Overview
They start by explaining what tokens are and why they are important. Many AI companies charge businesses based on the number of tokens their employees use. When employees use too many tokens, AI costs can increase very quickly.
Skip then explains how agentic AI is different from normal AI prompts. Instead of doing one task, agentic AI can work more independently. It can search for information, make decisions, check results, and repeat tasks many times. This can be very useful, but it can also use a lot of computing power and become expensive.
Next, they discuss several large companies. Uber reportedly spent its yearly AI budget in only four months, which led to strict monthly token limits for developers. Amazon stopped an internal AI leaderboard, and Microsoft canceled many internal Claude Code licenses after AI costs increased too quickly.
Finally, Skip and Dez talk about the bigger business impact. Companies are no longer focusing only on how much AI employees use. Instead, they want to measure how much useful work AI produces. This idea is called Inference Yield. This change could have a big effect on AI companies, especially companies like Anthropic and OpenAI as they prepare for possible future IPOs.
