Engineer runs five Grok Bots that write prompts for his coding agents
On day one of Grok Bot Galaxy, an engineer showed a five-bot team with separate memories that checks cloud agents every 5 minutes, writes the next prompt itself and spins up agents on red CI.
on day 1 of Grok Bot Galaxy an engineer showed a fleet of bots that writes the prompts for his coding agents five bots, split like a team: a chief of staff, craig on ui, steve on devx, hogan on infra, jenny on ops. same model underneath three engineers on one model because of context. each bot keeps its own memory, so a task lands on the one that already knows the area his p0 routine: > a bot checks every cloud agent every 5 minutes > an agent parks on sleep 300 while the tests finished in under a minute, 4 minutes gone > the bot kills the call and writes the next prompt itself ci runs the same loop. red build -> bot spins up an agent -> merge by his rules. on-call gets pinged only if it's still red after 10 minutes catch: a 5-minute check still lets an agent burn up to 5 minutes. and the loop needs a pass/fail signal, which he says himself
This post quotes another post: x.com/0xbobaaa/status/2092193059521843267