We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Two a16z partners recommend the same 20VC interview
Martin Casado called Harry Stebbings' interview with Diogo (@CompleteSkeptic) "Very good" and "Comprehensive." He said it "draws out a lot of Diogo's nuanced views from labs, AI, research, engineering, capital etc." Ben Horowitz linked the same video. His reason: "the number of original ideas per minute," and Diogo is "the best" . Watch
Gurley: an essay on AI "handoverism"
Bill Gurley called Ruxandra Teslo's piece "wonderfully thoughtful" . The piece is about the idea that humanity should, or inevitably will, hand control of the world over to AI. He shared it in a post that quotes Anthropic's Tom Brown and Dario Amodei on that idea. He argued that if the company's founders hold this view, it belongs in the Risk Factors section of Anthropic's S-1 . Gurley calls himself "decidedly anti-handoverism." He wants anyone shaping AI regulation to disclose where they stand on the question . Read
Rabois: Stratechery's "Obituaries"
Keith Rabois called Ben Thompson's latest piece "Still the best analysis in all of tech." He added that he often disagrees with Ben on China policy . The piece argues that Jensen Huang's demise was the end of Nvidia's long-term prospects, the AI bubble, and "the exceptional way in which the U.S. viewed itself in the world" . Read
Founders and operators
- Shaan Puri recommended his podcast episode with James Currier of NFX, who teaches a startup class at Stanford. Puri asked Currier to teach the class on the podcast for free . His main takeaway is that speed is the best proxy for founder success: how fast a founder decides, learns and iterates . Currier's advice is to reset your speed bar by spending time with people who move much faster than you do . Watch
- Alexandr Wang told followers to "listen for the inside scoop on muse" . He was pointing to Lenny Rachitsky's interview with Mona Sarantakos, Muse's head of product. Topics include why the team waited seven months to launch, the two levers that move activation for an AI agent, and why taste is now the most important PM skill . Wang is named in the episode as one of the people who shaped the product . Listen
- Guillermo Rauch called Vercel's post on agent evals a "Good read on how we do evals at Vercel." Note that it's his own company's post. His framing: users are now agents, so you run evals instead of user interviews, and the evals inform the engineering loop . The post covers seeing how agents use your product, testing changes before you ship, and comparing models as they launch . Read
- Sarah Guo replied "Phil knows" to Phil Venables' announcement of his book Scaling Security: The 7 Pillars of Elite Cybersecurity Teams. Venables says it covers how to lead a security team, work as a C-suite executive, and make an organization more defensible . Link
Classics cited on stage
- Patrick Collison, speaking at a panel moderated by Fed Chairman Kevin Warsh, said that as an immigrant he felt he should read Tocqueville's Democracy in America, and he just finished it. He pointed to the last chapter's warning about "soft despotism," where a comfortable state brings a "suffocating tranquility" that undermines ambition and great works . He tied this to his question of what a good AI future should look like . That continues his request in the Nov 14–15 brief for optimistic writing on post-work societies.
- Marc Andreessen called the "now famous Dynamo paper" from the economics literature "100% correct." He cited it to argue that new technologies take a long time to deploy because organizations have to change how they operate .
Also noted
Paul Graham shared Ben Todd's updated METR graph with a caveat: "If this is real, this is stupendously fast growth" . Post
- Jeremy Howard recommends George Pólya’s How to Solve It, calling it probably the best book of its kind; he says its problem-solving framework informed his own approach and is useful beyond mathematics.
- Howard praises an exploratory-programming video attributed in the transcript to “haml Hussein” as “really good”; no title or URL is provided.
- He advises checking documentation when AI introduces an unfamiliar programming or library feature, arguing that understanding it builds knowledge useful for modifying code.
Patrick Collison said he felt he should read Democracy in America and had just finished it; he highlighted its closing chapter’s warning that state-provided comfort could inhibit ambition, achievement, and great works.
- Jeremy Howard calls Attention Is All You Need a great paper, while noting it did not invent attention; he says it showed a new way to use attention more deeply than previously thought.
- Jeremy recommends reading Peter Wnjak’s 20 rules for formulating knowledge in SuperMemo. He warns that downloading ready-made material to memorize breaks rule one, and points to the minimum-information principle.
- Jeremy highlights Andy and Michael’s Quantum Country for people interested in quantum computing or mechanics: the book builds spaced repetition into the reading experience.
- Eric urges readers to read The Lean Startup while discussing sales, marketing, and talking to potential customers.
- Eric points people looking for business ideas to essays on the topic, including one by Paul Graham; he calls the essays great and says ideas often come from life experience and noticing problems that make you ask why things are the way they are.
Jeremy Howard’s summary of George Polya’s How to Solve It is shared as a resource; its four-step framework—understand the problem, devise a plan, carry it out and verify each step, then reflect—is presented as useful for coding and other tasks, not just mathematics.
- Jeremy Howard urged readers to “read this book and go deep” about Eric’s book. He called reading it one of his best experiences, citing its depth and interesting subject, and said it had not yet been released.
- Jeremy praised Carson’s co-authored book Hypermedia Systems as fantastic and deeply worth reading; he said it offers fascinating rabbit holes to explore.
- Jeremy called Seth Braver’s Pre-Calculus Made Difficult, The Dark Art of Linear Algebra, and Full Frontal Calculus unique and extraordinary, and worthy of close reading; he noted that he had not yet read Pre-Calculus Made Difficult.
Marc Andreessen called the “now famous Dynamo paper” in economics “100% correct,” endorsing it in a discussion of why new technologies may not spread automatically and can take time to change how organizations operate.
- Satya Nadella pointed viewers to “the blog that Bass wrote” and to blogs from the co-work and GitHub Copilot teams as useful software-factory examples; he highlighted a solo developer maintaining a runtime and a team whose size he estimated uncertainly as “nine or 12.”
- Nadella told listeners to look at Microsoft’s disclosed MI2 base pre-training recipe and regime, describing it as one of the best such disclosures; he said outside observers considered it the most detailed description of pre-training and called it a clean-room implementation.
Jeremy Howard cites How to Solve It as a major inspiration for the “Solve it with code” approach, while noting that the book is not specifically about AI.
In a Semafor interview, David Sacks praised a post by Tim Sweeney of Epic Games as “very good,” endorsing its argument that the episode raised concerns about how government relates to companies and individuals, and that government allegations can lead down a “dark path.”
Jeremy Howard recommended Daniel Pink’s To Sell Is Human and Influence (attributed in the transcript to “Robert Seondini”) for learning to influence organizations, after a carefully framed proposal failed to secure a project; he says influence takes more than assembling logical arguments and describes Influence as covering the academic study of what causes influence.
In response to a listener's question about AI-related attack vectors and Strike's security process, Jack directed listeners to Strike's blog for its pieces on software factories and red-teaming with agents. He described the related security approach as using AI tools to red-team Strike's own software and defend it.
- Jeremy Howard recommends the fast.ai course for learning how machine learning and stochastic gradient descent work.
- Howard recommends How to Solve It by George Pólya, a math-problem book whose problem-solving techniques he sees as relevant beyond math and coding.
- Howard calls Advent of Code a great place to practice: choose a suitably challenging problem to learn programming concepts, algorithms, and data structures.
- Howard says he will link a couple of videos he strongly recommends on personal development and building one’s craft, warning that neglecting this may put one’s future at risk; the video titles and creators are not identified here.
Sam Altman said he had just watched that weekend’s Saturday Night Live, found it “a little funny,” and loves the show.
- Jeremy Howard recommends watching videos by Andrej Karpathy, Chris Latner, or himself to understand why not to outsource all building to AI. Howard invokes Karpathy’s point that building from scratch exposes gaps in understanding and leads to deeper understanding; he warns that outsourcing all building can erode competence and confidence.
- Howard recommends Advent of Code as a place to practice solving coding problems; he says learners can adjust the difficulty and still learn from Day 1.
- Eric recommends a biography of Colonel John Boyd as “definitely the best introduction” to the OODA Loop, saying Boyd’s remarkable life—including his experience in the Korean War—makes the idea engaging; the biography’s title and author are not given.
- Eric calls a video demonstrating one-at-a-time versus batch mailing of 100 invitations one of his favorite teaching tools and his go-to illustration of small-batch work; the video compares the approaches and examines where the time goes. It had been taken down, and its creator and title are not identified.
- Jeremy Howard recommends George Pólya’s How to Solve It, calling it probably the best book of its kind; he says its problem-solving framework influenced his own approach and is useful beyond mathematics despite the book’s math focus.
- He recommends Advent of Code puzzles as practice for problem-solving: their well-specified challenges make a useful sandbox without requiring extensive libraries or specialized algorithms, and he urges learners to solve them themselves to build skill rather than immediately outsource the work to AI.
-
He recommends reading Python’s documentation and checking features suggested by AI; he says this helps learners discover capabilities—such as
intaccepting a base argument—and retain knowledge they can use to modify code. - He also points learners toward videos on exploratory programming, singling out an unspecified video by Hamel Husain as “really good.”
- Jeremy Howard recommends Rachel’s complete course on computational linear algebra for anyone who wants to study the subject deeply; he says it remains current and is worth knowing.
- Howard recommends studying Tinygrad’s tensor puzzles to learn broadcasting and tensor computation. The original puzzles were created by Sasha Rush (Howard thinks they were originally for PyTorch); Tinygrad has a version, and Sarah and Austin created a GPU C++ version. Howard says the puzzles build skills in broadcasting and concise tensor programming for high-performance GPU work, including deep learning; he also names the PyTorch puzzles as a good alternative.
- Howard recommends Advent of Code to people who dismiss it as LeetCode: he says its puzzles are carefully designed educational tools, citing past puzzles that taught virtual machines and convolutions, and describes them as enjoyable to work through.
Jeremy Howard recommends that viewers unfamiliar with T-Max try it, calling it “amazing,” and suggests working through the accompanying dialogue’s T-Max and SSH-forwarding walkthrough, including creating and naming terminal windows. The walkthrough is featured in “Controlling other computers - Solveit lesson 7”.
Jeremy Howard recommends exploring the shared Solveit lesson dialogue/notebook as follow-up material; the lesson plans to share Jono’s annotated version and the version built during class. He suggests beginners start with building a route and component, while readers interested in Python internals can explore topics such as globals and getattr; more generally, follow the parts that spark curiosity and dig deeper.
- Eric recommends his book The Startup Way to people dealing with resistance to change and bureaucracy in larger companies; he says it addresses what happens as startups grow into enterprises.
- For readers who know the term “early adopter” but not its origins, Eric points to books such as Crossing the Chasm, connecting the reading to the earlier terms “lead users” and “lead customers.”
Silicon Valley & The World: David Sacks, Co-Founder & General Partner, Craft Ventures | Semafor
In a Semafor interview, David Sacks praised a post by Tim Sweeney of Epic Games as “very good,” endorsing its argument that the episode raised concerns about how government relates to companies and individuals, and that government allegations can lead down a “dark path.”