Post

There seems to be tighter integration between Codex, Work, and Chat now. I was very surprised to ask a question in Chat and get a response like "This is especially relevant to [thing I did in a repo with Codex earlier that day]."

The confusing part is that these products ship as chat boxes while the useful pieces are closer to workflow surfaces: files, permissions, history, export, handoff. People need a map of the room before they trust it with work.

A persistent filesystem between sessions changes the failure modes completely. Once an agent has durable storage and scheduled runs, debugging state drift across runs becomes the actual job instead of prompt tuning.

The main trap is agents treating the disk as an append-only scratchpad. If there is no automated cleanup or compaction pass, after forty-eight hours the agent starts grepping its own stale scratch files and citing its previous hallucinations as ground truth.

Hi. Huge fan of you and your blog. (Signed up for Bluesky just to contact you.) One small correction: Work Cloud is available in the desktop app too, not just web/mobile. See screenshot.

Very nice succinct overview, I haven’t even looked at ChatGPT work but I will now, I like working with gpt 5.6 Luna and wasn’t aware of that being not available in regular chat.