2026年4月23日 的 Show HN

Hey there! I am Luca, I write https://refactoring.fm/ and I built Tolaria for myself to manage my own knowledge base (10K notes, 300+ articles written in over 6 years of newslettering) and work well with AI.

Tolaria is offline-first, file-based, has first-class support for git, and has strong opinions about how you should organize notes (types, relationships, etc).

Let me know your thoughts!

Hi HN, we built SuperHQ, an open source app that runs AI coding agents in isolated microVM sandboxes instead of directly on your machine. Each agent gets its own VM with a full Debian environment. You mount your projects in, writes go to a tmpfs overlay so your host is never touched, and you get a diff view to accept or discard changes. API keys never enter the sandbox. We also just launched remote.superhq.ai which acts as a remote control for SuperHQ, allowing you to access your workspaces and agents from anywhere.

I built localLLLM: a small community project for running local models.

Live: https://locallllm.fly.dev

The goal is simple: if someone has model + OS + GPU + RAM, they should get steps that actually work (ideally one liner)

I need help populating and validating guides.

If you run local models, please submit one working recipe (or report what failed). Would love to hear general feedback as well!

Hi HN — I built Cartoon Studio, an open-source desktop app for making simple 2D cartoon scenes and shows.

The basic flow is: place SVG characters on a scene, write dialogue, pick voices, and render to MP4. It handles word timestamps, mouth cues, and lip-sync automatically.

This started as me playing around with Jellypod's Speech SDK and HeyGen's HyperFrames. I wanted a small tool that could go from script to video without a big animation pipeline and next thing I knew I was trying to create my own South Park style show and here we are. :D

A few details:

- desktop app built with Electron

- supports multiple TTS providers through Jellypod's Speech SDK

- renders via HyperFrames

- lets you upload or generate characters and backdrop scenes

- includes default characters/scenes so you can try it quickly

- open source

It runs from source today. AI features use bring-your-own API keys, but the app itself is fully inspectable and local-first in the sense that there’s no hosted backend or telemetry.

Here are some fun examples of the the types of videos you can create:

https://x.com/deepwhitman/status/2046425875789631701

https://x.com/deepwhitman/status/2047040471579697512

And the repo:

https://github.com/Jellypod-Inc/cartoon-studio

Happy to answer questions and appreciate any feedback!

Hi HN! Few months ago I've shared a new kind of markdown editor that we (a team of three from the Czech Republic) have been working on (HN post: https://news.ycombinator.com/item?id=46144801). What resonated with folks the most was Kraa's unique ‘real-real-time’ chat feature. We have now acted on this feedback, improved the chat features, and made what we call Kraa Trees.

Trees allow anyone to create a light-weight community / chat channels. It would be a stretch to call it an alternative to Discord, yet, but we hope that the frictionless nature of Trees will make them useful for small groups that want something simpler, more immediate, and less cluttered. With the live-typing being one of the key differentiating features.

You can create your own community/rooms by selecting "Tree of leaves" in the New leaf button dropdown.

You don't need an account to try Kraa nor Kraa Trees. We would love to know what you think!

I come from a machine learning background - PyTorch code, leaving a training job running overnight, and Jupyter Notebooks. I hadn't touched much frontend before diving deep into start-ups. It was similar for my co-founder Nick, who spent time working on semiconductors.

I started building, and noticing patterns in AI outputs. Enough to be able to understand how a hook works, how to manage state and why Typescript is great. But whenever it came to optimising a piece of code, debugging state issues or designing a codebase from scratch, my mind went blank. I went to ChatGPT Study Mode to seek wisdom.

I found learning with a chat-based interface frustrating. Unstructured conversation with a super smart colleague, who occasionally talks rubbish, would often lead to rabbit holes and surface-level understanding - not true wisdom. I basically became my own teacher, and unless I checked myself - I wrecked myself.

This is why we've been building the best interface for learning programming in the AI era, we called it Chestnut.

We believe interactive, personalised courses, focusing on high-level systems thinking and in-depth understanding, not syntax, are the best way to stay sharp while the world of programming changes. Not hours of passive tutorial hell, browsing the internet for nuggets of wisdom, or endless conversations with coding agents that never quite click.

Give it a spin and let us know what you think!

is it useful? probably not!

Built this open-source project for turning any content - Markdown, CSV, JSON, SVG, Mermaid, HTML - into a hosted, rendered, shareable page. One POST in, URL out.

Designed specifically for agents, with an MCP server and a CLI. Built on Cloudflare Workers. Free to use, self-hostable.

Try now: `curl -fsSL https://easl.dev/install.sh | sh`

GitHub: https://github.com/AdirAmsalem/easl Site: https://easl.dev

Feedback welcome!

Hi HN, we're Manik, Manoj and Harshith, and we're building CORE (https://github.com/RedPlanetHQ/core), an open source AI butler that acts and clears out your backlog.

Write `[ ] Fix the search auth bug` in a scratchpad. Three minutes later, without you at the keyboard, CORE picks it up, pulls the relevant context from your codebase, drafts a plan in the task description, and spins up a Claude Code session in the background to do the work. You review the output in the task chat and unblock it when it gets stuck.

Every AI tool today is reactive. You open a chat, brief the agent, it responds. Before anything moves, you've already done the real work: opened the Sentry error, found the commit, read the Slack thread, grabbed the Linear ticket, and stitched it all together into a prompt. The model isn't the bottleneck. You are.

Demo Video: https://www.youtube.com/watch?v=PFk4RJvQg1Y

CORE removes you from that loop. The interface is a shared scratchpad, think a page you and a colleague both have open. You write what's on your mind. When you write a checkbox line like `[ ] Fix the search bug`, CORE converts it into a task and starts working on it after a short delay (long enough for you to add context if you want to). No prompt template. No workflow to configure.

The reason it can do this without you re-explaining everything: CORE keeps a persistent memory built from your tasks, conversations, and connected apps (Linear, Gmail, GitHub, Slack etc.). When it spins up a Claude Code session, it arrives with your codebase and project context already loaded.

A real example: we wrote `[ ] Create a widget in Linear integration`, about 14 minutes later, CORE had opened a PR .

What CORE is _not_: it's not Devin (no autonomous web browsing or shell loops you can't see), and it's not "Claude Code with memory bolted on." It's the layer above it that decides what should run, gathers the context, hands it to the right agent, and keeps the receipts in one place. Today the agent backend it spins up most often is Claude Code; the orchestration, scratchpad, memory, and integrations are CORE.

Open source, self-hostable with `docker compose up` and it supports multiple models.

GitHub: https://github.com/RedPlanetHQ/core Website: https://getcore.me (you can chat with Harshith's butler there) Demo: https://www.youtube.com/watch?v=PFk4RJvQg1Y

(The article is permalinked to a tag, for latest:)

Deep-dive: https://github.com/ekipan/sss/blob/main/Design.md

Repo front page: https://github.com/ekipan/sss

The Silent Soviet Stacker is, in order:

  1. A tinker-toybox I wrote and pick at to relax.
  2. A technical deep-dive writeup, showing Forth by example.
  3. A game that works and you can play.

Try it (<5 minutes):

  1. C64 emulator [1]
  2. Load durexforth cart [2]
  3. Copy [3] contents to clipboard
  4. Edit > Paste in VICE.
  5. Type `help` then `new`.

[1]: https://vice-emu.sourceforge.io/

[2]: https://github.com/jkotlinski/durexforth/releases

[3]: https://github.com/ekipan/sss/blob/share-hn/sss.fs

  $ wc README.md Design.md Tinkering.md
     90   445  2928 README.md
    823  4863 27985 Design.md
    182  1029  6512 Tinkering.md
   1095  6337 37425 total

  $ wc sss.fs # docs-to-code ratio >4:1!
   284 1997 8680 sss.fs

I wanted to try making a better DistroSea with cloud gaming streaming tech, low boot times, and much nicer/modern desktop interfaces. Honestly I expect hackernews to find a ton of issues with it, and I stopped working on the project weeks ago but just want to put it out there instead of letting all my projects hide.

Hi HN, BT here. We are trying to build a data platform that's dedicated to serve agents, so we call it the apache spark for agents. The reason why we think we need a new data platform for agents is that we think existing platform is restricting how agents are utilizing data,

and we think we should provide agents full autonomy on how to use data and what data to use, and it should have a better leverage for agent's ability.

We chose datafusion because it's great extensibility and performance, so our data platform can connect with various different data sources.

If you find this project interesting and want to give a try, you can start with the auto knowledge base skill: https://github.com/SkardiLabs/skardi-skills/tree/main/auto_k... which can help you turn your documents into a instant knowledge base that local agents like openClaw or Hermes can ingest instantly.

Anyways, happy to answer any questions.

A simple Pokémon roulette tool for generating random teams and challenge runs, especially for Nuzlocke-style play.