The twilight of the chatbots
One Useful Thing · Ethan Mollick · 2026-06-30
Ethan Mollick argues that AI capability is growing at better-than-exponential rates and that the dominant paradigm is shifting from interactive chatbots toward long-running autonomous agents that can complete weeks of human work in hours.
Appears in
Extraction
Topics: ai-agentsai-capability-progressfuture-of-workagentic-ai
Claims
- Multiple measurement frameworks (METR, UK AISI, GDPval, Epoch) show AI capability growing at better-than-exponential rates.
- Opus 4.7 built a software package representing 2–17 weeks of human engineering work autonomously in 14 hours at a cost of $251 in tokens.
- A quarter of OpenAI employees have at least four AI agents running simultaneously at least once per week.
- Domain expertise — not professional background — is the strongest predictor of success when using AI coding agents like Claude Code.
- The US government has blocked public access to Claude Fable and GPT-5.6.
- Open-weights Chinese models lag frontier US models by 6–12 months but are on their own exponential improvement curve.
- Organizations with AI plans written before winter 2025 are describing a system far less capable than what currently exists.
Key quotes
Opus 4.7, working on its own for 14 hours, was able to build a software package that would take 2-17 weeks of human engineering work (it cost $251 in tokens).
We are moving from a world where non-experts use chatbots to fill in gaps to one in which experts use agents to get work done.
The instability is what happens when institutions that move at the speed of people (or worse, committees) try to track a capability curve that is very much not human in nature.