AI Slop

AI Slop is low-value AI-generated output that creates more review, maintenance, and cognitive work than useful progress.

The problem is not that AI produces imperfect work. Engineering has always involved review, revision, and mistakes. The problem begins when generated output looks finished enough to enter a workflow, but adds little signal, weakens judgment, or creates work for someone else to clean up.

The Handbook

• Code Slop — Generated code that increases surface area without improving the system.
• Documentation Slop — Polished-looking prose that obscures decisions, repeats the obvious, or creates false certainty.
• Test Slop — Tests that create coverage theater, assert trivial behavior, or make real failures harder to see.
• Process Slop — AI-generated process, reviews, and planning work that add organizational friction rather than clarity.
• AI Tadka — A specific pattern in which unnecessary AI “improvement” makes working engineering worse.

The Standard

Useful AI output reduces the cost of understanding or changing a system. Slop does the opposite: it transfers cost into review, maintenance, CI time, or human attention.

The goal is not to reject AI assistance. It is to preserve the judgment required to decide whether an addition is worth keeping.