DSPy on the Jev hype: prompting and hand-written judgment code are both too extreme

lateinteraction · x · 2026-09-21

Luis Serrano (lateinteraction) retweets the DSPy team: it's exciting that everyone is learning "programming, rather than prompting, AI models" via Jev, but the deeper point is missed—using a frontier agent is like managing a person, and hand-crafting Jev code is like hand-writing assembly for a niche RISC machine. Both have their use, but both are generally too extreme for engineering good systems.

Related event: DSPy Team: Hand-written Prompts and Fully Managed Agents Are Both Premature Optimization(2 posts)→

Original post →

More from coding & agent

coding & agent channel →