Why INGRID is allowed to say "I don't know"

Every reasoning module in BlueCrow follows the same rule: separate what was actually measured from what's being inferred, and say so explicitly. A hot GPU under heavy load gets a different narrative than the same temperature at idle — but the severity never gets softened just because there's a plausible explanation. An unexplained reading gets an honest "I can't confirm the cause from temperature and load alone" instead of a guess dressed up as an answer.

That was a deliberate choice, and it costs something. A more confident-sounding assistant is, on the surface, more impressive. It's also more likely to be wrong in a way you can't tell from the outside.

INGRID is built to reason from what BlueCrow can actually observe on your machine, not to fill gaps with plausible-sounding filler. If she doesn't have a temperature reading, she says the sensor is unavailable — she doesn't estimate one. If a driver question needs current information she doesn't have locally, she looks it up instead of answering from a stale assumption.

I'd rather INGRID be right and occasionally uncertain than confident and occasionally wrong.