Breaking down Grok Bot's multi-bot division of labor
Based on xAI's docs, the author maps out a Grok Bot setup where chief, tester, docs, patcher and Jev bots split the work, leaving the human to read the summary, click approve and merge.
holy sh*t. most people still use Grok Bot like a chat window one prompt, one answer, then an hour of babysitting xAI's own docs describe the opposite: one owner bot per job, a small crew under it, and you only show up at the end. i drew the whole thing out: > chief: holds the full repo in Grok 4.7's 500K context and owns the job start to finish > tester: reruns whatever failed and names the line that broke > docs: pulls the real SDK page before anyone writes a guess > patcher: ships the fix to a branch and kicks CI again > jev: the referee. every yes/no fork goes through it, typed answers back in milliseconds the rule i would steal first: docs isn't allowed to act alone. it asks Jev first that one line is why the crew doesn't loop all night your part of the whole run: -> read the summary -> click approve -> chief merges one click per job. that's the entire human workload honestly the hardest part for me was keeping my hands off the keyboard bookmark it, the setup i run is below ⭣
This post quotes another post: x.com/savipww/status/2102720919185617314