When the platform offers an order, you have seconds. Assistant is being tested to detect the visible facts, show what it has actually read, and progressively add a recommendation that remains under the courier's control.
A 4 km order can be different for two couriers because their speed, local knowledge and current commitments differ. The Assistant goal is to make those differences visible from the courier's own history — but an uncertain reading must remain visibly uncertain.
Evaluating offers by eye: looks close, seems to pay well, it's peak time — I'll take it. A month later you can't remember why you earned less.
The floater should first show the exact facts it read. Recommendations are added only when the required facts and confidence are sufficient.
Working across platforms can create more options, but it also creates conflicts, safety risks and timing pressure. Any decision support must respect active commitments and the courier's own judgement.
The long-term goal is to compare detected offer facts in one courier-controlled view. A map line alone is not a bicycle route and will not be presented as one.
The Assistant is being designed to flag possible conflicts between active work and a new offer. The courier retains the final accept or decline decision.
Tools for finding and analysing a courier's imported conversation history are planned. They are not yet a public product promise.
Note: when working across multiple platforms, know their rules on simultaneous work — platforms apply their own sanctions for violations.