Summary

Phil Hetzel from BrainTrust discusses the emerging field of agentic development and challenges traditional assumptions about which technical discipline should own this space. The presentation focuses on the complexities of generative AI implementation, highlighting the gap between creating proofs of concept and successfully bringing AI agents to production. BrainTrust's platform addresses this challenge through two key pillars: evals (experimental tweaking) and observability, which help ensure agent reliability and performance in real-world usage scenarios.