Claude is the only one I trust with a long doc

I didn't plan to write this. Then the tool did something that annoyed me enough to type.

Claude writes the calmest, most careful long-form of the bunch, and I lean on that more than the feature list.

The moment it let me down was specific: refuses more often than it should on harmless things. That's the kind of failure you remember.

What changed my habit wasn't a new model. It was realizing I'd stopped double-checking, and the one time it mattered, it showed.

The tool isn't the problem. The habit around it is.