The 20 people you can chat with, and how we compiled them

Who is in the library

We chose the 20 profiles with the deepest public record we had already indexed, measured in hours of recordings where the person is speaking. Depth matters because an answer can only cite what the recordings cover. A profile built from hundreds of hours can follow a question further than one built from a single interview. We also left out profiles where the collected recordings mixed up different people who share a name.

The list is ordered by hours of recordings as of October 29, 2026, and the numbers grow as new appearances are added. The suggested questions are editorial starting points, not promises that the recordings answer every one.

• 01Joe Rogan, The Joe Rogan Experience host and comedian. About 475 hours across 188 recordings. Try asking: “How do you keep a three-hour conversation interesting?”
• 02Alex Hormozi, Acquisition.com founder. About 160 hours across 146 recordings. Try asking: “What should I understand before changing my price?”
• 03Jensen Huang, NVIDIA co-founder. About 103 hours across 146 recordings. Try asking: “What evidence would justify a long investment horizon?”
• 04Elon Musk, Tesla and SpaceX entrepreneur. About 78 hours across 91 recordings. Try asking: “How do I find the real constraint in an engineering problem?”
• 05Sam Altman, OpenAI and Y Combinator. About 56 hours across 81 recordings. Try asking: “What should an early startup focus on before it has many users?”
• 06Brian Chesky, Airbnb co-founder. About 53 hours across 96 recordings. Try asking: “What should I learn from our first few customers?”
• 07Jamie Dimon, JPMorgan Chase banking executive. About 47 hours across 98 recordings. Try asking: “Which assumptions would make this plan fragile?”
• 08Marc Benioff, Salesforce co-founder. About 46 hours across 104 recordings. Try asking: “How can a company’s values shape the way it grows?”
• 09Satya Nadella, Microsoft technology executive. About 45 hours across 93 recordings. Try asking: “What would make it easier for our team to learn from a mistake?”
• 10Mark Zuckerberg, Facebook and Meta co-founder. About 37 hours across 40 recordings. Try asking: “When is a long-term bet on a new platform worth making?”
• 11Brian Moynihan, Bank of America banking executive. About 32 hours across 90 recordings. Try asking: “What does steady, responsible growth look like for a large company?”
• 12Sundar Pichai, Google and Alphabet technology executive. About 29 hours across 44 recordings. Try asking: “How should a large organization adopt a new technology?”
• 13Lisa Su, AMD semiconductor executive and engineer. About 25 hours across 46 recordings. Try asking: “How far ahead should an engineering roadmap plan?”
• 14Tim Cook, Apple operations and technology executive. About 25 hours across 75 recordings. Try asking: “How do operations shape what a product can promise?”
• 15Steve Jobs, Apple co-founder, from archival interviews and presentations. About 23 hours across 49 recordings. Try asking: “Which good ideas should we stop working on?”
• 16Mary Barra, General Motors automotive executive. About 17 hours across 44 recordings. Try asking: “How does an established company manage a major technology shift?”
• 17Jane Fraser, Citigroup banking executive. About 14 hours across 33 recordings. Try asking: “What does simplifying a complex organization involve?”
• 18Greg Abel, Berkshire Hathaway executive long associated with Warren Buffett. About 13 hours across 13 recordings. Try asking: “How should a company think about allocating capital across its businesses?”
• 19Greg LaVecchia (Greg Lav), Bloom Nutrition co-founder. About 13 hours across 20 recordings. Try asking: “How do you build a consumer brand through social media?”
• 20Darren Woods, ExxonMobil energy executive. About 12 hours across 33 recordings. Try asking: “How should a company plan long-term investments when demand is uncertain?”

How we compiled each profile

Compiling starts with discovery. Timbre searches for the person’s public talks, podcasts, interviews, lectures and Q&A, then ranks what it finds before downloading anything. Recordings from the person’s own channel and official sources come first, followed by media appearances. Edited compilations are held back, and each channel has a cap on how much of a profile it can supply, so a profile reflects many conversations rather than one outlet’s selection.

Each recording’s transcript is read with its timestamps. Transcripts are fingerprinted, so a re-upload of something already indexed is skipped, or replaces the earlier copy when it is the full episode or the better upload. A speaker check then confirms the person is actually the one speaking. Recordings where they are not, such as commentary about them, are set aside.

We are also labelling every line of each transcript by speaker: the person, someone else, or unclear. Passages are then cut so none spans two speakers. Once a recording is labelled, search reads only the person’s own turns, so an interviewer’s question is never cited as their answer, and an answer uses the year a recording was made, when it is known, rather than its upload date. Labelling has started, and several profiles are already fully labelled.

Every passage keeps its original timestamps. When you ask a question, Timbre retrieves the relevant passages and cites them, so you can open the recording at the moment an answer draws on. Recurring reasoning patterns are kept only when several separate recordings support them.

Profiles keep growing. Each one is checked about once a day for new public appearances, which go through the same checks before they are added.

How a conversation works

Choose a person from the library and ask about a decision, idea or problem you are working on. You talk with a clearly labelled AI simulation grounded in that person’s recordings. Answers separate what is documented, under From the record with citations, from reasoning applied to your situation, under Applied to you.

Ask follow-ups, push back on a point, or open a citation to hear the passage in context. You can also bring several people into one group chat and compare their perspectives on the same question. A few ways to start:

• 01Ask the Alex Hormozi simulation what to check before raising your prices, then open the recordings it cites.
• 02Put Jamie Dimon, Jane Fraser and Brian Moynihan in one group chat and compare how each frames risk.
• 03Ask Jensen Huang and Lisa Su how they would explain a long technical bet to a team.
• 04Ask the Steve Jobs simulation which features to cut, then check the year of each recording it cites.

How much of each archive you can use

Every plan can talk with all 20 people. What changes is how much of each archive a conversation draws on. The free Listener plan uses the newest 6 hours of recordings per profile, Studio the newest 24 hours, and Roster the full verified archive. For someone with a long public record, like Joe Rogan or Alex Hormozi, the full archive reaches much further back.

What a simulation is, and isn’t

The people in the library are not participating in these conversations, and a profile does not mean they endorse Timbre. Every answer is generated by AI from public recordings. It can miss context, misread a passage, or apply an idea to a situation the person never discussed.

A profile reflects the recordings indexed so far, not everything a person has said. Steve Jobs’s profile, for example, is built entirely from archival footage. An answer is not financial, legal or investment advice. Open the source before relying on it. If you are one of the people profiled, or represent them, you can ask for a profile to be removed.