Why it matters
The panel's own framing, model costs evolving and infra choices becoming strategic, is essentially the tokenmaxxing thesis, discussed by a cloud provider, a model lab, and a VC in the same room.
The tokenmaxxing angle
Description explicitly asks 'how do you approach evals and reliability' and 'what does real portability look like' as models and costs evolve — a direct discussion of multi-model routing and switching costs.
From the organizers
Hosted by Madrona with AWS and OpenAI on the panel; agenda runs panel discussion (1-1:45pm) then startup spotlights (1:45-2:15pm), with evals, reliability, and model portability named as core topics.