# Hello, concurrent.cd - where we research agents and humans working together

> Every website has its origin story; this is concurrent.cd's, from early conversations about overlapping agents to a first public preview.

Field note · published 2026-10-26 · https://concurrent.cd/notes/hello-concurrent-cd

Every website has its origin story. This is mine: three small things that triggered me to start concurrent.cd:

1. A conversation with Steve Yegge about agents coordinating through shared state and merges, in the factory-like systems of [Beads](/tools/2026-10-05-beads) and [Gas Town](/tools/2026-10-14-gas-town), made it clear how much planning and replanning costs before any code lands, which is why this site will focus on practical ways to structure parallel work.

2. The realities of git merges and integration pain underscored how hard it is to keep many agents aligned. Anthropic's post on [redesigning its source control](https://www.anthropic.com/engineering/redesigning-source-control) around a new system, Scion, and the open clones that followed, showed how far teams are going to fix it.

3. The talk [“A Computer for Every Idea: Growing Trees of Live Machines”](https://www.youtube.com/watch?v=XpZCyQobP-c) by Ondrej Drapalik and Mish Ushakov of [E2B](/tools/2026-10-05-e2b), at AI Engineer Paris 2026, showed what it looks like to explore multiple futures at once.

Through all of it, one word kept running through my head: concurrency. Concurrent. CI/CD was built for humans, for a handful of people taking turns at one codebase. Delivery now has to work for many agents at once, and the concurrency isn't only between agents. It runs between agents and the people steering and reviewing them, and between the people themselves, whose ways of working change when each of them runs a fleet. That is where the name concurrent.cd comes from.

Over the last couple of years, people asked me how AI applies to DevOps. I answered them by saying, I'll get to that when AI coding in dev has changed sufficient. Sure you could already have AI in Ops, but now I feel we have to re-imagine the delivery. And that's exactly what I see happening in the industry. Once you start running a factory things change drastically and it's no more just adding AI to our current processes. 

**That's why I'm so excited I pulled this site together.**

My first attempt at grokking thins was a weekend after the first thoughts. I hijacked my talk ["The Rise of Agent Enablement: Coding Agents Don't Scale Themselves, Neither Do Your Teams"](https://www.youtube.com/watch?v=wxNQRGPnJ24&t=1140s) (from 19:00) to tie the ends together. Given it was a first draft, it was very rough and raised some eyebrows :) Now I will refine the work as I write here and present at other events for sure.

This site is where the research lives from here on. I considered adding it to [my blog](https://jedi.be/blog), but thought it would become hard to track. I'll publish my own [notes](/notes), summaries and takes, track the [tools](/research?kind=tool) I find, and keep the quotes, links and videos I come across in the [research log](/research), so you can follow the whole research as it unfolds, not just its conclusions. I'll keep at it as long as it's useful. 

If you have evidence, tools, or stories that help, I’d appreciate the help. You can share them and we canlearn together! Or a friendly nod to encourage me to keep going 😅

On to more explorations!

## Sources

- [The Rise of Agent Enablement: Coding Agents Don't Scale Themselves, Neither Do Your Teams (Devoxx, YouTube)](https://www.youtube.com/watch?v=wxNQRGPnJ24)
- [AI Engineer Paris 2026, Day 1 Livestream, Creativity Room (YouTube)](https://www.youtube.com/watch?v=XpZCyQobP-c)
- [Anthropic: redesigning source control (Scion)](https://www.anthropic.com/engineering/redesigning-source-control)