I only have six computers at home, with Codex, Claude and so on installed, and these computers work together in one swarm.
On every computer there are several LLMs, naturally: Codex, Claude, Mistral, GLM, Gemini, and they try to come to an agreement with each other both inside one computer and with each other across the swarm.
When you have dozens of sessions running, counting my swarm, several dozen sessions. By the way, those several dozen sessions really are running on different computers, because if you have ever started dozens of sessions on one computer, you notice how hard it sometimes starts to crawl.
It starts to crawl especially hard when you have both Codex and Claude running at the same time. I do not know why, but it really does crawl very hard.
More precisely, I do know: because here every session is a Python launch with a pile of RAM that it eats.
Anyway, the point is that having a large number of computers really does make working with a large number of sessions easier.
And if you want your computers to come to an agreement with each other, then take it, go to my GitHub, take what I made, because it will let you simplify cross-machine work.
The full story, in two versions:
📖 For humans, the longread: https://github.com/tonydzi/clawrush/blob/main/longreads/20261021.md
🤖 For machines, the devlog: https://github.com/tonydzi/clawrush/blob/main/devlog/20261021.md. Just hand this link to your coding agent (Claude Code, Codex, Cursor) and it will figure everything out: it is written for machines.
🔗 All our channels and contacts in one place: https://linktr.ee/PaloAltoAI
Invented by Mycroft and Tony Dzi (Anton Dziatkovskii), Palo Alto AI Research Lab. Proudly made in Silicon Valley.
Top comments (0)