TypeSafe's Jev model pitches 'machine-native intelligence' with RLCD-calibrated decisions
samcharrington · x · 2026-10-07
On the TWIML podcast, TypeSafe co-founder and CEO Diogo discusses Jev, the company's newly released model for embedding fast, reliable intelligence directly into software. Key points:
- He argues text-generation-optimized models are poorly suited to many decisions required for real-world automation, coining "machine-native intelligence"
- Explains how Jev differs from traditional classifiers and LLM-based approaches
- Introduces reinforcement learning from calibrated decisions (RLCD), with calibration and reliability as core to AI as a software primitive
- Discusses the relationship between models and code
Retweeter chrismarino called the questioning far deeper than typical fawning podcast interviews, though he remains skeptical on economics and adoption and less convinced it's "just a classifier."
More from Models
- Dev Complains Overnight Long-Running Tasks Keep Hitting Usage Limits — willdepue · 2026-10-07
- Gary Marcus on whether frontier LLMs can solve open math problems without symbolic harnesses — GaryMarcus · 2026-10-07
- From botching 9.9 vs 9.11 to tackling the hardest math problems in two years — Yuchenj_UW · 2026-10-07
- OpenAI claims 372 unsolved problems cracked, averaging about 3 hours each — i_dg23 · 2026-10-07
- Benchmark: OpenAI Decisions API costs 2x more, 5-10% worse than Jev — xeophon · 2026-10-07
- Cagliostro V3.5 135M Dethrones SmolLM2-135M on Open SLM Leaderboard at 27.49 — Megneous · 2026-10-07