In a recent interview on Lenny’s Podcast, Instagram head Adam Mosseri suggested that Meta may need to implement caps on AI token spending per engineer within a year or two. He argued that as AI costs rise, they could become comparable to employee salaries, necessitating budget controls similar to payroll or operational expenses. Mosseri emphasized that managing token budgets would involve allocating resources efficiently across teams, much like distributing GPUs, CPUs, and storage. While Meta currently lacks such caps, he believes structured limits could improve cost management. The executive also noted that companies like Uber and Microsoft have faced AI spending overruns, prompting changes in their approaches. Looking ahead, Mosseri expects token costs to decrease as AI model providers engage in pricing competition. Meanwhile, Meta has already reduced expenses by discontinuing a controversial token spend leaderboard, calling it a 'silly thing' that created little value.

cut Fable 5 token usage by rendering text context as images - teamchong/pxpipe