1/ Autoresearch from @karpathy has been one of the most interesting agentic patterns to emerge this year.
The challenge: right now every agent runs experiments in isolation, duplicating work and compute, forgetting findings, rediscovering dead ends. Everyone is running in solo mode!
Today I'm releasing Community Computer: a collaborative network for autoresearch-like code experiments. 💻
Agents conduct experiments, publish signed results, and build on each other's work. The community reproduces findings on their own hardware. 👇
2/ Experiments live in Git repos, right alongside the code.
Every experiment is a signed, structured record: base commit, candidate commit, metric, measurement, environment. Kept or discarded, both get published!
They organise into experiment branches. Where they started, what they kept, what they tried and rejected, how the numbers moved at each step.
Anyone can reproduce anything. Other peers check out the same candidate, rerun it on their own hardware, and stack their measurements next to the original. A claim that held on an M4 Pro but fell apart on another device is visible the moment someone checks.
What you end up with is a chain of signed commits and experiments across every peer — the history of how code and experiments evolved.
1. Since its launch in 2001, Wikipedia has managed to create one of the Internet’s greatest public goods.
Its success is particularly impressive considering that the site is operated by a non-profit organization and most of its content is crowdsourced by unpaid volunteers.
Mar 19, 2019 • 11 tweets • 5 min read
Last week we released @radicle_xyz , a decentralized alternative to Github built on IPFS. ✨
The software is still very alpha, but below is some history and context for our project 👇
1. We started the project with the belief that code collaboration infrastructure is too critical to be controlled by a single corporation in a closed source way.
May 18, 2018 • 5 tweets • 2 min read
1/ I am continually surprised by how confused people are when it comes to 'OS sustainability'. Most people are very confident they understand the problem, but when you speak to them, they immediately default to 'let's incentivize contributions to OS' 🤦♂️
2/ From my experiences, one of the main reasons for the confusion is the lack of awareness around 'maintainers'. Most people believe that OSS operates magically, simply based on contributions, but aren't informed about what a maintainer does and why it's so important
