Sara Wimmercranz recently opened a work meeting with contrast framing, a linguistic trick chatbots love. This is not a morning meeting, she began. It’s a reset. She had ingested so much AI-generated prose that it was coming out of her mouth. The Journal’s A-hed is that this is no longer a joke about other people. Conversation at her Stockholm venture firm, she told the paper, increasingly sounds like Claude or ChatGPT. In Sweden, she said, the culture would just be quietly listen. The model prefers fawning. She catches herself calling people brilliant.
The leak has a vocabulary. Will Barnett, 42, who recruits for a Bay Area startup, still swears he speaks like a human — then edits his own Slack so it sounds less like a model, which he notes is the same workflow as using one. Daniel Wolkowitz, at an AI startup, says he can no longer tell whether a message was written by a colleague with AI or a colleague who has started writing like it. The tells keep arriving: it’s not this, it’s that. These are the problems. This is how it happens.
Cave-man speak, and a whiteboard that bills you for it
Allie K. Miller, an AI adviser to Fortune 500 companies, was in a founders’ group chat that had a name for the hyperdirect prompting style: cave-man speak. It was bleeding into everyday talk. Why no bullets? She has a new appreciation for anything written before ChatGPT’s November 2022 launch, and for words a model would be unlikely to spit out. Cacophony, she told the paper, wistfully. Deepan Mehta, 33, built the equivalent of a swear jar — a whiteboard. Genuinely, load-bearing, smoking gun, X is the new Y, and hyphenated contractions nobody would normally use all earn a tally. Most marks at the end of the week? “You pay for happy hour.” David Bohigian, 56, a Washington nonprofit CEO, has a drier theory: the voice is the old business-book style, now at scale. He had read hundreds of those books before AI. The models have read those and more.
Our read
Behavioral: the tool you spend the day with becomes the voice you take into the room, and money conversations are not exempt. If the plan review starts to sound like a prompt — contrast framing, three bullets, a fake reset — it is worth noticing, because the household’s actual decisions are messier than that. A raise, a claiming age, a 529, a roof: none of those is a morning meeting that needed to be a reset.
The practical habit is Mehta’s whiteboard, applied to the statement. When a money sentence feels too smooth, ask whether a person said it or a model would have. Then put the number underneath. Fifteen minutes, bring the statement, talk like a person.