In Ballet's evaluation agent test on 1,998 complex support messages, Jev cost $0.095 and 110 seconds vs Claude Sonnet 5's $4.34 and 220 seconds. Sonnet was about 1 point more accurate, but Jev was 46x cheaper.
Same 1,998 complex support messages. Same three questions. @typesafeai Jev: $0.095 and 110 seconds. @AnthropicAI Claude Sonnet 5: $4.34 and 220 seconds. Sonnet was about 1 point more accurate. Jev was 46× cheaper. That decision is what Ballet's evaluation agent does. It
The author shares a Jev usage insight: it seemed tricky at first, but clicked when framed as 'not letting the AI write text.' Using inquiry handling as an example, the conventional approach dumps everything on GPT or Claude to analyze, while Jev takes a different approach.
The prototype uses AI to estimate hesitation or interest from cursor back-and-forth movements or lingering before clicking, observing customer demeanor and supplementing explanations or suggesting comparisons.
The author built an e-commerce real-time customer service demo using Jev and gpt-live-1, which instantly suggests recommended products based on the user's speech during a conversation, delivering a great experience.
