Most people read the Anthropic chart as red vs blue.
I added purple.
Red = current use.
Blue = theoretical coverage based on today’s job design.
Purple = how the job itself expands once more of the routine is handled.
The story is not just compression.
The pie gets bigger.
This #Anthropic chart looks less like a labor warning and more like a deployment map.
Most people will focus on where AI is already showing up.
I’m looking at the much larger surface area where capability exists but workflow hasn’t caught up.
@clairevo I’ve felt this for a while. Jira tickets magically closing. Engineers making features ready for review.
Coding agents are super powerful, but they don’t generate the same cognitive dissonance for people who already had someone else coding for them.
If I hear about one more AI agent for PR reviews…
Not everyone lives in git.
Most of the economy runs on spreadsheets, meetings, PDFs, and human ambiguity.
Did AI just have its “reveries” moment?
In Westworld, reveries were fragments of prior experience that shaped behavior.
Anthropic launched Claude agent “dreaming.”
Agents review past sessions, memory, transcripts, mistakes and adjust.
Reveries started as a feature too
@thsottiaux@steipete Codex has real enterprise potential, but admin needs to catch up. Enterprise Owners need a clear Codex admin path. Ownership ≠ admin took weeks to uncover. Need:metadata observability, central controls, managed runtime hooks, and Intune/managed deploy. Make rollout governable.