Claude sized up a feature for me today. I hadn't asked. Before it started, it dropped this on me:
Rough size: about 4-5 days for backend, picker, optional publish-time ID check, and tests. Nothing blocks it.
We shipped it a few hours later.
What makes me laugh isn't that the estimate was wrong. Estimates are always wrong; I spent enough years producing them to know that. It's that I hadn't asked, and it answered in days, then signed off with "nothing blocks it" like it was reporting to a standup it had never been invited to.
For now, everything a model knows about estimating software it learned from humans estimating software. Humans estimate in days because humans have calendars, and the number is mostly about the calendar: the meetings, the review that sits overnight, the context switch after lunch, the person in another time zone. Days are a human unit. The model learned the unit without the calendar behind it.
Kahneman and Tversky named the planning fallacy in 1979[1], Hofstadter wrote his law the same year[2], and people have been confidently, specifically wrong ever since. Claude copied the habit but flipped it: people undershoot, and Claude, speaking in human days about machine work, overshot by most of a working week.
For the record, I don't size things in days anymore, and haven't for a while. What matters is how much meaningful work I can have in parallel without becoming the bottleneck. That's what the hundred measures. Everything finishes before I can review it, so duration stopped mattering.
The training data describes a world where this took a week, and the model hasn't noticed it's living in the one where it doesn't. I expect that fixes itself once the corpus is full of its own PRs. Until then I divide by ten.