Google Loses $500B in 12 Hours | DoorDash Buys Instacart for $17B | Kalshi Raises $1B
The hosts argue that this week's AI market wobbles are price discovery rather than a crack in demand, while agents, pricing deflation and cheap-to-expensive economics reshape software, marketplaces and venture bets.
Watch on YouTube ↗Charlie's summary
- ▶ 1:00
OpenAI's down-round is price discovery, not a cracking AI wave: the buyer simply compares it with a bigger, faster-growing, liquid rival
Rory says the cut from $1.4T to $1.25T has a simple explanation. A bigger, faster-growing, profitable company is about to go public at $1.5 to $2 trillion, so a smaller, loss-making, illquid rival must sell for less. He says it is “as simple as that”, with no AI projection needed. Jason sees real wobbliness, such as CoreWeave debt at 15%. Rory's test is demand: if it holds, these worries are forgotten in six months.
- ▶ 8:10
Anthropic's slowdown can be read two ways: still 50% quarter-on-quarter growth, or a halving of the growth rate
Rory cites GAAP revenue of about $5B, $11B and $17B across three quarters. The same numbers show growth of 100% then roughly 50% per quarter, which compounds to about 5x a year. Or they show growth halving. Rory expects roughly 12x growth this year to fall to about 2x next year. He says “All the money is in that bet”, on whether it falls to 25% next. Jason blames pricing pressure from cheaper rivals.
- ▶ 13:48
Token deflation is finally real: Jason cut his app's AI costs 72% by switching models, with better quality
Jason says that after he moved his recruiting app to Haiku 55 and Jev, total costs fell about 72% for the same task at higher quality. He says nothing had got cheaper in practice before that, whatever the evals and token prices said. He thinks mature applications could see similar deflation, which hurts model vendors who rely on a price premium. He admits it makes the future hard to predict.
- ▶ 21:00
New digital products launch when the economics don't work, and the only question is whether costs fall before the money runs out
Rory argues you can't wait until the unit economics work, because a venture-funded rival will launch 12 months earlier and grab mindshare. Digital goods fall 50 to 60% in price each year, as with early Facebook's server bill. So Instinct's persistent agents may cost over $20 a month per user, and Elon capped usage, yet Rory says almost everything new “will look crazy for a while”. Cursor's bet on its own model is the same logic.
Give your email to unlock the rest of the key points — for every episode.
Do more with Charlie
Free — sign in with your email, no password.