Every Outage Is a Broken Mental Model

Incidents are usually described as failures of software.

Often they are failures of understanding.

Every outage exposes a mismatch between how someone believed the system worked and how it actually behaved.

The System Did What It Could Do

Production systems rarely violate physics. They follow dependency paths, failure modes, capacity limits, and configuration states.

The surprise comes from our model being incomplete.

A dependency was hidden. A retry amplified load. A cache masked an assumption. A region failed differently than expected.

Postmortems Should Update Models

A good postmortem does more than identify root cause.

It asks which mental model failed. What did we believe? Why did that belief seem reasonable? What evidence contradicted it? Where else might the same assumption exist?

AI as Model Checker

AI can help compare documented architecture, telemetry, incidents, and change history to surface contradictions.

But it can only help if the organization preserves enough reasoning for comparison.

The Goal

The point of incident response is not merely restoring service.

It is improving the organization's understanding so the same surprise becomes less likely next time.

Nexplane is open source. If this resonated, star the repo — it helps others find it.
⭐ Star on GitHub