{"id":118934,"date":"2026-09-20T01:56:10","date_gmt":"2026-09-20T01:56:10","guid":{"rendered":"https:\/\/youzum.net\/typesafe-ai-releases-jev-a-system-one-model-that-returns-typed-calibrated-decisions-instead-of-text\/"},"modified":"2026-09-20T01:56:10","modified_gmt":"2026-09-20T01:56:10","slug":"typesafe-ai-releases-jev-a-system-one-model-that-returns-typed-calibrated-decisions-instead-of-text","status":"publish","type":"post","link":"https:\/\/youzum.net\/es\/typesafe-ai-releases-jev-a-system-one-model-that-returns-typed-calibrated-decisions-instead-of-text\/","title":{"rendered":"TypeSafe AI Releases Jev: A System One Model That Returns Typed, Calibrated Decisions Instead of Text"},"content":{"rendered":"

The ChatGPT moment in 2022 taught AI to talk to people. One of its builders now bets the next moment is AI that talks to software, not people. TypeSafe AI<\/a> released Jev<\/a>. Jev is transformer-based, but it is not a large language model. It does not generate text. You send a state and typed questions. It returns typed decisions with probabilities that code can branch on.<\/p>\n

Is it deployable?<\/strong> Yes, as a hosted API in early access behind a waitlist. TypeSafe has not published weights, a parameter count, or a self-hosting option.<\/p>\n

What is a System One Model?<\/strong><\/h2>\n

The name borrows from Daniel Kahneman\u2019s split between fast intuition and slow reasoning. TypeSafe team argues RLHF tuned models for human preference. That produced chat, and overconfidence and mode dropping. Those flaws keep a human in the loop.<\/p>\n

Jev uses a new stack: a new architecture, a parallel sampler, and Reinforcement Learning for Calibrated Decisions (RLCD). TypeSafe has not disclosed the architecture. <\/p>\n

How the Jev API Works<\/strong><\/h2>\n

One endpoint handles everything: POST https:\/\/api.typesafe.ai\/v1\/systemone<\/code>. The body carries state<\/code>, model<\/code>, and a map of questions<\/code>. The docs<\/a> define 3 question types.<\/p>\n

\n\n\n\n\n\n\n\n
Primitive<\/th>\nAsks<\/th>\nReturns<\/th>\n<\/tr>\n<\/thead>\n
Choice<\/td>\nPick 1 option from a list<\/td>\nchoice<\/code>, probabilities<\/code>, confidence<\/code><\/td>\n<\/tr>\n
Score<\/td>\nRate against ordered levels<\/td>\nscore<\/code>, probabilities<\/code>, confidence<\/code><\/td>\n<\/tr>\n
Noul<\/td>\nIs this statement true?<\/td>\nnoul<\/code>, a probability from 0 to 1<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n

Questions run in parallel and in isolation against the same state. TypeSafe says adding questions barely changes response time. A Choice supports up to 255 options.<\/p>\n

\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
from typesafe_sdk import Choice, Noul, TypeSafeClient\n\nclient = TypeSafeClient()  # reads TYPESAFE_API_KEY\nr = client.system_one(\n    state=ticket,\n    questions={\n        \"department\": Choice(\n            instructions=\"Which team should handle this\",\n            criteria={\"billing\": \"Payment issues\", \"technical\": \"Bugs\"},\n        ),\n        \"is_urgent\": Noul(instructions=\"The message conveys urgency\"),\n    },\n)\nprint(r.answers[\"department\"].choice, r.answers[\"is_urgent\"].noul)<\/code><\/pre>\n<\/div>\n<\/div>\n
Install with pip install typesafe-sdk<\/code> (Python 3.10 or later). A JavaScript SDK ships as @typesafe-ai\/sdk<\/code>. The quickstart<\/a> also covers cURL and an agent skill for Claude Code.<\/p>\n
Confidence is the Product<\/strong><\/h2>\n
Every Choice and Score answer carries a confidence<\/a> value from 0 to 1. TypeSafe derives it from the shape of the probability distribution. In the docs example, billing<\/code> wins at 0.84. Confidence is only 0.596, because technical<\/code> still holds 0.159.<\/p>\n
The docs suggest 3 paths. Act on high confidence. Review the middle. Send low confidence to a human. Thresholds should scale with the cost of a wrong action.<\/p>\n
Pricing, Speed, and the Benchmark Fine Print<\/strong><\/h2>\n
Jev costs $42 per billion input tokens. TypeSafe quotes existing LLMs at $0.20 to $10 per 1M input tokens. In its recorded demo, Jev finished in 0.114s for $0.000081. GPT-5.6 Terra took 8.566s for $0.013880.<\/p>\n
The TypeSafe team claims it to be 193.6x faster and 444.6x cheaper. Those figures come from TypeSafe\u2019s own workflow evals<\/a>. But hold on here are some things to keep in mind:<\/strong><\/p>\n