Enterprise AI Bottleneck Is Evals, Not Models

Databricks' chief AI scientist Jonathan Frankle argues that models are already good enough for enterprise tasks and the real bottleneck is evals. Others add that domain experts who know what "good" looks like, armed with the best models, are the true value engine.

2026-10-24 ~ 2026-10-24 · 2 related posts

• Databricks' Frankle: enterprise AI's bottleneck is evals, not models — curious_vii · 2026-10-24
• Domain experts, not evals, know what good looks like: arming them with the best models is the real value engine — curious_vii · 2026-10-24