When you use an AI tool made by a model maker, it will always use that maker's model. That's fine when their model is best. It's a problem the rest of the time — which is often, because no one leads at everything.
The structural conflict
A model maker can't route you to a competitor; their business is selling their own model's usage. So even a great first-party app is structurally limited to one model's strengths and blind spots.
What neutrality buys you
A neutral tool has no model to sell, so it's free to pick the genuine best for each task — and to switch as the leaderboard changes. You get the best available result rather than the best result from one vendor.
Why it compounds
As models specialise further, the gap between "best single vendor" and "best of all vendors" widens. Neutrality isn't a feature you'll outgrow; it gets more valuable as the field fragments.
That's the bet behind Ensemble: stay neutral, route to the best, assemble the result.