It arrives as comfort. It is a warning about the shelf life of your verdicts.

Read as optimism, the line is about the machine getting better. Read as written, it is about you — and about the day you decided what the machine could not do.

That verdict is still running your company.

Someone tried it eighteen months ago on the task that mattered, got something embarrassing back, and closed the tab. The finding became a sentence in a meeting. The sentence became a policy nobody wrote down.

Nobody has re-run the test since, because the test already has an answer.

Ask around your own office and you will hear the exact phrasing — it hallucinates, said with certainty, about a version retired before the last two hires arrived.

That is how a company ends up holding a considered position on a tool that no longer exists.

It cuts the other way too. The workflow you built around today's limits — the review step, the second pair of eyes, the person who cleans up after it — is scaffolding with a shorter life than the budget line that funds it.

The hard part of Mollick's sentence is the tense. Worst you will ever use: today's version is the floor, and the floor rises faster than the quarterly cycle that decides what to do about it.

So the useful question was never what it can do. It is when you last checked, and on what.

So today, pull up the one task you wrote off — the analysis, the draft, the deck, the code you decided it botched — and run it again before lunch, unchanged. Whatever comes back is your floor, not your ceiling.