# Spending computation where approximation fails

ID: adaptive-computational-resolution
Canonical: https://rezonansapp.com/nodes/adaptive-computational-resolution
Version: 1
SHA256: bcee4ad5d162eed572b44ffc35ee27698e85a9fcfd2b9ea5c1c2aad5b0895cb7
Epistemic kind: reconstruction

## Question

Can a cheap local approximation identify where expensive computation is actually needed?

## Formulation

Use regularity-sensitive approximation to allocate fine computation to difficult regions of a fractal visualization.

## Tension

A cheap indicator must detect its own failure well enough to justify skipping expensive work.

## Challenge

How are missed difficult regions counted alongside runtime savings?