Ethan Mollick is an associate professor at the Wharton School of the University of Pennsylvania, where he studies the effects of AI on work, entrepreneurship, and education. He is best known as one of the most widely read explainers of practical generative AI use, through his Substack “One Useful Thing” and his 2024 book “Co-Intelligence: Living and Working with AI,” a New York Times bestseller. He co-directs Wharton’s Generative AI Labs, which build prototypes and run research on human-AI collaboration. He was named to TIME’s list of the most influential people in AI.
Recent
-
2026-08-10 Announced “Co-Existence: The Next Phase of AI” (Penguin/Random House, Oct 20, 2026) – a follow-up to Co-Intelligence (2024 NYT bestseller). Also named to TIME100 Creators 2026. co-existence.ai · TIME
-
2026-07-16 Updated his “harbor towns through history” benchmark to add Kimi K3, Inkling, and GPT-5.6 Pro, then made the interactive simulations publicly playable. X
-
2026-07-15 Flagged the anti-slop paradox: reading a viral “anti-slop” markdown file for AI agents, he found it was “basically 100% AI written” and would “hyperstition your outputs into mega-slop.” Bottom line: build your own skills/instructions with a real sense of taste rather than importing an AI-generated template. X
-
2026-07-15 “Between having AI as a subject for infinite speculation and using AI as a writer to produce such speculation at scale, it is the Golden Age of thinkpieces.” Then asked Fable to turn his tweet into a thinkpiece; called the result “pretty funny.” X · X
-
2026-07-13 Flagged OpenRouter token-share graphs as a poor proxy for real AI usage in an agentic world: “I think OpenRouter is not a good measure of actual model usage in a world of agentic tools… We really need better data indicators for AI!” — noting that usage is shifting to Codex/Code/Cowork without surfacing in token-share metrics. X
-
2026-07-13 Called for clearer AI roadmap communication, specifically about Anthropic’s Claude Code weekly rate limit extension policy: “Planning for how to use AI is a lot easier if there is some clarity about what to expect. Even a ‘we fully intend to keep extending this week by week but may need to stop under the following conditions’ would be better.” X
-
2026-07-09 Shared first GPT-5.6 demos after embargo lifted: compared Sol to Fable as “faster but works with you in steps more” vs Fable’s preference to go off and execute autonomously. Also flagged that OpenAI hasn’t reported Sol’s score on GDPval — their own benchmark for hard long-horizon tasks — calling the omission “confusing.” Was an early tester under NDA. demo tweet · Sol vs Fable comparison · GDPval thread
-
2026-07-02 Tweeted OpenAI internal usage data showing agentic task completion spreading beyond engineering to knowledge workers broadly: “The chatbot era is over, and agentic systems are coming to tasks beyond engineering.” Citing the same shift documented in his June 30 Substack piece but with new org-level usage numbers as evidence. X
-
2026-06-30 Published “The Twilight of the Chatbots” on Substack: argues the chatbot era is over and agentic systems are coming to work beyond just engineering. Cites METR and UK government estimates of programmer-hours-per-prompt as the key capability metric to watch. oneusefulthing.org
-
2026-06-28 Posted about a large China study showing AI homework use hurts test scores: “AI tutoring to support student learning is good, using AI to replace the effort of homework is bad.” X
-
2026-06-21 Documented the rapid capability arc in a single tweet: “June 2024: LLMs couldn’t count the r’s in strawberry. July 2025: LLMs got gold at the Math Olympiad. May 2026: LLMs solved one of the best-known questions in combinatorial geometry.” Referencing an OpenAI model cracking the Erdős planar unit distance problem (open since 1946). X
-
2026-06-19 Flagged the bright spot in OpenAI’s leaked 2025 financials: “OpenAI is profitable on serving customers with 40%+ gross margins. But training remains incredibly expensive.” Framed automating AI research as a cost-reduction play, not just a capability one. X
-
2026-06-09 Pushed back on a paper arguing prior ML systems’ lack of productivity savings predicts lower investment in current AI systems; called it “super weird” and noted it cites a “mostly fake MIT study.” X