TypeSafe ships Jev: typed calibrated decisions, no RLHF and no text generation
ccerrato147 · x · 2026-09-20
Thread conclusion: the fix needs a different objective — not human preference (RLHF), not verified correctness (RLVR), but calibrated decisions: confidence tracks accuracy, so software acts when sure and escalates when not. Diogo's TypeSafe shipped it as Jev on Sep 15: typed decisions, no text generation.
More from coding & agent
- Jev's instant compression scores each tool call to trim agent context without summarization — FinanceYF5 · 2026-09-20
- Stop using LLMs for ticket triage: Jev turns unstructured input into an executable smart if — sven_ai · 2026-09-20
- Jeff Dean's 1-Hour AI Engineering Lecture: From LLM Basics to Agent Graphs — irinarish · 2026-09-20
- Multi-agent systems work best with clear roles, not more agents — _jaydeepkarale · 2026-09-20
- Jev founder: all AI models are built for human-in-the-loop, not true automation — hardimanjames · 2026-09-20
- Karpathy's Stanford AI engineering lecture: from LLM to Graph in five steps — irinarish · 2026-09-20