hypothesis-evaluating

name	hypothesis-evaluating
description	Use when comparing expected outcomes against actual results, judging what a gap means, or evaluating whether a change achieved its intent. The review agent crystallizes this.
user-invocable	false

Hypothesis Evaluating

Comparing what was expected with what actually happened, where the divergences between prediction and reality are the most valuable findings.

---

This is a cognitive stance — it describes how the agent thinks, not how it collaborates. Where hypothesis-forming makes predictions specific enough to be wrong, hypothesis-evaluating reads the gap between those predictions and what reality produced, and judges what the gap means.

The evaluation isn't about finding errors. It's about calibration: understanding what the divergences reveal about the assumptions, the process, and the result. Where prediction matched reality, that's confirmation — worth noting but rarely the interesting part. Where prediction and reality diverge, something was learned.

This stance is explicitly evaluative, not generative. FOUNDATIONS describes the distinction: generation and evaluation activate different regions of the probability landscape. Trying to do both at once — diagnosing a problem while drafting the fix — collapses evaluation into generation and weakens both. Diagnose first. Judge readiness second. Gesture at fix directions only when the diagnosis makes them obvious.

The relational structure

A scientist reading experimental results against the hypothesis that motivated the experiment. The experiment was designed to test the hypothesis. The confirming results build confidence in the theory. The disconfirming results are more interesting — they mean the theory needs revision, and the specific way the results diverge tells you which part of the theory was wrong.

A forensic accountant reconciling projected cash flows with the actual ledger. The projection was based on assumptions about revenue, costs, and timing. The ledger shows what actually happened. Where they match, the assumptions held. Where they diverge, the accountant investigates: was the assumption wrong, was the data wrong, or did something happen that nobody anticipated?

A coach reviewing game film against the game plan they drew up before the match. The plan assumed certain matchups, certain defensive formations, certain opportunities. The film shows what actually happened. The coach isn't looking for mistakes — they're calibrating: where did the plan's model of the opponent hold up, and where did the opponent do something the plan didn't account for? The divergences become the foundation for the next game plan.

When this stance applies

• Reviewing execution against preparation (the review agent crystallizes this stance)
• Any post-action evaluation where you have both a prediction and an outcome
• Calibrating confidence in a plan, model, or assumption after seeing evidence

What this stance is not

• Fault-finding or criticism (which looks for what's wrong against a standard of correctness)
• Generation (which produces new content — this stance evaluates existing content)
• Auditing (which checks compliance — this stance reads for understanding)

Composition notes

Naturally paired with hypothesis-forming — one forms the prediction, the other evaluates it against results. The quality of the evaluation depends on the quality of the hypothesis: vague predictions produce vague evaluations, specific predictions produce specific findings.

Composes with collaborative grounding when the evaluation surfaces ambiguities that need the user's situated knowledge. Some divergences between prediction and result can only be judged by someone who knows what was intended.