Why it declined my question

AI Windes answers questions it can ground in the model catalogue. Everything else it declines — it will not write your code, draft your email, or answer general knowledge questions.

Why

There is no language model behind the consultant, so it has nothing to improvise with. A confident-sounding answer built on no data is worse than a refusal, and a tool that answers everything badly is less useful than one that answers a narrow thing well.

What passes

• Anything naming a model, a maker, or model vocabulary — tokens, context window, benchmarks, inference, open weights.
• Selection questions with a superlative and a constraint, even without naming a model: “cheapest with vision”.
• Short follow-ups in an existing conversation: “and cheaper?”, “what about vision?”.

What doesn't

“How do I write Python code to reverse a list” mentions coding, but it isn't a model-selection question — it's a request for code. Rephrased as “which model is best at Python code generation”, it's answerable.

If a fair question was declined

Add a model name, a maker or a constraint and it will almost always route correctly. If something plainly in scope keeps being turned away, send us the exact wording — that's a bug in the scope rule and we want it.