80% of OpenAI and Anthropic's Revenue Comes From 1% of Customers

Eric and Neil say the rule of 7 has grown toward the rule of 28, and that buyers touch a brand 8 times for a sub-$100 B2C buy and 23 times at $10K-plus. A gelato shop stranger who knew them from different places makes the case that you never know where people will find you. They cover why YouTube drives leads that nobody can attribute, then Meta's $14.9B deal for 49% of Scale AI and the 200-person AI team that Alexander Wang, the former Scale AI CEO, leads. Zuckerberg's line that talent density beats big teams for AI is tested against Meta's 70,000 staff. Neil explains his private equity mindset and his debt at SOFR plus 2%, and they close on the failures that come with big bets. Attention, talent and the cost of big bets.

Eric and Neil start with X Money, which pays 3% cash back and 6% interest, and argue that the offer is under marketed and would get more people investing. They test an AI-written blog post on a client, who took it for human writing, then cover a time-on-page chart that compares human and AI content. They debate whether writing is thinking and whether a real name still beats AI copy. Dan Shipper's claim that newsletters are dying, from a writer with over 120,000 subscribers, leads to a debate on email, AI summaries in the inbox and whether YouTube is S-tier. They close on every channel getting more competitive over time. Human voice still earns the trust.

Eric and Neil open with Tesla's Roadster flying car, which they treat as a marketing stunt that keeps Tesla in the conversation. The AI fluency gap follows, with Karpathy's funnel, where most people barely touch an LLM, and agents that still misfire. Next is X Money and the newsletter debate, then a blog post test that a client took for human writing. The back half covers the rule of 28, Meta's 200-person AI team inside a 70,000 staff company, and why YouTube leads are hard to attribute. The show ends on Accenture's $29 monthly AI cap, Ramp's AI spend index and Bending Spoons' cuts at Evernote. Attention, trust and spend with a reason.

Some agencies win clients by inviting decision makers to Formula One boxes, but Neil prefers the way banks play the wine and dining game. Bank of America sent his CFO to a World Cup game with the hotel paid, and Wells Fargo offers suites and seats after the deal is done, just to say thanks. Then a competing bank tried to match Wells Fargo's refinance offer at the last minute, after months of saying no. Eric shares a group text pitch dressed up as comparing notes, and Neil tells the worst pitch he's heard: a financial advisor chasing a friend's $25M house sale and eight-figure profit with one cherry-picked account and a stock that barely moves down. A blunt episode on relationships and selling.

Eric built thumbnail, editing and posting skills, and his long-form YouTube videos went from about 500 views to 15,000 to 25,000. Some of the team wouldn't use them, so they had to let people go, and the one person who stepped up is now fixing the Marketing School channel's trends, thumbnails and 30-second retention. Neil says AI optimizes for average, which still beats a below-average team, but you need humans in the loop. They cover Bain's 94% stat, why Neil would rather hire great marketers and teach them AI, and a senior SEO candidate who read straight off a Claude template. Then Neil's Zappos story, how Uber outspent Lyft, why AI rewards whoever raises billions, and Neil's Muse outreach to M&A advisors. An honest episode on AI skills and hiring.