The Wall Street Journal — masthead est. 1889, house organ of American capitalism — posted this weekend that the era of AI “tokenmaxxing” in corporate America is over, and that companies are now embracing “thrift-maxxing.”

133K views · Metrics as captured July 27

Read that twice. Not a trend explainer, not scare quotes doing embarrassed work — the Journal’s own social desk, narrating a macro story about enterprise software budgets, in the suffix. One precision for the record, because we’re sticklers now: the coinages appear to be the social team’s framing rather than the article’s body text. That distinction matters and also doesn’t — the account speaks to 20 million followers with the paper’s logo on it, and it assumed every one of them could parse “tokenmaxxing” cold.

The actual story is real

Behind the vocabulary, the reported piece describes a genuine turn. Companies fed up with ballooning AI costs are trading down to cheaper models — including Chinese ones — and shopping à la carte instead of defaulting to the premium tier. A Cursor executive notes the most powerful models are wasted on mundane tasks. A senior technology executive at a major financial firm delivers the money quote: the free-money period for AI is definitively over. Some companies reportedly burned their entire annual AI budget in the first quarter.

Longtime readers will recognize the era being pronounced dead — it’s the one our tokenmaxxing ranking documents, the startups whose entire product is spending LLM tokens at scale. Just seventeen days ago we watched a well-followed tech poster advise that the smartest career move was to join whoever is “agent maxxing” with unlimited tokens. That post now reads like a artifact from a previous geological layer. Unlimited tokens went from hiring criterion to liability between two sweeps of our own feed.

The suffix survives the spending

Here’s what interests us more than the capex cycle: when the behavior died, the language didn’t revert. The Journal didn’t write that companies are “cutting costs” — the plain English that served business journalism for a century. It wrote that they’re thrift-maxxing. The bubble deflates, and the vocabulary just… swaps prefixes and keeps going.

That’s a stronger signal than any single adoption we’ve logged. The Pentagon borrowing the suffix was an event; Elon deploying it as an insult was a milestone; but the WSJ using it to describe the absence of excess means the framework has become load-bearing. In maxxing grammar, you never simply stop doing something — stopping is itself an optimization with a name. Restraint is a stat now. The most institutionally sober voice in American media just told executives they aren’t tightening belts; they’re maxxing thrift.

The forums built a language for compulsive optimization. Corporate America adopted it on the way up, and — this is the part nobody predicted — kept it on the way down. There is no exit from the skill tree. There are only new branches.

For where the AI money goes next, read today’s companion piece: the oracle thesis — because while the enterprises thrift-max, an AI researcher just named looksmaxxing as the industry’s next great consumer market.