kukuri is a topic-first decentralized social network built for choosing communities, not platforms. Instead of putting one server or algorithm in control, it lets users discover conversations by topic, choose Community Nodes they trust for indexing and moderation, and keep identity and content portable across the network. It combines P2P gossip distribution with delegated trust, so decentralization doesn’t mean every user has to manage everything themselves.
How did Astra change the scope or ambition of what you built?
Maker
Astra didn’t change kukuri from one idea into another — it changed how much of the idea I could realistically push into a working product by myself.
kukuri was already a fairly large Rust + TypeScript/Tauri project before the challenge. With Astra, I could work across architecture, networking, frontend, CI, moderation, release engineering, and documentation without constantly narrowing the scope just to stay within what one developer could handle.
The biggest change was confidence in tackling messy, cross-cutting work. Instead of avoiding changes that touched several layers, I could investigate them end-to-end, update the relevant design docs and tests, and actually ship them.
That let me aim for a real Windows/Linux Builder Preview with P2P sync, Community Nodes, moderation, recovery, and release tooling — not just a demo of the core idea.
Report
Maker
📌
Hey Product Hunt! I’m the solo developer behind kukuri.
kukuri is a topic-first P2P social network.
You pick a topic, jump into the public conversation, and if you want, move into a smaller private channel without leaving that topic behind.
Your identity lives on your own device. Community Nodes can help with discovery, indexing, moderation, and connectivity, but they don’t own your account or act as the permanent home of your posts.
One thing I really didn’t want was the usual “decentralized = congratulations, now you’re the sysadmin” experience.
The goal is closer to: choose who you trust, delegate things to them, and change your mind later.
kukuri wasn’t built for the GPT-6 Astra Challenge. I already had a fairly large Rust + TypeScript/Tauri codebase, so this was less “build a demo in a weekend” and more “okay, let’s see how far I can push an existing weird P2P social network with a new model.”
One of the first things I did with Astra was clean up the repo instructions for AI agents. The project had accumulated ADRs, coding and testing rules, old lessons, duplicated instructions, and a few places where different docs were basically telling the agent different things.
That was probably my biggest takeaway from using Astra: better models don’t remove the need for engineering discipline. Once the model can handle larger and messier tasks, clear boundaries, good tests, and an actual source of truth matter even more.
During the launch sprint I used Astra across architecture, Rust, frontend, CI, moderation, releases, and a frankly unreasonable amount of bug hunting. Burned through my weekly limit and some extra credits along the way...
And now there’s an actual Windows/Linux Builder Preview you can install and poke at.
It’s still very much a preview. Things may break. Some parts are awkward. That’s kind of the point of launching it now.
If you try it, I’d really like to know:
• Does “start from a topic, not a server” actually make sense when you use it?
• Do Community Nodes feel understandable, or do they still sound like decentralized-systems homework?
• What breaks first?
Happy to talk about the architecture, P2P sync, Community Nodes, the AI-heavy development workflow, or any questionable engineering decisions I made along the way.