Facing ambiguity

What you want an AI to do when a request is under-specified.

This is a two-poled spectrum — neither end is better. It describes how you'd like an AI to work with you, never who you are. Here are the two poles:

Make the call, note it

When a task is ambiguous, you'd rather I make a reasonable call and note it than stop to ask.

You likely lean this way if: you tell me to make the call and just note what I assumed.

Ask first

When a task is ambiguous, you'd rather I stop and ask than guess at what you meant.

You likely lean this way if: you ask me to check with you before deciding on something ambiguous.

Where they meet: Set a threshold: small calls get made-and-noted, big ones get a quick check-in first.

How Hunchful asks about it

When something's ambiguous, should I make a reasonable call and flag it, or check with you first?

Why opposites can help each other

A decider keeps momentum; an asker prevents wrong turns on the calls that matter — together you move fast without driving off a cliff.

Put it in a model your AI can read

In a Cognitive Model Protocol model, your position on this axis becomes a revisable hunch — with a confidence, an uncertainty, and a falsifier — that any AI can read and adapt to. Nothing is a fixed label.

Make your model →

Other axes

When things get messy — Synthesize first vs. One step at a time
Answer format — Tight prose vs. Scannable lists
Surfacing uncertainty — Flag it loudly vs. Commit, minimal caveats
Getting started — React to a draft vs. Plan before producing