the mass-vs-radius chart from Lineweaver & Patel's "All Objects and Some Questions" (2023). It plots the scale of everything in the observable universe on a single log-log chart

This was an experiment in extending the utility of a science infographic via hyperlinks, navigability and tooltips that detail the meaning of specific parts.

There are likely still spatial/numerical issues, happy to take corrections.

The original paper: https://pubs.aip.org/aapt/ajp/article/91/10/819/2911822/All-...

SmolVM is an open-source local sandbox for AI agents on macOS and Linux.

I started building it because agent workflows need more than isolated code execution. They need a reusable environment: write files in one step, come back later, snapshot state, pause/resume, and increasingly interact with browsers or full desktop environments.

Right now SmolVM is a Python SDK and CLI focused on local developer experience.

Current features include: - local sandbox environments - macOS and Linux support - snapshotting - pause/resume - persistent environments across turns

Install: ``` curl -sSL https://celesto.ai/install.sh | bash smolvm ```

I’d love feedback from people building coding agents or computer-use agents. Interested in what feels missing, what feels clunky, and what you’d expect from a sandbox like this.

Hey HN — I'm Aaron, and I built Steadcast because I was frustrated that I listen to 15+ hours of podcasts a week and retain almost none of it.

Steadcast is a native macOS podcast player (Swift, SwiftUI, AVFoundation — no Electron) designed around one idea: podcasts are how a lot of us learn now, but no podcast app actually treats them that way.

What makes it different:

• Learning domains, not genres. The app is organized around 10 knowledge domains (AI, science, history, philosophy, psychology, etc.) with 90+ editorially curated shows. No true crime, no celebrity gossip.

* On-device transcription. Every episode gets transcribed locally using the macOS 26 Speech framework. Every word becomes searchable. No cloud processing, nothing leaves your Mac.

•AI summaries. Apple Intelligence generates summaries that capture the key ideas from each episode.

• Memories. Press Ctrl+Cmd+K while listening and Steadcast saves a timestamped bookmark with the full transcript context around that moment. So when you hear something worth remembering, you can actually find it later.

• Smart Speed. Intelligent silence trimming with a live counter showing exactly how much time you've saved.

• Keyboard-first. Global hotkeys, full keyboard navigation — built for people who work at their Macs all day.

It's free on the Mac App Store: https://apps.apple.com/app/steadcast/id6760033638

Requires macOS 26. Built as a solo project. I'd love feedback on the curation approach especially — is a podcast app that has opinions about what you should listen to a feature or a bug?

Hi HN,

Two things inspired me to build this: Andrej Karpathy recently talked about going back to RSS feeds — "higher quality longform and a lot less slop intended to provoke" (https://x.com/karpathy/status/2018043254986703167). Around the same time, Marc Andreessen shared his information source: 25% X, 25% podcasts with practitioners, 25% conversations with AI, 25% old books (https://x.com/pmarca/status/2030949039345467570).

This made me realize how much it matters to control what you let in. But I've used to the feed-style reading habit (each social media does like this). I just don't want a random meme or rage bait to hijack my attention out of nowhere.

So I built FeedSense — a private recommendation system. You pick your sources, and it builds a "For You" feed from only those sources. The system takes content quality, publish time, and your read history into account (still tuning it), but it will never surface anything outside the sources you chose

TestFlight beta on iOS: https://testflight.apple.com/join/56VPEeeS

https://feedsense.cc

Happy to talk architecture or the recommendation approach if anyone's curious.

RMBG-1.4 + SAM running client-side via ONNX Runtime WASM. ~2s on laptop, works on mobile. Your image never leaves the browser.

Built this as part of allplix.com. 19yo student in France, solo project.

Happy to talk about the WASM pipeline or the pain of running ML models in a browser tab.

macOS's built-in cycle window shortcut (⌘` / ⌘@) has always been flaky for me. Probably not a Show HN, but if it annoyed me this much it might be annoying some others.

Only tested on the latest macOS — would appreciate any reports from other versions.

There are many website for searching cheap flights, but I don't know any for searching cheap train fares. I decided to build one myself. Three providers are currently covered: European Sleeper, NightJet, RegioJet.

The best part about coding agents is I get to indulge in making games again. They are so good at basic html / js games that I'm surprised we don't see another flash-game-like renaissance

Reword was originally built as a way for me to learn to beat my friends in the game Anagrams (itself a variant of Bananagrams). The goal is simple -- given a root word and a few additional letters, try and find a workable anagram.

Some elements are inspired heavily from wordle, including the once-per-day randomization and the share emoji.

Would love feedback!

I built a desktop app to manage my job search because I got tired of juggling Numbers, Notion, and Pages to track applications and different versions of my resume.

Kin is a private, offline-first tool for managing the job search process in one place:

- track job applications and conversations - maintain a master resume and cover letter - generate tailored resumes and cover letters per application - store notes, contacts, and interview history - (optional) Chrome/Brave extension for saving job listings - calendar integration for interviews

Everything runs locally. No accounts, no cloud sync, and your data never leaves your machine.

It’s free and open-source: https://github.com/ruslanora/kin

Still early, but feedback is welcome, especially from people who’ve tried managing job searches with Notion, spreadsheets, or similar tools.

Airwave is a small self-hosted project that uses yt-dlp + ffmpeg to create a single live MP3 stream from supported sources.

All listeners connect to the same stream endpoint, which avoids drift entirely.

Supports YouTube, SoundCloud, Mixcloud, and Sonos.

Curious what people think about the approach.

I’ve been working on the A2A Utils [1] package, a comprehensive set of utility functions for using A2A servers (remote agents). It has been three weeks in the making, and is based on nearly a year’s experience using A2A servers in production. The A2A MCP Server [2] and OpenClaw A2A Plugin [3] are wrappers around it, and it will power other packages in the near future. If you are planning on using A2A servers, I highly encourage you to read through the documentation for 30 minutes. I promise it will save you at least 10x that time!

A quick introduction to A2A. A2A standardises agent discovery, communication, and authentication. The protocol introduces the concept of an Agent Card, a JSON object containing the agent’s name, description, skills, accepted inputs, outputs, and authentication methods. It also defines a Message, a Task (a unit of work), an Artifact (the output of a Task), and a Part (a fundamental object for Messages and Artifacts that can represent text, data, and files).

Most people associate A2A with multi-agent systems, but it also serves as a framework for deploying agents. You get a standardised API endpoint for sending text, data, and files from any A2A compatible SDK, UI, or app. Agent Cards can be used to catalog and search agents. And A2A was explicitly designed for long-running, asynchronous tasks with support for polling, streaming, and webhooks. To learn more, see What is A2A? [4]

Check the GitHub discussion out, where I compare sending a Message and polling a Task with the A2A SDK and A2A Utils. I think it's much more elegant! I would love to hear your thoughts and welcome feedback.

Links:

[1] A2A Utils: https://github.com/a2anet/a2a-utils [2] A2A MCP Server: https://github.com/a2anet/a2a-mcp [3] OpenClaw A2A Plugin: https://github.com/a2anet/openclaw-a2a-plugin [4] What is A2A?: https://a2a-protocol.org/latest/topics/what-is-a2a/#core-ben...

I started building this for my dad to help track stock for his small manufacturing business, including reorder levels and answering "can I make x of this?". It's grown to include bill of materials, stock takes, purchase orders, and production orders, and aims to compete with spreadsheets rather than existing tools.

It's a local desktop app (Electron and SQLite), so there are no ongoing subscriptions or data sovereignty concerns, though you do pay the Electron RAM tax.

There's a 30-day trial with no feature restrictions at https://kitted.site.

I'd be delighted to answer any questions and take any feedback.