Skip to content
JevDirectory.org
OfficialSites & GuidesVerified 2026-09-20

AI Primer: RLCD and Calibration

Why TypeSafe trains decision models with RLCD instead of RLHF: calibrated probabilities where 0.2 outcomes happen about 20% of the time, and the case for machine-to-machine automation.

Open docs.typesafe.ai
Category
Sites & Guides
Published by
TypeSafe AI
Author
—
Added
2026-09-20
Tagsofficialdocsevaluationconfidencearchitecture

Highlights

  • RLCD, reinforcement learning for calibrated decisions, is a third post-training path alongside RLHF and RLVR.
  • Calibration means outcomes given 0.2 probability occur around 20% of the time across many predictions.
  • RLHF rewards can produce sycophancy and hallucinations; TypeSafe cofounder Diogo Almeida co-invented RLHF.
  • TypeSafe expects future AI automation to be roughly 99% machine-to-machine and 1% human-facing.

Watch out

Calibration is a property of groups of predictions, not a promise about one answer, so keep a review path for the individual cases that matter.

More like this

System One
The idea behind Jev: models that return typed decisions and probabilities instead of generated text, named after Kahneman's fast-thinking System 1, with confidence to route around uncertainty.
Sites & Guides#official#docs#architecture
Official
API Reference
The raw HTTP contract behind every SDK: POST a state plus typed noul, choice, and score questions to /v1/systemone, and get one answer per question, with error codes and retry guidance.
Sites & Guides#official#docs#api
Official
Models
The model card for jev-1.13.0: $42 per billion input tokens with free output, a 64k context, 250k tokens per second, and how to list the models your account can call.
Sites & Guides#official#docs#models
Official
Back to all resources

From the community

Posts from builders shipping with Jev right now.

Vercel's fx safety reviewer, 18x faster

Guillermo Rauch
Guillermo Rauch
Vercel
@rauchg
·

We're seeing extraordinary results from @typesafeai. Default mode in 𝚏𝚡 is auto, with a safety reviewer analyzing every command. That reviewer runs on GPT Luna today. Jev is up to 18x faster (p95) *and* more accurate. It's coming to @vercel AI Gateway and likely new default.

Pranit
Pranit
Vercel
@fazxes

We benchmarked fx auto mode (safety) classifier with @typesafeai's Jev. tl;dr: ~5-18x faster and more accurate than 𝚐𝚙𝚝-𝟻.𝟼-𝚕𝚞𝚗𝚊, our current top choice

Image
Reply
Read 138 replies

Jev lands on OpenRouter