Tokenmaxxing is so over. It's all about modelmaxxing now.

Ah, the great tech pivot: first, every employee was ordered to 'tokenmaxx'—bludgeon every task with AI until the tokens flow like cheap wine. Now, the very same founders are wringing their hands over 'wasted tokens' and demanding 'modelmaxxing' instead. Translation: we told you to go wild, but actually, you were supposed to use the right model for each job, obviously. Because nothing says 'strategic vision' like careening from one vacuous management fad to its opposite.

This is the classic Silicon Valley shuffle: invent a problem you created, then sell the solution. Tokenmaxxing was never a coherent strategy—it was a bet that sheer volume would unlock value. Surprise: it mostly just racked up API bills. Now modelmaxxing promises precision routing, as if engineers needed a memo to know you don't need GPT-4o for a calendar reminder. The consultants are rubbing their hands—this is a whole new matrix of decisions to monetize.

And what happens when modelmaxxing also disappoints? We'll pivot again, probably to 'sensemaxxing' or some other portmanteau that makes founders feel like they're steering. The real story is that companies still haven't figured out where AI actually adds value, so they keep rebranding the guesswork. But hey, at least the token waste is now a learning experience—on someone else's cloud bill.