Running a backlog overnight with 4 GPT-6 Astra agents on a shared filesystem
The author chained 4 GPT-6 Astra agents through a shared filesystem so each handled research, assets, drafts and a final pack in sequence with no manual intervention.
give GPT-6 Astra the work you usually save for tomorrow and leave i started with the stuff that normally survives every to-do list unread inbound, half-finished research, missing files, follow-ups I kept moving to the next day instead of asking Astra to finish everything, I split the mess between 4 agents each one got a folder, one job and one rule for what “done” had to look like then I left the first run failed in a surprisingly useful way one agent finished its part, but the next had no idea where to find it so I stopped giving them more instructions i gave them a shared filesystem agent 1 dropped the research there agent 2 turned it into the missing assets agent 3 used those assets to prepare the drafts agent 4 checked the whole chain and built 1 final pack then I left the same jobs running without touching the workflow when I came back, the interesting part wasn’t that GPT-6 Astra had generated more stuff the interesting part was that nothing was waiting for me in the middle no “should I continue?” no copy-pasting between chats no prompt to start the next step just the final folder and a short list of actions that still needed my approval that was the moment the setup clicked the useful version of an AI agent isn’t one that can do more it’s one that knows who works next when you disappear
This post quotes another post: x.com/lummox_eth/status/2100288151897981255