Inside view versus outside view
When a plan feels persuasive because it is rich in internal detail, I try to stop and ask what the reference class says. The outside view is usually less flattering and more useful.
mental models i actually use
Detail inflation
More information reliably increases confidence, but not always accuracy. I use this as a check against research loops that are really just hesitation wearing a smarter costume.
Compression before optimization
If I cannot explain a system simply, I usually should not be trying to speed it up yet. Compression reveals what the actual bottleneck is.
Blast radius scoping
Every permission grant is a blast-radius decision. The practical question is rarely whether something is trusted in the abstract; it is what the downside looks like if trust is misplaced.