Hey HN i’m a platform engineer and my day job I mainly focus on kubernetes clusters and terraform with the occasional C#, Typescript when building services, so this could well be completely out of my depth. Jolt started as an experiment to see whether I could build my own network at all. I’ll also be honest that most of this code was agentically written. However the high level decisions and architecture were mine based on research, experimentation, trial and error. Canary testing was one of the consequences of that when you build this way you need something slapping your AI-driven confidence back to reality.

This year I asked myself a question... Is it possible to have a platformless world? A world where no platform controls your data, mines it for ad targeting or bans you and takes away your audience with it. So I started working on that as a challenge just to see if it's possible... and I ended up with Jolt.

Jolt is a p2p distributed content syndication network. Lol that's a mouthful, it's basically my solution to platformless content distribution. Someone can publish content using a cryptographic identity and others can discover that content. Apps can build on top of the network however they are decoupled from the audience or the account.

Jolt achieves this with a daemon capability model kinda like how mobile apps ask for permissions, but for your identity and content.

Apps don’t own your identity. They ask for a scoped session from the daemon (eg. "publish under /pastes/*") that you approve or decline. You can uninstall the app and you still keep the audience and the data.

I'm gonna give you the jist of the work in this post and talk about the architecture in a comment but check out the RFC section on the github pages site if you want a deep dive.

When I was building this the actual hardest part was not protocol design or crypto. Funny enough it was testing. LAN is deceptively simple and things just work; why? mDNS will find your peers instantly and make you feel cracked. Only to be humbled when you do a 3 node canary test to try to replicate the messy real world on a small scale.

The 3 way canary tested my laptop on my home broadband, a relay in hetzner and my mac running on a mobile hotspot, finicky I know but that's all I had. I was trying to answer a hard question at v0: 2 nodes on different networks could communicate and more importantly behind actual NATs and how.

The first attempt was via the relay so I knew that worked but I wasn’t really satisfied. I wanted direct communication and I spent about 2 days getting QUIC hole punching working for direct node to node communication.

That being said im not fully sure it holds up in the wild, the realist developer in me doesn’t think so. One thing I'm sure about is that one canary test doesn’t cover all edge cases. I would need more nodes and people on the network telling me something is broken thats how I work lol and that's why I am posting this.

You can jolt find here https://alexanderwanyoike.github.io/jolt/

So does this thing even work. Well I can happily say that I was able to build 2 applications on it Spoke (https://alexanderwanyoike.github.io/spoke/) my attempt at building a platformless social network and Pastey (https://alexanderwanyoike.github.io/pastey/) a simple pastebin like application when you want to send data to different places. Spoke is more maintained and I plan on getting Pastey to the same level of polish. They are still very early “hello world” prototypes.

Otherwise if you wanna chat add me up on spoke fsjj2nvibhmztsw2w3klgkbjorqbze3qsi2tcwzbs5pepdigohiq.jolt ;)