# When the plan changes, does an adaptive heuristic choose better work?

Reasoning Under Change: Evaluation, Python. Repository created Oct 2026.
Code: https://github.com/jlov7/reasoning-under-change-release
Page: https://jasonlovell.ai/projects/reasoning-under-change

Exact work selection under budgets and revisions; the adaptive heuristic did worse, 43 to 59 of 64.

## Design choice

Solve the small problem exactly on every revision, and test an adaptive heuristic against a simple reactive baseline rather than assume it helps.

## In practice

For a small queue, exact selection is cheap. Measure an adaptive layer before trusting it.

## What I built

- Exact selection over a prerequisite graph within named budgets
- Replanning that keeps completed work
- Counterfactuals: what forcing a task in would displace
- Standard library only; the negative result is kept

## Where it stops

At most 20 tasks, four named budgets, capacity 0 to 128 and a one-second planning deadline. The simulation is authored. That the adaptive approach works better is not established, and this is not general AI reasoning.

---
By Jason Lovell. Independent work, separate from my role at PwC.