Hexapodia LXIV: Understanding the Anthropic IPO & the Near-Term Future of AI

Noah is or is cosplaying an Anthropic-tech-financial bull. Brad is or is cosplaying a bear. But he reflects that he was a bear on the long-run supergrowth prospects of Google, Facebook, and Amazon too. And he was wrong. Dan Wang moderates a bull-vs-bear debate on the coming Anthropic IPO. Brad argues that Anthropic has exactly one proven paying business, coding agents. Everything else it does is either going to be give away free by the platform monopolists or depends for its long-term profitability on a “secret sauce” nobody has yet discovered. Noah argues the real prize is a change in industrial organization. If frontier AI makes business decision-making learnable, every firm will need to tie its capital structure to a frontier lab. Will the surplus from modern advanced machine-learning models (MAMLMs) flow to Anthropic’s shareholders, or will it be competed away to users and to hyperscalers determined to keep Anthropic from sitting between them and their customers (the “Netscape problem”)? We will see. Ideas:transaction-cost theory of the firm, economies of scope, contestable markets, Shannon information, stochastic parrots, the House of Intellect, the bitter lesson, the real-ASI as the Anthology Super-Intelligence pyramid-of-dwarfs that is the collected writings of humanity over the past 5000 years. Brad sent Noah “starting soon…” messages on every channel except the one he was watching! Coding agents are Anthropic’s one proven, monetized business.Search and summarization are wonderful but are given away free by platform monopolists for stickiness. Brad's has a bear case on LLMs.They are good at saying the next likely word because language is redundant; when ahumansays an unexpected word they're importing world-model or non-linguistic thought; when amodelsays one, it has drawn a random number from the tail of a distribution. One is information, the other is weird noise. The dark alternative.Maybe we are the sub-Turing ones: that it is actually language that is thinking and speaking through us, in which case we have no value added vis-à-vis our machines and Peak Horse-style replacement is the real question. Noah has a bull case on LLMsandon the Anthropic IPO.The core move is: business decision-making is mostlyconditional forecasting("if I enter this market, how will I do?"), and AI now wins forecasting competitions against almost any human forecaster. Most business jobs don't need creativity:"a very uncreative, very smart person who knows absolutely everything about the industry" will do the job well — and that's a description closer to a frontier model than to a human worker. Noah’s real thesis is about capital structure.In the Anthropic bull case world, every firm musttie up its capital structurewith a frontier AI company and become in some sense its subsidiary. The Excel analogy makes it concrete.Imagine premium Excel was available only to Microsoft subsidiaries, and everything done in Excel improved the product forother subsidiariesbut not for arm's-length buyers. Economies of scope across every industry. "AI is in the process of being invented".The biggest error to avoid: in June 2025 an Anthropic agent ran a vending machine that gave outfoie grasfor nothing. But its goal was pleasing its programmers and interlocutors, not making money. AlphaGo's move 37 is the template:humans built bounded systems with verifiable rewards (go, chess) leaving most of the "doors in the House of Intellect" connected to those symbolic constructins unopened for centuries. AI self-play in such formal systems opens the doors no human had. Brad's Bronze Age/Iron Age analogy.Coinage (~600 BC) madeanyonea possible counterparty, and market exchange became a superhuman distributed intelligence — Athens put its requirement tons of tin out to bid and it arrived from Cornwall via the Rhône, nobody in Athens knowing where it came from. The AI transition is the same kind of leap, in 40 years instead of 500. Brad's concession that undercuts his own skepticism.His contestable-markets and "IP edges are transient" argumentsshouldhave kept Google, Facebook, and Amazon from where they are; he doesn't understand their secret sauces, so he can't say Anthropic won't end up in their class. AI as the digital Walter Cronkite.Trained on the whole distribution, AI is "constitutionally moderate" and brings facts; you cannot make it a conspiracy-theory nut in one domain without making it a nut in general (Grok's "Mecha-Hitler" outcome). AI looks like asanity agentwhere social media proved to be aninsanity agent. Housekeeping. The next episode is scheduled forTuesday November 17th at 2:00 PM PST, with Ryan Avent on his book about faith. Anthropic. 2026.FormS-1 Registration Statement.SEC. https://www.sec.gov/Archives/edgar/data/2130357/000162828026068676/anthropicpbc-sx1ax1.htm. Avent, Ryan. 2026.In Good Faith: How the Nature of Belief Shapes the Fates of Societies.New Haven: Yales University Press. https://ryanavent.substack.com/p/an-answer-to-the-meaning-question. Bender, Emily M., &al. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–616. https://doi.org/10.1145/3442188.3445922. Booth, Robert.2025. “British AI Startup Beats Humans in International Forecasting Competition.”The Guardian. September 20. https://www.theguardian.com/technology/2025/sep/20/british-ai-startup-beats-humans-in-international-forecasting-competition. Borowiec,Steven. 2016/ “AlphaGo Seals 4–1 Victory over Go Grandmaster Lee Sedol.”The Guardian. https://www.theguardian.com/technology/2016/mar/15/googles-alphago-seals-4-1-victory-over-grandmaster-lee-sedol. Capoot, Ashley.2026. “Anthropic Releases Amended IPO Prospectus with Q3 Financials.”2026.CNBC. November 2. https://www.cnbc.com/2026/11/02/anthropic-ant-s-1-quarter-financials.html. Chen, Jackson. 22026. “Anthropic Files to Go Public”.Engadget, October 30. https://www.engadget.com/2296724/anthropic-files-to-go-public/. DeLong, J. Bradford. 2022.Slouching Towards Utopia: The Economic History of the 20th Century.New York: Basic. https://www.noahpinion.blog/p/book-review-slouching-toward-utopia. DeLong, J. Bradford. 2026.Grasping Reality.SubStack. http://braddelong.substack.com. Klee, Miles.2025. “Grok Calls Itself ‘Mecha-Hitler,’ Spouts Antisemitic Comments.”Rolling Stone. July 11. https://www.rollingstone.com/culture/culture-news/elon-musk-grok-chatbot-antisemitic-posts-1235381165/. Patel, Dwarkesh, 2025. “Andrej Karpathy — AGI Is Still a Decade Away.”Dwarkesh Podcast. October 17. https://www.dwarkesh.com/p/andrej-karpathy. Smith, Noah. 2026.Noahpinion.SubStack. http://noahpinion.blog. Stern, Joanna. 2025. “We Let AI Run Our Office Vending Machine. It Lost Hundreds of Dollars.”Wall Street Journal.December 18. https://www.wsj.com/tech/ai/anthropic-claude-ai-vending-machine-agent-b7e84e34. Wang, Dan. 2025.Breakneck: China’s Quest to Engineer the Future.New York: W.W. Norton. https://en.wikipedia.org/wiki/Breakneck:_China%27s_Quest_to_Engineer_the_Future. Vinge, Vernor. 1992.A Fire Upon the Deep. New York: Tor Books. https://en.wikipedia.org/wiki/A_Fire_Upon_the_Deep.