Matt Turck podcast unpacks Jev: TypeSafe's machine-native AI and its RLHF critique
mattturck · x · 2026-10-16
VC Matt Turck sits down with TypeSafe's CEO one month after the Jev launch to unpack what Jev is: 'machine-native AI' built for software, optimized for intelligence per dollar rather than chat. Key threads: RLHF trained models to please humans (sycophancy, refusals), RLVR benchmark gains don't equal reliable automation, and TypeSafe's RLCD approach optimizes for machine-native intelligence using in-house training data on open-weight models. The CEO embraces the 'it's just a classifier' framing, argues fast-followers copy the interface but miss the intelligence, is skeptical of recursive self-improvement, and invokes Jevons paradox to argue practical automation can avert an AI winter.
Related event: Matt Turck Interviews TypeSafe CEO on Jev One Month After Launch(2 posts)→
More from Companies & People
- AI billboards take to the LA skies as startup hires plane to tow ad banner — TarlonKhoubyari · 2026-10-16
- SUNY Albany to offer Topological Deep Learning course, from Perceptrons to WL test — burny_tech · 2026-10-16
- Palantir opens third year of merit-only Meritocracy Fellowship applications — eliano · 2026-10-16
- Why hiring 'technical people who can create content' is so hard, from a Cerebras dev — charles_irl · 2026-10-16
- Sentry to host System Update panel on agentic coding with OpenAI, Linear, DoorDash leaders — zeeg · 2026-10-16
- Anthropic gives METR employee-level access, waives right to block its risk assessments — TechCrunch AI · 2026-10-16