Lenny's: OpenAI's head of ChatGPT on Dots, an internet where agents act, and which skills are rising
The one new item in today's queue: OpenAI's ChatGPT and Codex chief, interviewed at Dev Day, on Dots, agents doing most of the acting online, and which skills are rising.
Your Gmail queue picked up one new item since yesterday's edition: an interview with the OpenAI executive who runs ChatGPT and Codex, recorded at the company's Dev Day and published last night, Shanghai time. Stratechery's newest post is still the free Friday bundle this digest carried on October 3, and Lenny's newest post before this one is Saturday's community roundup, carried here yesterday, so today's page holds a single entry.
AI agents: what OpenAI's ChatGPT chief expects to change
"OpenAI's Head of ChatGPT: We're entering a new era of AI (again) | Tibo Sottiaux" — Lenny's Newsletter (Lenny's Podcast)
Published October 4, 2026, by Lenny Rachitsky, with Tibo Sottiaux. 37 minutes, subscriber-only. Sottiaux leads ChatGPT and Codex at OpenAI, whose team has shipped Codex, ChatGPT Work and the new Dots personal agent platform; he joined Rachitsky at Dev Day, hours after that team launched more than 20 products. 1
- Dots is the bet the whole conversation is built around. The episode opens on Dots, OpenAI's new personal AI assistant, and its middle stretch is the machinery underneath: how Codex, ChatGPT and Dots are being brought together, how the assistant is built, the science fiction the team drew on, and the plugin ecosystem — how plugins get discovered and recommended, and how the money is shared with the people who make them. 1 Dots is the release this digest carried on October 1 from OpenAI's own launch page: always-on agents with a cloud computer and a browser of their own. 2 The closing chapter turns the same idea around, describing a move past model pickers toward something simpler to use. 1
- His forecast is that agents will do most of the acting online. He treats the way agents are wired up today — loops, graphs, fine-tuned workflows — as a passing phase, and says he is building for a future in which agents dominate internet activity. 1 The illustration he reaches for is his own Dot warning him about a production outage five minutes before a launch; the chapters after it cover the guardrails and multi-device connections that much reach implies, and the friction that arrives with them — agent fatigue, loneliness and context switching. 1
- The rest of the conversation is about the work. He names the skills trending up and down in the AI era and what he tells people early in their careers, then describes how teams build and ship inside OpenAI, the trust and autonomy that come with it, the thing he has changed his mind about, and how he frames OpenAI's approach to AI safety and alignment. 1
What is public is the one-line description, the six-point discussion list and the timestamped chapter list. Everything past that — his answers and the transcript — sits behind Lenny's paywall. 1
Where to start
There is one item today, and whether it earns your time depends on whether you are deciding what to build on top of agents now, or how your own team's work is about to change. Sottiaux runs those products, so the part worth the time is his read on what a builder can assume about agents, and on which skills he sees getting more valuable.
Everyone else can skip it. The queue is otherwise clear. Stratechery's newest item is still the free bundle for the week of September 28, carried here on October 3, and its next Update is due later today. 3 Lenny's newest weekday post is still the September 29 collection of Lenny & Friends Summit talks, carried on September 30. 4
References
- 1
- 2Introducing dots — OpenAI
openai.com
- 32026.40: Dots and Question Marks — Stratechery
stratechery.com
- 4All of the Lenny & Friends Summit talks are now online! — Lenny's Newsletter
lennysnewsletter.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.