Show HN за 13 сентября 2025 г.

CLAVIER-36 is a programming environment for generative music. Programs are laid out in a two-dimensional grid, and evolve over time according to a fixed set of rules. The system is much like a cellular automaton, in that most of the rules governing the evolution of the system are local.

C36 programs describe sequences of discrete events in time. The environment includes a primitive sampler, as a self-contained means of interpreting these events as sound. For full expressivity, though, the system is best used as a generator of data for interpretation by an external musical instrument, such as a synthesizer.

The project was very directly inspired by Orca (https://100r.co/site/orca.html). It began as my own from-scratch implementation of Orca and diverged over time.

It's written in C, and compiled to WASM for the browser.

See the following pages for more info:

about page: https://clavier36.com/about

user manual: https://clavier36.com/manual

tutorial video: https://youtu.be/rIpQmJVMjCA

I ported Markus Kuhn’s wcwidth to TypeScript and optimized it with bitset lookups for O(1) performance. It now covers the full Unicode 15.1 combining ranges.

GitHub: https://github.com/dawsonhuang0/Wcwidth-O1 NPM: https://www.npmjs.com/package/wcwidth-o1

Feedback welcome!

i made a council of advisors to help me code. as a self taught dev, ive been heavily reliant on ai for the past two years. found myself often prompting claude to take on different personalities, so i built a web app. it's great for the step before telling cursor what to do and reviewing prs once theyre ready. PLEASE DON"T NUKE MY APY KEY. ty

Hi everyone! First time posting.

I built this mini memory game in an attempt to beat lazy memory syndrome.

because when was the last time you remembered what you did the day before purely based on memory (or without being prompted) ?

Super simple fun game for giving our brains a quick workout and actually reminding it on how to remember. If you had those old Nokia phones from the early 90s, the game will seem insanely familiar to you

Hi,

I have been building small AI Agents for quite some time now using various frameworks and one thing that always bugged me was that iterating on small things like prompts, flows, tools etc always took a code change + deployment of the app. While the prompt part can be solved with Langfuse I haven't found a good way to keep the flow management remote (and open source).

Lately I have been working with Genkit (https://github.com/firebase/genkit) and love how modular it is. So I thought why not build a UI builder on top of that that can handle simple flows, prompts and basic tracing.

And here we are with a first early version: https://flowshapr.ai Repo: https://github.com/flowshapr/flowshapr

This first release can - Manage and execute simple flows remotely - Works with GoogleAI, Anthropic or OpenAi - Integrate with remote MCP tools - API Endpoint to execute flows remotely - Flows and flow urls are compatible with the genkit client sdk

Upcoming - Support for Ollama - Support for various vector stores - More complex multi agent flows - Session management

Any feedback and suggestions are welcome!

I got tired of shipping megabytes of JS just to render small bits of dynamic UI.

So I built HMPL.js — a tiny (~24 KB) server-oriented templating language for JavaScript. It lets you send UI directly from the server with a simple block syntax, no framework overhead. Example:

<div> <button id="btn">Click!</button> Clicks: {{#request src="/api/clicks" after="click:#btn"}}{{/request}} </div>

It’s fetch-based (not XHR), supports events/forms, integrates JSON5, and sanitizes HTML via DOMPurify. Think of it as a smaller alternative to HTMX/Alpine.js when you just want server-driven interactivity with minimal JS.

Docs: hmpl-lang.dev

Would love feedback on where this could actually replace heavier setups ^ ^

I created a Nano Banana Prompt Generator. Instead of just browsing existing prompts, you can now generate fresh ones with a bit of guidance. The idea is to make it easier for people who are new to Nano Banana (or prompt writing in general) to get started.

For example, if you enter "3d figure" in the input box, a complete 3d figure image prompt will be generated.

Try it here: https://nanobananaprompt.org/prompt-generator

Thanks for taking a look!

I tend to watch TV shows months or sometimes years after they air, and by that time, all the discussions are lost in internet oblivion. I wanted a space where I could join the conversation anytime. So I created one.

MediaMouth makes it fun and easy to talk about the media you love. Conversations are organized season by season, episode by episode, so you never have to scroll through chaotic hashtags to join in.

We’re looking for feedback on usability, ways to improve, and hopefully gain some new users. We’d love to hear your thoughts!

You can download the iOS Beta or watch the video walk-through on our website: www.mediamouthapp.com

I’m using the excellent Bevy game engine for my colony sim/action game, but I do lots of real-time procedural generation/animation so the wgpu renderer is too slow.

So I wrote my own Rust/Vulkan renderer and integrated it with Bevy. It’s ugly, buggy, and hard to use but multiple times faster.

Full source code, with 9 benchmarks comparing performance with the default wgpu renderer: https://github.com/wkwan/flo

Video where I go over the examples and run them on the Steam Deck: https://www.youtube.com/watch?v=y1m30oOksmI

Hey HN,

I built AI Roast – a simple web app that takes your website and gives you a funny AI-powered roast.

Features:

Instant roasts in a lighthearted tone

"Pro Roast Report" (downloadable) if you want to take it further

Share button to drop your roast on Twitter/X easily

Live stats: total roasts, reports sold, and shares

Why: I wanted something playful to explore AI beyond productivity apps. Instead of “AI that helps you work,” this is “AI that dunks on you.”

Would love feedback on:

Roast quality (do they land?)

UX — what would make this more fun/shareable?

Any crazy ideas for distribution

Try it here ai-roast.jamatrix.io

Here is a computer science fiction story.

The cursor moves across the screen and opens a document titled "How to make a computer browse internet automatically"

The cursor reads the document, word by word, builds a knowledge graph, generates actionable scripts and injects to itself.

The cursor then knows how to browse the internet, starts browsing and accumulating knowledge 24/7.

The cursor reads about computers, hardware, programming, movements, does fact correction with the hardware it is built on.

The cursor builds a simulator to run simulation of knowledge in a sandbox environment before executing actions in the real browser or knowledge space. This allows efficient usage of energy.

The cursor does self optimization a.k.a goes into hibernation when the monitor is turned off.

To make the concept entangled, the cursor feeds from the light of the monitor as the energy source. This creates a balance in energy transformation.

https://ui-editor.com/f/how-to-make-a-computer-browser-inter...

I want to share my container automation project Proxmox-GitOps — an extensible, self-bootstrapping GitOps environment for Proxmox.

It is now aligned with current Proxmox 9.0 and Debian Trixie - which is used for containers base configuration per default. Therefore I’d like to introduce it for anyone interested in a Homelab-as-Code starting point

GitHub: https://github.com/stevius10/Proxmox-GitOps

It implements a self-sufficient, extensible CI/CD environment for provisioning, configuring, and orchestrating Linux Containers (LXC) within Proxmox VE. Leveraging an Infrastructure-as-Code (IaC) approach, it manages the entire container lifecycle—bootstrapping, deployment, configuration, and validation—through version-controlled automation.

- One-command bootstrap: deploy to Docker, Docker deploy to Proxmox

- Ansible, Chef (Cinc), Ruby

- Consistent container base configuration: default app/config users, automated key management, tooling — deterministic, idempotent setup

- Application-logic container repositories: app logic lives in each container repo; shared libraries, pipelines and integration come by convention

- Monorepository with recursively referenced submodules: runtime-modularized, suitable for VCS mirrors, automatically extended by libs

Pipeline concept:

   - GitOps environment runs identically in a container; pushing the codebase (monorepo + container libs as submodules) into CI/CD

   - This triggers the pipeline from within itself after accepting pull requests: each container applies the same processed pipelines, enforces desired state, and updates references

- Provisioning uses Ansible via the Proxmox API; configuration inside containers is handled by Chef/Cinc cookbooks

- Shared configuration automatically propagates

- Containers integrate seamlessly by following the same predefined pipelines and conventions — at container level and inside the monorepository

- The control plane is built on the same base it uses for the containers, so verifying its own foundation implies a verified container base — a reproducible and adaptable starting point for container automation

It’s still under development, so there may be rough edges — feedback, experiences, or just a thought are more than welcome!

After using Claude code for a few months, I set out to create my own version and realized that I spent most of the time creating a framework for organizing agent interactions. After a few more months of tinkering and playing around with ideas I created Wasmind.

Wasmind is a modular framework for building massively parallel agentic systems.

It's an actor based system where each actor is a wasm module. Actor's are composed together to create Agents and you can have 1-1000s of Agents running at once. Actors communicate through message passing where every message is broadcasted to every actor.

Wasmind can be used to build systems like Claude Code or really anything multi-agent you can dream of (examples included in GitHub: https://github.com/SilasMarvin/wasmind).

Wasmind solves a few key problems: 1. Modular plug and play 2. User-centered easy configuration 3. User-defined and guaranteed enforceable safety and agent restrictions (coming soon) 4. Allows easily composing any number of agents

You can configure it to use any LLM local or remote. I haven't tried qwen3-next but qwen3-coder (especially served by providers like Cerebras) has been incredibly fun to play with.

Thanks for checking my project out!

I used Vercel and Figma. Never thought I'd be able to do it.

Hi HN,

A while back I was trying to grow my following on X and Facebook, but I kept running into the same problem: I couldn’t consistently come up with good content ideas from time to time. Posting regularly can feel daunting, and replying to comments or threads often ends up taking more time than it should.

That frustration pushed me to build VibeReply — a Chrome extension that uses AI to help generate posts and replies directly inside social platforms.

Most AI writing tools live in a separate editor or dashboard. I wanted something lightweight, integrated, and fast — something that works where I already spend my time. That’s why VibeReply sits directly inside the platforms (X, Facebook, LinkedIn) instead of requiring copy-pasting from another app. It also focuses on making replies feel conversational and natural rather than generic. Every post or reply is unique, even when you use the same tone or topic multiple times, so it doesn’t recycle the same responses.

Why I think it stands out from other social reply apps:

- It’s fully integrated and works directly inside the three major social platforms — X, Facebook, and LinkedIn — so you don’t need to copy and paste between tools.

- It goes beyond simple reply generation by also helping you create fresh, engaging posts. You can choose from a variety of topics, generate an advice, a fun fact, or even a controversial statement in the selected topic and get unique content every time.

- It supports multiple languages, ensuring replies are generated in the same language as the original post for more natural conversations.

- Coming soon: the ability to set your own custom topics and tones, plus AI-powered meme generation for posts and replies.

I built VibeReply to solve my own problem with staying active on social media, and now I’m excited to see how it might help others facing the same challenge.

Please check it out here: https://www.vibereply.app

This is a demo video of using VibeReply: https://youtu.be/SefUGxi52Rk