Green tests, clean PR, and silence when you ask why.

You have seen it in review. Clean PR, green tests, and the author goes quiet when you ask why. A team at MIT Media Lab put EEG on that silence. People wrote essays with and without an LLM. The LLM group showed weaker neural connectivity across 32 brain regions, and minutes after finishing, 83% of them could not quote a sentence they had just produced. In the brain-only group, 11% could not.

It is a preprint, N=54, and the task was essays, not code. The exact numbers may not transfer. The mechanism, your team can already feel. The study.

Comprehension debt: the gap between what gets shipped and what the team can still reason about.

Green tests prove the code runs. They do not prove anyone on the team can change it. It shows up the night an on-call engineer opens a file nobody can explain. You can see where it builds, though: the PRs that landed fast with almost no discussion despite a large diff. In your next review, take the cleanest, quietest one and ask the author to walk you through a decision.

One for your next review:

A query, `SELECT * FROM events WHERE status = 'active'`, is slow on a 10-million-row table. The `status` column is indexed. Why might the database still scan the whole table?

You are listening for whether they think in terms of selectivity, not just "add an index." A strong answer says the index is close to useless when almost every row matches. A weak answer keeps trying to tune the index. If the room's only instinct is "add an index," that is the gap.

(The mechanism: an index only helps when the filter is selective. When almost every row is `active`, scanning the table once is cheaper than jumping to the index and back for nearly every row.)