>The wager the title refers to is that the big labs cannot build everything on top of their own models. Their customers are increasingly sick of paying for tokens, feeding the labs their data, and effectively training the company that may one day replace them, a dynamic one founder describes as not wanting to spend his life drinking Anthropic and OpenAI's water. So Guo bets on the application layer and on the picks and shovels, like Baseten, which lets companies run their own models and has grown revenue twentyfold in a year as token prices spiked.