Online evaluation

Judging a model's real quality, latency, and cost on live production traffic rather than static offline benchmarks, so that swapping the model behind an agent is decided by observed deltas.

grounded in: Trend themes 'Agent cost/perf migration: Teams are quantifying real speed and cost deltas from swapping the model behind a production agent' and 'LLM hype backlash: separating genuine usefulness from