Noumena splits errors into two categories. What are they, and how is each meant to be handled?

Internal system/infrastructure errors are caught and handled behind the scenes; user errors (invalid actions) are surfaced clearly to the right audience.

Error handling is one of the most overlooked parts of software, and Noumena models it in two dimensions:

1. Internal system and infrastructure errors — crashes, resource limits, and the like. These are typically not exposed to end users; they are caught and handled internally, backed by full observability (metrics, alerting, diagnostics) so operators can resolve and monitor them.
2. User errors — a user performs an unexpected or invalid action. These should reach the user, clearly. The Runtime's require conditions are the main safeguard here: a failed assertion becomes a custom error returned as a REST response for the frontend or integration service to present.

The design principle is matching the error to its audience: hide plumbing failures, expose actionable user mistakes.