Tibo on Lenny's Podcast: give your Dot goals and feedback, not hand-built loops

#What changed

What Tibo said:

• On the Oct 4 episode recorded at DevDay, Tibo said that setting up and fiddling with loops and graphs is not where agents are heading. The Dots bet is an always-on agent that understands your goals and preferences and learns from feedback.
• Dots has no model picker. The only thing you configure is which channels you want to talk to it through. He said he wants reasoning-effort settings and model choices gone from the app as fast as possible.
• His own agent team grows and shrinks with model breakthroughs. Since getting an ultra-fast model he runs fewer agents in parallel and is back in flow.
• Codex writes most of his analysis code: trends, business questions, and how recent launches are performing.
• His Dot pinged him five minutes before the DevDay live demo to say ChatGPT production was down. It connected the event, the demo, and the production system on its own.

#Who is affected

Dots users (Pro at launch) and Codex users who build agent loops or multi-agent setups.

#What to do now

• Instead of writing a long step-by-step routine for your Dot, state the outcome you want and your preferences, then correct it when it gets something wrong so it learns.
• Connect the channels you actually check. That is the main setup Dots asks for.
• Tell your Dot what matters this week, like launches or big meetings, so it can flag related problems early. Keep production access behind guardrails, the way OpenAI runs its specialist dots with extra guardrails and monitoring.
• If you run many Codex agents in parallel, Tibo's pattern suggests trying a faster model with fewer agents before adding more.

#What is not confirmed

Verification: checked against the episode transcript at roughly 2:03 to 3:58, 8:20, and 20:56 to 22:18. The transcript is auto-generated, so points are paraphrased, not quoted. The loops point is Tibo's stated product direction, not a measured result.

#Sources

• Lenny Rachitsky's episode post on X: https://x.com/lennysan/status/2106772428861149683
• Lenny's takeaways thread: https://x.com/lennysan/status/2107131339355181240
• Episode page: https://www.lennysnewsletter.com/p/openais-head-of-chatgpt-were-entering
• Transcript: https://podscripts.co/podcasts/lennys-podcast-product-career-growth/openais-head-of-chatgpt-were-entering-a-new-era-of-ai-again-tibo-sottiaux