No Priors opens the room where AI gets built

In April 2023, Sarah Guo described an experiment. She and Elad Gil would record an episode of No Priors without a guest, just the two of them talking about technology, startups and AI. She asked listeners for questions. One wanted to know about unusual applications. Another asked about hiring prompt engineers. Someone else wondered whether specialized models could reshape finance and healthcare. The questions were already outrunning the vocabulary.

That little invitation captures the appeal of the media business they had launched two months earlier. AI was moving quickly; people needed a place to ask better questions. No Priors offered access to the people making consequential decisions, plus hosts who had reasons to understand the answers. Its product was a conversation, distributed through familiar podcast apps and YouTube. The doorway was ordinary. The room behind it was unusually well connected.

Two investors, one open door

Guo announced the launch as a “side project”. The February 2, 2023 trailer promised interviews across the AI stack, connecting technical progress with business implications and predictions. That is a useful founding choice: choose a field broad enough to keep changing, then organize it around people who have actually built something.

The hosts brought different kinds of experience. Guo founded the AI investment firm Conviction in 2022 after roughly a decade at Greylock. Gil co-founded Mixer Labs, acquired by Twitter, and Color Health. He also wrote High Growth Handbook, a practical book about running companies after they find a market. Between them, research questions meet questions about hiring, distribution, customers and capital. A technical breakthrough has to survive all four.

Their expertise also establishes the show's perspective. These are investors with portfolios, beliefs and professional relationships. Gil's public investment list includes OpenAI, Anthropic and Mistral. Knowing that helps a listener interpret the conversation. Access can yield better questions; proximity can also make a thesis feel more settled than it is. The sensible listener keeps both possibilities in view.

“Curiosity is alpha.”Sarah Guo, on her personal website

The product is the second question

There are plenty of ways to hear that a new model is impressive. No Priors becomes useful when the conversation moves downstream: what bottleneck remains, who buys the result, and what happens when an experiment reaches production? Its audience includes builders, investors and technically curious business people trying to connect developments that arrive in separate news feeds.

Consider the October 2024 conversation with Braintrust founder Ankur Goyal. The episode moves through evaluating AI applications, assembling a team, choosing languages and working with customers. That combination matters. A model can look convincing in a demonstration while leaving a business with the harder job of deciding whether its answers are reliable enough. The interview gives that operational work a place in the story.

The format also accommodates a wider lens. Andrej Karpathy's September 2024 appearance joined autonomous driving, robotics, model bottlenecks and education. His March 2026 return explored coding agents and AutoResearch. Read together, the episodes offer a way to watch an expert's agenda change. They are dated records of what seemed urgent, rather than a timeless instruction manual.

Back to the same room

Repeat interviews are an underappreciated feature of this kind of media. Jensen Huang appeared in April 2023 and returned in November 2024. A listener can compare what the NVIDIA chief chose to emphasize at different moments. The value comes from the interval: technology develops, markets respond, and yesterday's central question becomes today's assumption.

In June 2026, the show joined Latent Space for a conversation with Satya Nadella at Microsoft Build. Guo, Gil and swyx shared the interview. The subjects included AI platforms and the developer ecosystem. Here, a media brand assembled around two hosts could work alongside another specialist show inside a major industry event. The collaboration made its access visible.

By October 2, the public feed contained 181 entries, including the trailer. Its newest conversation featured Fractile founder Walter Goodwin discussing chip design and future workloads. The previous month ranged from diffusion models to transforming incumbent businesses. That range is the editorial proposition: AI has consequences across the stack, and understanding one layer often requires visiting another.

Pay with attention

For the audience, the direct price is simple: the published episodes are freely accessible. The scarce input is time. Recent conversations with Goodwin and Sequence Holdings' Michael Lee run approximately 36 and 43 minutes. A weekly habit is manageable; an indiscriminate attempt to swallow the entire archive is a different undertaking.

The visible model is publishing: a recurring show, video distribution, clips and email sign-ups. Apple lists Conviction as its creator. That association places it near the venture capital world, but a venture fund's capital is not podcast revenue. For understanding the product, the important exchange is attention for informed access. The listener gets context; the hosts build an ongoing public conversation around their field.

Within AI media, its alternatives include Latent Space and The Cognitive Revolution. Think of No Priors as the investor-and-builder conversation on your listening shelf. It is particularly useful when you want to connect engineering choices with company formation and market structure. Someone seeking a step-by-step coding tutorial will need documentation and practice alongside it.

Borrow the question, do the experiment

Here is a practical way to use the show. Pick one episode tied to a decision you face. Write down the guest's main claim, the constraint they describe and the evidence that would make you reconsider. Send those three lines to a colleague. A shared question is more useful than forwarding forty minutes with “thoughts?” attached.

Then do something small enough to check. Evaluate your own application. Talk to a customer. Compare the same guest's earlier forecast with the later interview. Predictions deserve that return visit. No Priors earns its place when listening produces a sharper experiment. The founders may open the door, but the useful work begins when the listener leaves the room.