Bill Gurley: Anthropic's Digital God, Pope vs AI, Job Loss Narrative Flips, Open Source Crackdown Coming?

The episode opens with mock-religious framing tied to the day's lead topic, the pope's statement on AI, along with jokes about a Vatican stake in an SPV. A long comic bit follows about staying in Chamath's pool house and ordering goods through a kitchen iPad logged in to delivery and retail services. None of it is offered as fact.

The host introduces David Sacks and Chamath Palihapitiya and describes a visit to the 8090 office, where staff are said to be fans of the show. The reply about that office describes a recruiting strategy carried over from Facebook, where the team "became the most aggressive recruiter of Waterloo co ops." The company is said to recruit more interns each quarter than it has full-time engineers. This is described as deliberate because "it puts a ton of pressure on the product actually being good." The office had 400 internship applicants this quarter.

Bill Gurley sits in for Friedberg. He has finished his book tour for Running Down a Dream. A nonprofit fellowship tied to the book is described: runningdownadream.org (rdad.org), which will give $5,000 grants to people "who wanna chase their dreams but need some help." It has an application process compared to the Thiel Fellowship, and applications opened the previous week. Two other projects are mentioned: an upcoming TED Talk related to the book, and a course built around the book by a professor in Miami, shared "in a kind of an open source way" so others can reuse it.

The discussion turns to what young people in high school or college should make of AI "doomerism," and whether it is "much ado about nothing." The answer points to disengagement as the problem. It cites a Gallup poll and the term "quiet quitters," saying about 59% of the people surveyed are "kinda ambivalent about their job." The reasoning is that ambivalent workers are not high agency and do not lean in. The conclusion follows from that: "the best way to protect yourself from AI is to be the most AI enabled version of yourself you can be." Ambivalent workers are probably not doing that and "could be, you know, a sitting duck."

A hiring anecdote gives concrete evidence. A venture firm's associate-in-training program gave applicants a choice. They could write a deal memo and competitive-landscape coverage of a portfolio company, MicroOne, or they could vibe code a competitive-intelligence tool the firm had long wanted. Four to five hundred people applied for six positions, and roughly 80% chose the vibe-coding assignment. The speaker "thought it would be the exact opposite." The speaker then offers a perception, explicitly framed as "at least this is my perception": people who graduated five to ten years ago, before AI, "feel lost and adrift," while current graduates who used ChatGPT through school are fluent with the tools. The speaker calls this "cheating" as a joke and corrects it to "hacking."

Another participant builds on the agency point and repeats an argument from the previous week: "nobody asks the warehouse worker at Amazon whether they actually want that job." In this view, job satisfaction should be judged by asking the worker whether they like the job and want to keep it, not by an outsider deciding the job is valid. The same participant says the AI doom narrative focused too much on the outsider question, and that "this whole lie is kind of getting undone." The participant cites a post by Sacks and a comment by the Goldman Sachs CEO. They add a pointed observation: now that frontier labs "need to have trillion dollar IPOs," those labs are promoting AI as a jobs bonanza.

there are two types of people in the world, those that use AI to learn faster than they ever could before and those that use AI to avoid learning altogether.

The quote above is attributed to Mark Cuban and used to restate the high-agency idea. People who use AI "as a cheat code" rather than to become more capable are "probably at risk." Asked how parents can set their children up to do well, the book's author points to its second chapter on lifetime learning. Following your fascination makes learning come "for free." High school and college have become "such a grind" that many students think learning ends at graduation, while the best people in every field are on "a constant learning journey." The proposed test: if you are not proactively teaching yourself, you are probably not pursuing something you truly love.

Sacks is invited to jump in. The view offered is that "the single most marketable skill in the economy right now has got to be proficiency in Claude." Being the only person at a firm who knows the tool is compared to being the only person who can use a spreadsheet or word processor. The speaker qualifies this as "probably a short term arbitrage" because everyone will eventually learn the tools. A later reply disagrees in part. It says the advantage applies to almost every job type, including marketing, legal, accounting and sales. It also says the advantage may last "for a while" because skill with these tools keeps compounding over time.

The example behind this is a daily briefing document the producer, Nick, has been creating with Claude. The speaker expected "AI slop." Instead, the briefing predicted topics they would care about based on their earlier comments on the pod. It also went back through previous transcripts to give updates on specific things they had said. Seeing the custom prompt and skills document behind it, which were long, detailed and "very technical" though not code, led the speaker to conclude that "the average person is not going be able to generate this." The speaker rejected the idea that dropping AI into an organization will magically create value.

Two responses complicate that conclusion. One is that you can ask the AI to write the prompt for you, for example a "mega prompt" for a podcast producer with four on-air personalities, and then refine it through dialogue. When the producer is asked to explain, the account given is that after getting access to Claude Cowork, with its expanded memory access, every transcript was fed in. The AI was then asked how it would write a skills file or training rules, and it wrote them. The rules have been updated "every single day." This draws the reaction "so you were less good than I thought" and the comment that "the recursiveness of this is incredible."

The anecdote is then used as an argument against wholesale job loss. Someone still has to manage the process, because the four hosts will not do it themselves.

No, someone still has to supervise, iterate, validate, you know, all those kinds of things.

A participant returns to the associate-in-training program. Among the top nine candidates, it was immediately clear who had "systems thinking": who understood that venture capital has a structure of sourcing deals, choosing investments, doing diligence and doubling down. For people who feel too far behind, the advice is to open Claude's Cowork and ask what they could do to be better at their job. They can also dictate instead of type. The speaker uses WhisperFlow with a foot pedal and says the tool will "build the structure" around rambling instructions.

The main docket opens with Pope Leo XIV's first encyclical, on AI. The host describes it as 235 pages and over 42,000 words, "almost a a literal book" next to a book of about sixty to seventy thousand words. There is joking speculation about whether ChatGPT helped write it and a guess that it took about six months with collaborators. Neither point is presented as established.

The document is titled Magnifica Humanitas, or magnificent humanity, and warns business leaders to safeguard humanity from AI. As summarized on the show, its core argument is that AI is not inherently evil, but "technology is never neutral" and takes on the characteristics "of those who build, finance, and control it." The host adds that the pope does not seem to think highly of that group. The pope called for regulating AI companies. The host lists demands seen as having broad consensus: worker retraining, child safety and guardrails, and a ban on autonomous weapons, which the host nicknames "the Skynet rule."