🔍 Read the full analysis: Jev And The Promise Of 'System One' AI In Practical Applications on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
TypeSafe has released Jev, an AI model based on ‘System One’ thinking, designed for decision-making in software. It offers faster, cheaper, and more reliable automation, challenging traditional language models. Its effectiveness and accuracy are still being evaluated.
On September 15, 2026, TypeSafe announced the launch of Jev, a new AI model designed for decision-making within software, marking a significant shift from traditional language models that generate free-text responses. This development introduces a class of AI called ‘System One’ models, optimized for automation, with potential implications for enterprise AI and software automation.
Jev is built to produce structured, typed decisions with associated probabilities, rather than free-form text. It responds to structured questions with specific answers, such as choices, scores, or yes/no probabilities, enabling direct integration into automated workflows. The model is based on a new training approach called Reinforcement Learning for Calibrated Decisions (RLCD), developed by Diogo Almeida, co-inventor of RLHF techniques used in ChatGPT. Jev answers in milliseconds, at a claimed cost of approximately $0.042 per million tokens, and is marketed as significantly faster and cheaper than traditional large language models (LLMs).
TypeSafe argues that most internal enterprise decisions are System 1 judgments—quick, intuitive responses—making Jev more suitable than text-generating models, which often require human oversight due to issues like overconfidence and hallucinations. The company claims Jev has ‘zero hallucinations’ in the narrow sense of schema compliance, though accuracy remains an area of ongoing evaluation.
Jev vs. LLMs: who should make the call?
Jev, from TypeSafe AI, is a “System One” model. It doesn’t write text. It returns a typed decision with a confidence score that your software can act on directly.
Same support ticket, two kinds of answer
“This ticket appears most likely related to billing, although it could also concern account settings or a recent plan change. I would suggest reviewing the invoice history before…”
A person reads it, or code has to parse the prose.
team: "billing"Software reads it and acts. Nothing to parse.
How they differ
| LLM | Jev | |
|---|---|---|
| Output | Text written for people | A choice, a score or a yes/no probability |
| Speed | Seconds per call | 70–500 ms* |
| Price | Input and (pricier) output tokens | $0.042 per million input tokens, output free* |
| Knows when it’s unsure | Often sounds confident when wrong | Confidence score on every answer |
| Explains its answer | Yes | No, which matters for audits |
| Best at | Reasoning, writing, open questions | Routing, tagging, scoring, duplicate checks |
* Vendor-reported. TypeSafe also claims up to 194× faster and 445× cheaper on its own selected workflows.
Accuracy is something you build
Jev is far cheaper and faster, but not more accurate than frontier models. How you phrase the question matters a lot.
TypeSafe’s benchmark scores agreement with two frontier models rather than verified ground truth. The five-question result used weights fitted on 1,000 labelled examples.
The real idea: a confidence dial you control
“duplicate listing”, confidence 0.62
Raise the threshold for fewer mistakes and more manual review. Lower it for more automation and more risk.
Only use Jev when all four hold
Good fits
- Routing tens of thousands of support tickets a day
- Flagging duplicate listings in a product catalogue
- Replacing a keyword filter that mis-tags half its matches
Poor fits
- Drafting customer emails or release notes
- Reviewing a few high-stakes contracts a month
- Anything that needs a written explanation
Implications of Structured Decision AI for Enterprise Automation
The introduction of Jev signals a potential paradigm shift in enterprise AI, moving from generative text models to decision-focused systems optimized for automation. By providing fast, reliable, and cost-effective decision-making, Jev could significantly reduce operational costs and improve the reliability of automated workflows. This approach challenges the dominance of traditional LLMs in enterprise settings, suggesting that many internal decisions can be better handled by specialized, schema-conformant models.
However, the effectiveness of Jev depends on the accuracy of its decision-making, which is still being validated through independent testing. Its ability to integrate seamlessly into existing systems and handle complex, nuanced judgments remains to be seen, but the approach offers promising avenues for more robust and efficient AI-driven automation.
enterprise decision automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on AI Decision-Making and Industry Shift
Over the past three years, major AI model launches have focused on improving reasoning, context length, and code generation, primarily through large language models like GPT and Claude. These models excel at generating human-like text but face challenges in reliability, overconfidence, and hallucinations, often requiring human oversight for critical decisions.
TypeSafe’s Jev represents a departure from this trend, emphasizing structured, schema-based responses that are more aligned with software decision-making needs. The company’s approach is rooted in psychological theory—Daniel Kahneman’s System 1 and System 2 thinking—and aims to replace slow, deliberative reasoning with rapid, calibrated decisions suitable for automation.
This shift reflects a broader industry debate about the most effective ways to embed AI into operational workflows, especially in enterprise contexts where reliability and speed are paramount.
“Jev replaces free-text generation with structured decisions, making AI more reliable for automation.”
— Diogo Almeida, co-inventor of RLHF
Current Validation and Accuracy of Jev
Independent testing indicates that Jev’s accuracy varies depending on the task. In a phishing email classification test, Jev scored 62.6%, compared to Claude Haiku 4.5’s 81.3%. On a more granular question set, Jev achieved 95% accuracy, but its probabilities were sometimes overconfident or underconfident. The benchmark results rely on agreement with frontier models rather than direct correctness, and the company acknowledges that accuracy depends on task design and prompt engineering.
Therefore, while Jev’s speed and cost benefits are clear, its reliability in complex or high-stakes decisions remains an open question, requiring further validation across diverse use cases.
Next Steps for Adoption and Validation
TypeSafe plans to expand testing of Jev across different enterprise workflows to better understand its accuracy and reliability. The company will also refine its training methods and calibration techniques to improve decision confidence. Industry observers expect early adopters to pilot Jev in areas like customer support, fraud detection, and operational automation, where speed and cost are critical.
Further independent evaluations and real-world deployments will determine whether Jev can fulfill its promise of replacing traditional LLMs in decision-critical applications, potentially reshaping enterprise AI strategies in the coming months.
Key Questions
How does Jev differ from traditional language models?
Jev produces structured, typed decisions with associated probabilities, rather than free-text responses, making it more suitable for automation and decision-making within software systems.
What are the main advantages of Jev?
Jev offers faster response times (70 to 500 milliseconds), significantly lower costs (around $0.042 per million tokens), and improved reliability for decision-making, with claims of ‘zero hallucinations’ in schema compliance.
What are the current limitations of Jev?
Its accuracy varies by task, and independent tests suggest it still makes errors, especially on complex or nuanced questions. Validation in diverse real-world scenarios is ongoing.
Will Jev replace large language models in enterprise AI?
While promising, Jev is unlikely to replace all LLMs but may serve as a specialized tool for decision automation where speed, cost, and reliability are critical.
When will Jev be widely available?
TypeSafe has begun pilot programs and plans further testing in the coming months, with broader availability depending on validation outcomes.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
