Voices
Guillermo Rauch says Jev is 18x faster at p95 than GPT Luna in Vercel's fx command reviewer and will likely become the default
Rauch posted that fx runs in auto mode with a safety reviewer analyzing every shell command, that the reviewer currently runs on GPT Luna, and that Jev is up to 18x faster at p95 and more accurate in that slot. He said it is coming to Vercel AI Gateway and is the likely new default. This is a concrete production swap rather than a benchmark: the workload is a per-command classifier in the hot path of an agent, which is exactly the shape Jev's decision-only design targets.
↳ Follow the thread