Why We Made Jev — Diogo Almeida, TypeSafe Co-founder & CEO

Name: Why We Made Jev — Diogo Almeida, TypeSafe Co-founder & CEO
Uploaded: 2026-09-22T00:00:00.000Z
Duration: 8542 s
Channel: Latent Space
Description: Days after Jev's launch, TypeSafe CEO Diogo Almeida explains System One Models: machine-native AI for software, not chat. He covers RLHF's collapse into instruction-following, calibration, rejecting public benchmarks, and why reliability and intelligence per dollar decide what AI automates.

LSLatent Space@LatentSpacePod

Full transcript

English

Summary：Days after Jev's launch, TypeSafe CEO Diogo Almeida explains System One Models: machine-native AI for software, not chat. He covers RLHF's collapse into instruction-following, calibration, rejecting public benchmarks, and why reliability and intelligence per dollar decide what AI automates.

Watch on YouTube

Core points (3)

1RLHF shifted models toward instruction-following and broke calibration for machine use.
2TypeSafe treats public benchmarks as gameable and optimizes for intelligence per dollar.
3Reliable small decisions, decomposed workflows, and versioned APIs beat chat wrappers.