Why it matters
Agent-driven traffic is exactly the bursty, non-deterministic load this talk addresses; if caching and perf-test assumptions don't hold for agent workloads, your cost and latency numbers are both wrong.
The tokenmaxxing angle
Unpredictable agent calls break the caching layers that keep inference cost down. This session is about building performance tests that catch that failure mode before it becomes a spend surprise in production.
From the organizers
The Test Tribe Meetup lists speaker Gargi Majumder at 307 West 38th Street, Studio 1401, 6-8pm; the description frames the talk around caching failures and latency spikes under AI-driven workloads.