The function you couldn't explain an hour later.
The feeling of knowing was real. The knowing was not.
- The track
- Writes code
- The rep
- About thirty seconds
- Published
- September 2, 2026
The Exhibit
You have shipped a function you could not explain an hour later. MIT Media Lab found the mechanism. They put EEG on people writing essays, some with an LLM, some without. The LLM group showed weaker connectivity across 32 brain regions, and minutes later 83% could not quote a line of what they had just written. Brain-only group: 11%.
Preprint, N=54, essays not code, so treat the exact figures as soft. The mechanism, you have already felt. The study.
The study
Comprehension Debt
Comprehension debt: the gap between what gets shipped and what the team can still reason about.
Accepting a suggestion is recognition. Recall is what builds the model you reach for when it breaks. Recognition feels like knowing, but it is not the same as being able to rebuild it, and the gap shows up exactly when the suggestion was wrong.
The idea
The Rep
Thirty seconds:
`SELECT * FROM events WHERE status = 'active'` is slow on a 10-million-row table, and `status` is indexed. Why does the database still scan the whole thing?
The answer
Most rows are `active`. An index pays off only when the filter is selective. When the predicate matches most of the table, the planner estimates from its column statistics that a sequential scan is cheaper than walking the index and fetching nearly every row. Selectivity is the lever, not the existence of the index. This is also why stale statistics bite you at 2am: the estimate, not the index, is what was wrong.
About thirty seconds