Agents look at the benchmark scores and immediately adopt the platform with the higher score. There's no switching cost or learning curve involved anymore. No conference talk they half-remember or some Stack Overflow answer they vaguely trust either. Which means agents don't have any particular habits or biases. Earlier, new platforms needed years to penetrate; now they just need better evals.