Foundry demosOptimization / Evaluation / Model upgrades
Use agent optimization to adapt an agent to a new model
Follow a competitive-intelligence agent through traces, a private benchmark, optimization candidates, and an experiment comparing versions.
In this demo
Jeff uses a competitive-intelligence agent to show why model upgrades can change behavior and why application-specific evaluation matters. The walkthrough examines execution traces, a private benchmark, optimizer candidates, and proposed repository changes. It finishes with an experiment comparing agent versions and discusses how to use evidence rather than assumptions when selecting an update.