Context rot

Definition

A term used to describe deteriorating AI performance as its working context becomes crowded with outdated, duplicated, conflicting or irrelevant information.

Put simply

Giving AI more information doesn't necessarily make it better.

What boards and leaders should understand

Larger context windows and persistent workflows create a temptation to keep adding information.

But if useful instructions or facts become buried among low-quality context, performance may deteriorate.

So when an AI system starts producing poorer results, the model itself may not be the problem. The information being accumulated around it may be.

Questions to consider

  • Where are long-running AI workflows accumulating information faster than we are managing its quality and relevance?