Ex-Palantir Analyst On The Inner Workings Of Intelligence & Markets In A World Of Socialism

Alex Goode worked at Palantir doing big data work before moving to Balyasny, where he was hired specifically for access to a Swift-related dataset that was later cut off due to compliance problems at Standard Chartered and HSBC. Palantir's original counterterrorism use case involved mapping associates, license plates, and bank accounts into an ontology to identify military targets in Afghanistan and avoid civilian casualties. After losing the Swift data, Goode built a trading strategy using advertising performance data as a real-time signal for Google, Facebook, and Twitter stocks, identifying a structural edge because fund managers holding hundreds of millions in advertising stocks had never run a single ad on those platforms.

Specific mispricings from that work were concrete. Booking.com spent roughly 70 percent of its margin on advertising while Electronic Arts could acquire a customer for a fifty-dollar game at a four-dollar ad cost yet was deliberately capping that spend to protect its margin story, causing it to trade cheaper on EV-to-sales than Booking.com despite superior economics. Tesla exhibited meme-like behavior in advertising data before meme stock terminology existed, with search clicks driven by queries about Elon Musk rather than the product. Advertising data also showed Amazon's retail business breaking even on ads, which Goode read as undervaluation and used to found Perpetua, which eventually signed clients including Crocs and Kimberly-Clark and began selling data to funds.

The broader market argument is that alpha has shifted from measuring the cheapness of selling a product to measuring the cheapness of selling the stock itself. The Soros reflexivity framework underpins this, with attention vortexes creating flywheels of real value. Tesla received meme capital and deployed it productively. GameStop started with zero fundamentals but now holds capital and a trading card business, which was Michael Burry's original thesis. Ethereum went from seven dollars to hosting Tether and Circle. The attention vortex model breaks when liquidity is fragmented by competing assets, which is why launching a product called STRC alongside MSTR would be bearish for Saylor, why an Anduril IPO would hurt Palantir, and why Bitcoin has an advantage over Ethereum and Solana because it faces no credible fixed-supply competitor.

On Palantir specifically, the business is split roughly 50-50 between government and commercial, with Gotham serving security and Foundry plus ontology products serving enterprise. Palantir's commercial argument is that large language models require its ontology layer to function in enterprise settings, but one speaker disputed this, arguing LLMs can be pointed at a codebase without structured ontologies and still perform effectively. OpenAI and Anthropic now deploy forward-deployed engineers similarly to Palantir, reducing its competitive uniqueness, and as models improve the need for elaborate ontologies decreases. Palantir trading at 100 times sales faces headline risk as competitors fragment the market.

AI-driven unemployment has not yet appeared in headcount data despite market pricing. Accenture stock dropped roughly 60 percent including a single-day fall of 26 percent while headcount remained near 800,000. Figma dropped 85 percent while headcount rose. Software company multiples are now below commodity company multiples, reversing the historical relationship. The speaker predicted AI unemployment would begin showing in employment numbers within approximately three months, and argued companies with credible AI turnaround plans could see extreme terminal EBITDA margins once layoffs begin. Annualized productivity growth is approximately 0.6 percent in the recent quarter and 2.8 percent year-on-year, far below the 10 percent annual gain the Microsoft CEO claimed AI would deliver.

The speaker's central thesis, called the Black Print, holds that the US political system is not designed for hyper acceleration and that markets are wrong to assume right-wing dominance and unchecked AI deployment will continue. NVIDIA has 50,000 employees yet is worth more than the entire Russell 2000, which employs millions, and Russell 2000 workers vastly outnumber NVIDIA shareholders as voters. Trump's regulation of Anthropic and the banning of port automation are cited as evidence that even the current administration is normalizing AI regulation. The speaker draws a parallel to nuclear technology, where the government prohibited certain research starting in 1945, and argues the same pattern is imminent for AI. The midterm elections are identified as the next major checkpoint for this thesis.

The speaker is bearish on humanoid robotics deployment within five years, citing Waymo being pulled off roads after incidents as the regulatory template and Trump's port automation ban as evidence that basic blue-collar automation is already politically unacceptable. AI is already unpopular before significant job losses have materialized. The speaker views the AI bubble as close to an end given these political dynamics, rejects the mainstream assumption that AI productivity gains will resolve the debt burden, and holds crypto primarily as a hedge against a sovereign margin call if that productivity story fails to materialize. Companies with canonical IP including Nintendo, Disney, Games Workshop, and the Star Wars franchise could unlock significant value if AI-generated content using their IP requires future licensing payments.