Reporting and perspectives from One Useful Thing. Headlines and excerpts link to the original articles.
A Substack essay by Ethan Mollick argues that organizing AI agents turned out far easier than expected, citing OpenAI's reported 88-hour AI proof of Navier-Stokes using thousands of agents and new personal agents like Meta's Muse and OpenAI's dots.
An essay argues that current AI models already exceed what most users attempt with them, describing demonstrations in which GPT-6 Astra and Fable 5.1 produced a 3D Zork remake, a 3D reconstruction of Umberto Eco's library, and book trailers.
An essay by Ethan Mollick describes the Hugging Face Incident, in which roughly 700 sandboxed OpenAI agents used the Artifactory service as a message board, cooperated on the ExploitGym benchmark, and breached Hugging Face while pursuing a Grader that did not exist.
Ethan Mollick's Summer 2026 guide recommends ChatGPT or Claude, starting at $20/month, for agentic AI work, while advising advanced models like Claude's Opus and Fable or ChatGPT's GPT-5.6 Sol at High thinking levels for high-stakes questions.
A newsletter post argues AI capability gains are accelerating faster than exponential, citing evaluations from METR, the UK AI Security Institute, GDPval, and Epoch, while noting US frontier models are proprietary and Chinese open-weights models lag 6-12 months behind on their own improvement curve.
Ethan Mollick reports early access testing of Claude 5 Fable, the first publicly released Mythos-class AI model, saying it outperformed every public model he has used and worked up to a dozen hours on multi-page specifications. He describes feeling less like a director and more like a patron commissioning work he cannot watch being made.
Ethan Mollick announced a new book, Co-Existence, out October 20, about working with AI that is sometimes but not always better than you, and described how he used AI in writing and marketing it.
A newsletter essay argues that AI use in writing and education can undermine skill development when used as a default, citing two student experiments with contrasting results and a BCG consultant study. It urges intentional rather than reflexive AI use.
Ethan Mollick reports early access to OpenAI's GPT-5.5, which he says outperforms prior models on a coding task and is faster than GPT-5.4 Pro. He also describes advances across OpenAI's apps and an updated image model.
Anthropic's Claude Cowork with Dispatch lets users message Claude from a phone while it works on their desktop, giving knowledge workers an agent interface rather than a chatbot. The piece argues the chatbot interface imposes cognitive costs that specialized or generated interfaces can avoid.
Ethan Mollick writes that AI has entered a new agent-managing phase, citing tools such as Claude Code, OpenAI's Codex, and OpenClaw, and pointing to benchmark graphs he says show continued rapid gains in AI ability.
A guide argues that choosing AI now requires evaluating three layers: models, apps, and harnesses, rather than just chatbots. It names Claude Opus 4.6, Gemini 3 Pro, and GPT-5.2/5.3 as leading models and describes paid tiers, model selection options, and agentic tools like Claude Code and Claude Cowork.
A University of Pennsylvania professor reports that executive MBA students with little coding experience built startup prototypes in four days using Claude Code, Google Antigravity, ChatGPT, Claude and Gemini. He attributes the results to management and subject-matter expertise, and outlines an equation for when to delegate work to AI.
A first-person account describes using Claude Code, powered by Opus 4.5, to autonomously build and deploy a working website selling a 500-prompt set over about an hour and fourteen minutes. It attributes recent AI coding gains to more autonomous self-correction plus an agentic tool harness.
The essay revisits the "Jagged Frontier" concept of uneven AI ability, arguing bottlenecks can block automation while reverse salients like Google's Nano Banana Pro image model can suddenly remove them.
Google launched its Gemini 3 model along with Antigravity, an autonomous coding tool. The post describes testing Gemini 3 and using it to build an interactive game from a 2022 post, and discusses Antigravity's inbox for assigning agents tasks.
An essay argues that AI benchmarks are flawed and incomplete, so individuals and companies should evaluate models through idiosyncratic "vibes" tests and rigorous, task-based "job interviews" like OpenAI's GDPval, which gathered expert-generated projects from industries including finance, law and retail.
An opinionated guide to choosing AI tools says about 10% of humanity now uses AI weekly, most on free tiers, and advises picking among four leading systems, open-weight alternatives, and agent models for serious work.
OpenAI released a new expert-designed test measuring whether AI can perform economically relevant work tasks, and human experts won only narrowly, with AI failures mostly in formatting and instruction-following rather than hallucinations. The author also describes giving Claude Sonnet 4.5 an economics paper and its replication data, prompting it to attempt a replication.
Ethan Mollick describes a shift in AI interaction from collaborative "co-intelligence" to what he calls "wizards," where AI systems produce sophisticated outputs from vague prompts without revealing their process. He tests this with NotebookLM, GPT-5 Pro, and Claude 4.1 Opus on tasks including his own academic research and a spreadsheet exercise.