Software Stocks Implode, Claude's Hit List, State of the Union Reactions, Trump's Tariff Pivot

The episode opens with a mock announcement of a "conspiracy corner" episode covering 9/11, flat earth and the JFK assassination, with Alex Jones as a pretend guest. The hosts then tease each other about view counts for an episode recorded during ski week with only one of the four regulars present, and about members' solo projects, including one host's own YouTube channel, which another host calls an "escape hatch". The band comparisons that follow ("The Beatles", "a little Yoko Ono situation") are banter, not news.

The hosts say the show's AI bot picked the dispute between the Department of War and Anthropic ("Dario versus Hagset") as the top topic. They defer it because the decision was due at the end of the day Friday, when the episode was released, and say they will discuss it the following week. The agenda they set instead: an expanding "Claude kill list" and "an AI fan fiction substack" that hit markets on Monday.

The discussion lists three February announcements from Anthropic, each followed by a fall in a different group of stocks. The hosts call Anthropic "three for three in tanking different market sectors in February" and joke that it has taken over the job of tanking the market from Brad Gerstner.

February 3: Anthropic announced a legal plug-in for Claude Cowork. Thomson Reuters, LexisNexis and LegalZoom were all described as down at least 10% since then. February 20: Claude Code Security was announced in a limited research preview. Security stocks fell, including CrowdStrike and names the transcript renders as "Cloud for Okta". February 23: Anthropic said Claude can modernize COBOL databases. COBOL is described as the oldest coding language in the world, still used in banking, payroll, government and health care.

The COBOL announcement comes with specific figures: COBOL runs 95% of ATMs in the US and powers Social Security payments, and 85% of all COBOL code runs on IBM machines. IBM is said to have fallen 13% on Monday, its worst day in a long time. The transcript garbles the comparison date and the size of the market-cap loss ("since 2,031 billion dollars in market cap losses"), so those numbers cannot be read reliably from the source.

The panel asks whether investors are just looking for an excuse to trim positions after markets hit all-time highs, or whether AI is really going to compress these stocks. The answer offers two explanations, one tactical and one strategic, without saying how much weight each one deserves, and says both are happening.

The tactical explanation is that many "smart money" hedge funds are degrossing on a large scale: cutting position sizes and taking on less risk. In a degrossing cycle, longs become less long and shorts become less short, which puts general downward pressure on prices. The speaker admits they don't know exactly why it is happening and suggests it could be driven by the structural change described next.

The structural explanation is about how markets price confidence in future cash flows. In a normal market, the debate is about when a company's cash flows (the examples are Coca-Cola, Eli Lilly and Meta) stop being highly reliable. Markets express that judgment in three ways. The first is the price-to-earnings multiple, whose inverse is the yield: 20 times earnings equals a 5% yield. The second is the revenue multiple. The third is the weighted average cost of capital (WACC), the rate used to discount twenty to thirty years of earnings back to today. A high WACC discounts future cash flows heavily. A low WACC assumes they will last.

We used to debate when. This is no longer a when moment. The market is very much in an if mode.

In this "if" mode, investors ask whether cash flows will last at all, whether they could "fall off a cliff in year three", and whether an AI model could wipe out a business without warning. That is event risk the market does not know how to price, so shareholders demand a large margin of safety. The illustrative figures are a P/E of 40 cut to 20 and 20 cut to 10, revenue multiples cut from 10 times to three times, and the WACC raised from 6% to 12 or 13%.

The most obvious knock-on effect, according to this analysis, is on how tech companies recruit and keep staff, because the squeeze falls on cash flows that are tied directly to stock-based compensation. The speaker adds that many of these companies may survive, but nobody knows for how long, and until that becomes clearer, investors "have to give yourself room to be wrong." A short follow-up clears up the wording: the old question was when cash flows would disappear, and the new one is "will they even exist?"

The second market driver discussed is a fictional Substack post set during a 2028 "global intelligence crisis". It was posted on a Sunday night, is described as getting 28,000,000 views on X, and is blamed for Monday's sell-off. In its scenario, companies adopt AI successfully, cut staff and raise margins. The summary compares this to Amazon trimming white-collar staff. The cuts then remove consumers' spending money, and a death spiral follows: firms deploy more AI to protect margins, cut more staff, and the economy collapses. Unemployment reaches 10% and the S&P falls from its highs.

The post also speculated that AI agents would avoid 3% interchange fees by settling transactions in stablecoins. Financial stocks fell on Monday: Amex down 8%, Capital One down 8%, Mastercard 6% and Visa 4%. The piece got pushback, notably for its claim that AI agents would "vibe code" their way to displacing DoorDash. The hosts call that claim silly to anyone who runs a network-based business.

One participant questions whether the post's virality was entirely organic. They cite "another post that just came out", which reported that the article's authorship was changed after publication to add a co-author. That co-author was described as a managing partner of a $262,000,000 SEC-registered hedge fund with short positions in companies the report named. The claim is presented as second-hand, attributed to that other post, and is not verified in the discussion.

The same participant points to Derek Thompson's article "Nobody Knows Anything", which they connect to a famous line by Hollywood screenwriter William Goldman. Its argument, as relayed here, is that nobody really knows what AI will do in two years, let alone twenty, so commentators fall back on "science fiction writing masquerading as analysis." Thompson describes the AI conversation as a marketplace of competing science fiction narratives. He says this is not because the technology is a parlor trick. It is because uncertainty is high and real-time data on AI's macroeconomic effects is scarce.

very serious conversations about AI are often more literary than genuinely analytical