Why it matters
Flyte's workflow tracing and rerun features let builders see exactly which steps in an agent loop consume tokens, making inefficient loops visible instead of hidden in logs.
The tokenmaxxing angle
Building the loop's LLM-call and tool-step components inside Flyte surfaces per-step execution and retry costs, practical groundwork for tracking where agent spend actually goes.
From the organizers
Organized by Sage Elliott for AI Builders and Learners - Seattle in Bellevue, with starter code and tool integrations provided and projects presented at the night's end.