Writing about multiplayer AI coding.

Notes on what changes when a whole team works with one agent in one live session, and what we are building to make that the way teams ship.

Three agents, one repo: the hackathon problemDoors at 10, demos at 6. What breaks at hour three of a hackathon, what it costs by four in the afternoon, and what changes when the team shares one live session.

Anthropic wrote the AI-native SDLC playbook. Here is where a team runs it.The playbook's artifact chain, intent to spec to plan to review, is right. It is also a checkpointing protocol, and checkpoints are what teams write when they cannot share the session. The same lifecycle as a team sport.

Every prompt has a name on it. So does every commit.How authorship works when a whole team works with one agent on one machine: commits follow the prompter, pushes never carry a personal token, and pull requests open as you. And the thing we deliberately chose not to do.

Single-player agents made one person faster. The team got left behind.Every AI coding agent shipped so far is built for one person in one session. The case for a shared live session your whole team prompts at once, and why we built Coshell around it.

Multiplayer sessionsOverlap detectionPer-drive economicsBuilt on OpenCodeRunning on your own infraWorking with one agent as a team

Get your team in the same session.

Open a drive in the browser, or attach from the CLI. Your whole team works in one shared live session.