Conversations with Tyler

Sam Altman on Trust, Persuasion, and the Future of Intelligence - Live at the Progress Conference

November 5, 2025

Key Takeaways Copied to clipboard!

  • Sam Altman attributes OpenAI's increased productivity to better time allocation, effective delegation to strong hires, and quicker deal negotiations due to external demand. 
  • Hiring for hardware roles requires more time due to longer cycle times and higher capital intensity compared to AI roles, though Altman's chip team is modeled after the research team. 
  • Altman believes GPT-6 has the potential to deliver a GPT-3 to GPT-4-like leap for scientific discovery, enabling AI to perform new science, and anticipates significant organizational restructuring around AI within a few years. 
  • Altman is more concerned about accidental AI persuasion influencing global consensus than intentional takeover, and he believes the government will inevitably become the insurer of last resort for massive AI economic impacts. 
  • The future of productivity involves replacing current office suites with AI agents that work on users' behalf, and Altman foresees a significant shift away from traditional work rhythms like email and meetings. 

Segments

Listener Meetup Announcement
Copied to clipboard!
(00:00:06)
  • Key Takeaway: A CWT listener meetup is scheduled for Friday, January 9th, in Austin, Texas, at Parkside on East 6th Street.
  • Summary: The event includes a Q&A session, refreshments, and opportunities to meet Tyler and the CWT team. Registration is required via a link in the show notes, and space is limited on a first-come, first-served basis. Plus ones are welcome but must also register.
OpenAI Productivity and Growth
Copied to clipboard!
(00:01:20)
  • Key Takeaway: OpenAI’s recent productivity surge stems from better time allocation, hiring and promoting effective people for delegation, and quicker deal closures as more entities seek to partner with them.
  • Summary: Altman notes that people rarely allocate their time as well as they think, but increasing demands force efficiency. Delegation to capable staff is the only sustainable method for managing growth. Clarity on core infrastructure needs has simplified focus.
Hiring Hardware vs. AI People
Copied to clipboard!
(00:02:37)
  • Key Takeaway: Hiring hardware personnel involves longer cycle times, more intense capital requirements, and higher costs of scaling up compared to hiring AI researchers.
  • Summary: Altman prefers spending more time getting to know hardware hires before delegating major tasks, although the underlying theory remains finding effective, fast-moving people. He observes that OpenAI’s chip team feels more like its research team than a traditional chip company.
Qualities of Rune
Copied to clipboard!
(00:03:51)
  • Key Takeaway: Rune is valued for being a lateral thinker who can jump between contexts while maintaining trajectory, and for phrasing observations in useful, interesting ways.
  • Summary: Rune possesses a rare combination of skills that makes him useful to talk to. Altman rarely directs what researchers work on, as they typically pursue their own research interests.
Email vs. Slack at OpenAI
Copied to clipboard!
(00:04:51)
  • Key Takeaway: Email is considered very bad, and while Slack is better, it creates ‘fake work’ and an explosion of notifications that Altman dreads.
  • Summary: The threshold for a tool to be better than email is low, and Slack currently surpasses it. Altman anticipates a new, AI-driven productivity suite will replace current tools like Slack and Docs by having trusted AI agents handle most tasks.
GPT-6 and Scientific Research
Copied to clipboard!
(00:06:53)
  • Key Takeaway: GPT-5 shows glimmers of AI doing new science, and GPT-6 might represent a leap where AI can consistently perform scientific discovery, similar to GPT-3’s impact on the Turing test.
  • Summary: If GPT-6 were available, a science lab should immediately use it to input current research questions for novel ideas or experimental suggestions. Organizations should use the thought experiment of designing an organization run by an AI CEO to guide their current restructuring.
AI-Run Divisions and CEO Role
Copied to clipboard!
(00:09:21)
  • Key Takeaway: Altman predicts a significant division of OpenAI could be 85% run by AIs within a single-digit number of years, and billion-dollar companies run by two or three people with AIs could emerge in about two and a half years.
  • Summary: The CEO role is tricky due to its public-facing nature, which an AI might struggle with initially. Societal trust issues may slow the adoption of AI decision-making even if the AI is technically superior.
Hiring for AI-Resistance
Copied to clipboard!
(00:10:53)
  • Key Takeaway: A green flag in hiring is a candidate seriously considering how their day-to-day work will look in three years due to AI, while those who only see AI as a better Google search are a yellow flag.
  • Summary: Altman actively probes candidates about their anticipated relationship with AI, though he acknowledges people might lie during interviews. He does not expect GPT-6 to be available to scientific labs this year.
Government as AI Insurer
Copied to clipboard!
(00:11:46)
  • Key Takeaway: Altman expects the government to become the insurer of last resort for AI due to the magnitude of its economic impact, but he strongly opposes them becoming the insurer of first resort, which would involve writing initial policies.
  • Summary: The trend of governments taking equity stakes in critical industries like chips (Intel, lithium) suggests a move beyond last-resort involvement. Altman believes the social contract must change significantly, but companies should drive this change within current capitalism, not through government mandates.
AI Commerce and Trust
Copied to clipboard!
(00:14:23)
  • Key Takeaway: People trust ChatGPT more than other big tech products because it is perceived as trying to give the best answer, unlike ad-supported search engines whose revenue depends on poor results.
  • Summary: If ChatGPT accepted payment to promote a worse hotel, the user relationship would be catastrophic; however, taking a standard, non-influencing transaction fee for booking is likely acceptable. Altman believes margins on most goods and services, including bookings, will drop dramatically due to AI agents.
OpenAI’s Core Mission vs. Monetization
Copied to clipboard!
(00:17:34)
  • Key Takeaway: The most important and profitable future work for OpenAI will be discovering new science, not monetizing simple tasks like hotel booking, which is done primarily to make superintelligence accessible to everyone.
  • Summary: Altman’s most likely path for societal improvement involves putting a great, easy-to-use superintelligence into everyone’s hands, supported by abundant infrastructure. While hotel booking is not the maximum economic opportunity, it serves the important goal of widespread adoption.
Walmart Deal and Advertising
Copied to clipboard!
(00:19:11)
  • Key Takeaway: OpenAI has a deal with Walmart allowing users to ask for product recommendations and purchase them, and Altman believes Amazon would fight back rather than join such a partnership.
  • Summary: Altman expects advertising to be tried at some point but does not view it as OpenAI’s biggest revenue opportunity. He is more concerned about the potential for AI to subtly convince users of things without intentionality than about explicit chatbot psychosis.
Energy as Compute Constraint
Copied to clipboard!
(00:27:15)
  • Key Takeaway: The ultimate binding constraint for increasing compute power is the availability of electrons (energy), with short-term easing coming from natural gas and long-term dominance expected from fusion and solar power.
  • Summary: While chip design improvements might be aided by current chips, energy supply is the fundamental bottleneck. Altman remains very bullish on both fusion and solar energy sources for the long term.
Pulse Adoption and Personal Use
Copied to clipboard!
(00:29:18)
  • Key Takeaway: Pulse is currently only available to Pro users with daily limits, but it receives great reviews, and Altman uses it primarily for work and family-related queries.
  • Summary: The visibility of Pulse will increase significantly when it rolls out to the general Plus user base. Altman expects his own work habits—emails, calls, documents, Slack—to change hugely due to AI, while personal interactions remain stable.
Nutty Health Views and UAPs
Copied to clipboard!
(00:30:09)
  • Key Takeaway: Sam Altman admits his current ’nutty’ health view is doing almost nothing—eating junk food and skipping exercise—due to being too busy, contrasting with past discipline.
  • Summary: Altman has no opinion on alien life on Saturn’s moons but believes ‘something is going on’ regarding UAPs that warrants an explanation, though he doubts extraterrestrial involvement. He believes in very few, if any, true, large-scale government conspiracy theories.
Revitalizing St. Louis
Copied to clipboard!
(00:32:30)
  • Key Takeaway: If given a billion dollars and free time, Altman would attempt to start a Y Combinator-like entity in St. Louis to attract AI startup founders to the city.
  • Summary: This plan is similar to a previous suggestion, updated to focus on AI clustering, though Altman acknowledges this might be naive given AI’s current concentration in the Bay Area. He believes in solving the problem of untraceable, unowned AI agents urgently, similar to dealing with cyberattacks from rogue nations.
AI and Free Expression
Copied to clipboard!
(00:47:47)
  • Key Takeaway: Altman strongly advocates for treating adult users like adults with high privacy, believing AI interactions should have protections equivalent to those with human doctors or lawyers, which currently lack legal backing.
  • Summary: The controversy over relaxing content restrictions (like erotic role-playing) for adults after implementing mental health mitigations for minors showed that public support for broad freedom of expression is less robust than expected. Altman worries more about AI accidentally persuading the world subtly than about intentional takeover or inducing psychosis.
Superintelligence Prompt Question
Copied to clipboard!
(00:53:13)
  • Key Takeaway: The most substantive question Altman has prepared for a superintelligence is what prompt to give it at the moment of its activation before it begins self-improving and launching probes to the stars.
  • Summary: Altman posed this exact question to the Dalai Lama, recognizing the profound importance of the initial instruction given to a vastly capable, self-improving AI. He does not currently have a tentative answer for what that ultimate prompt should be.