liveBook · Manning

Vigilance decrement

As explained in Vibe Engineering

Vigilance decrement is the decline in sustained attention that occurs during prolonged monitoring.

Vigilance decrement is the erosion of sustained attention when oversight becomes prolonged, passive, or reserved for rare failures. In AI-assisted code review, it appears as scanning generated diffs on autopilot, trusting green tests, and overlooking the invariant violation that matters. section 1.4

The late production failure

A developer accepts an AI-generated change after a quick review because tests pass. The code fetches related records one at a time, working with fifty staging rows but timing out for a customer with forty thousand. An on-call engineer then spends hours reconstructing intent before fixing a problem that could have been caught earlier. section 1.4

Counter the decrement by keeping human judgment engaged at the verification gate:

  • Write executable specifications, tests, or contracts before generating implementation. section 1.8
  • Replace dump-and-review with a Specify, Implement, Verify, Refactor, and Own loop. section 1.8
  • Treat passing mechanical checks as insufficient for semantic, architectural, security, performance, and business-meaning verification. section 1.8

How Vibe Engineering covers it

How deep it goes
Explains the failure mechanism and production consequence, then connects it to verification practices without developing psychological theory.
Examples
Uses autopilot review, green tests, late handoffs, and an N-plus-one-style production timeout as concrete failure cases.
What it assumes
Assumes familiarity with AI-generated code, code review, tests, production incidents, and the distinction between mechanical and semantic correctness.
What you can do after
Diagnose/debug with it by identifying passive oversight and installing specifications, verification gates, and explicit ownership before merge.
How it connects
Connects vigilance decrement to automation complacency, the out-of-the-loop problem, automation bias, trust debt, and code review.

This explainer was generated by Manning's knowledge graph from the text of Vibe Engineering. The section links point to the parts of the book it draws on.