AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Jev Discusses The Shift Toward 'System One' AI And Its Potential Impact on ThorstenMeyerAI.com

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get the latest gadgets delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

TypeSafe announced Jev, a decision-oriented AI model inspired by Kahneman’s ‘System One,’ designed for rapid, structured decisions in software automation. The development signals a shift from traditional language models toward decision-specific AI, with potential cost and speed advantages.

On September 15, 2026, TypeSafe announced the release of Jev, a novel AI model that shifts focus from text generation to structured decision-making. This development challenges the prevailing assumption that large language models (LLMs) are necessary for all enterprise AI tasks, emphasizing speed, cost-efficiency, and decision reliability. The move is significant because it signals a potential paradigm shift in how AI is integrated into software automation, especially for routine judgments.

Jev is described by TypeSafe as a System One Model, inspired by Daniel Kahneman’s concept of fast, intuitive thinking. Unlike traditional LLMs that generate free-form text, Jev produces typed answers with associated probabilities, such as indicating a support ticket is likely a billing issue with 94% confidence. It handles questions of three types: choices, scores, and yes/no probabilities, and returns results in under 500 milliseconds at a fraction of a cent per input token. The model is built to eliminate errors caused by output formatting or hallucinations, focusing purely on decision accuracy.

Jev’s training method, called Reinforcement Learning for Calibrated Decisions (RLCD), aims to address issues associated with RLHF, such as mode dropping and overconfidence. Its design reflects a belief that most internal decisions within enterprise software are System One judgments—quick, automatic assessments—rather than slow, deliberate reasoning. This approach contrasts sharply with the dominant trend of deploying large language models for conversational AI and text generation tasks.

TypeSafe claims Jev is significantly faster and cheaper than existing models, answering in 70 to 500 milliseconds and costing roughly $0.042 per million tokens, with no output charges. The company asserts that most internal enterprise decisions are System One judgments, and automating these could drastically reduce costs and increase speed, expanding the scope of what can be automated.

While TypeSafe markets Jev as having ‘zero hallucinations,’ this claim is narrowly defined. Jev cannot produce off-schema answers or malformed data structures, but it can still select an incorrect answer within its allowed options. Its accuracy depends heavily on how well it is trained and the specific workflow context, with independent tests showing a range of results from 62.6% to 95% accuracy depending on question complexity and test conditions.

At a glance
breakingWhen: announced September 15, 2026
The developmentTypeSafe unveiled Jev on September 15, 2026, a new AI model that produces typed decisions instead of text, aiming to transform enterprise automation by focusing on decision-making speed and reliability.

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

A typical LLM

“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.

Jev
team: "billing"
confidence 0.94threshold 0.80: auto-route

Software reads it and acts. Nothing to parse.

How they differ

LLMJev
OutputText written for peopleA choice, a score or a yes/no probability
SpeedSeconds per call70–500 ms*
PriceInput and (pricier) output tokens$0.042 per million input tokens, output free*
Knows when it’s unsureOften sounds confident when wrongConfidence score on every answer
Explains its answerYesNo, which matters for audits
Best atReasoning, writing, open questionsRouting, 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 own workflow benchmark
Jev (ties Claude Sonnet 5)
67.8%
Independent test: 2,000 phishing emails
Jev, asked one question
62.6%
Claude Haiku 4.5
81.3%
Jev, split into five narrow questions
95.0%

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

Jev decides
“duplicate listing”, confidence 0.62
Above: act automatically. Tag, route, merge. Most of the volume ends here.
Below: escalate the unsure few to an LLM or a person.

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

High volumeThousands of small judgments, not a handful of big ones.
Narrow questionRelevance, category or duplicate checks. No reasoning needed.
Cheap errorsA wrong answer costs little, or unsure cases go to something smarter.
Heuristic failureA keyword rule is visibly getting it wrong.
All four true: Jev is a strong candidate
Any one false: use an LLM, or keep your rule

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 ‘System One’ AI for Enterprise Automation

The introduction of Jev marks a potential turning point in enterprise AI by shifting focus from language-based interactions to structured decision-making. This could lead to significant cost reductions, faster processing times, and more reliable automation of routine judgments. If widely adopted, it may reduce reliance on human oversight for simple but critical decisions, reshaping workflows across industries. However, the model’s accuracy and applicability in complex scenarios remain to be fully validated, making this an important development to watch.

Amazon

enterprise decision automation AI tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on AI Shifts Toward Decision-Focused Models

Over the past three years, major AI model launches have emphasized improved reasoning, longer context windows, and enhanced code generation, primarily through large language models like GPT and Claude. These models have become the default for conversational AI and complex text tasks. However, industry experts have raised concerns about the limitations of LLMs, especially their tendency toward hallucinations, overconfidence, and the need for human oversight. In response, some researchers and companies have explored alternative approaches focused on decision-making rather than text generation.

TypeSafe’s Jev builds on this trend, drawing inspiration from cognitive psychology and Kahneman’s ‘System One’ concept, which describes fast, automatic thinking. The company argues that most internal enterprise decisions are ‘System One’ judgments and that AI should be optimized to support these quick, routine assessments rather than emulate human-like conversation or reasoning. The model’s launch with $40 million in funding and the involvement of AI pioneer Diogo Almeida underscores the seriousness of this shift.

Prior efforts to improve AI reliability have centered on refining LLMs, but Jev’s decision-centric approach represents a fundamental rethinking—prioritizing structured, calibrated outputs over free-form text.

“Jev is designed to produce decisions, not words. It’s more like calling a function in your code than chatting with a colleague.”

— Diogo Almeida, CEO of TypeSafe

Amazon

decision-focused AI software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Uncertainties in Jev’s Performance and Adoption

While Jev shows promising speed and cost advantages, its real-world accuracy and reliability in complex or high-stakes scenarios are still being evaluated. Independent tests have produced mixed results, with some questions answered with over 95% confidence and others significantly less so. Its ability to generalize across diverse enterprise workflows remains unproven, and questions about long-term robustness and integration into existing systems are still open. Furthermore, the company’s claims about ‘zero hallucinations’ are based on narrow technical definitions, and broader reliability issues have yet to be fully addressed.

Amazon

structured decision-making AI

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Jev and System One AI Development

TypeSafe plans to release more detailed case studies and conduct broader independent testing to validate Jev’s performance across different industries and decision types. The company is also working on refining training techniques to improve accuracy and calibration. Adoption will depend on how well Jev integrates with existing automation platforms and whether its decision accuracy meets enterprise standards. Industry observers will be watching for real-world deployments, potential limitations, and the evolution of decision-focused AI models in the coming months.

Amazon

fast decision AI models

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How does Jev differ from traditional large language models?

Jev produces structured, typed decisions with probabilities instead of free-form text, making it more suitable for automation and decision-making tasks within software systems.

What are the main advantages of Jev?

It offers faster response times (70-500 ms), lower costs (around $0.042 per million tokens), and avoids issues like hallucinations and formatting errors common in language models.

Can Jev handle complex, high-stakes decisions?

Its performance in complex scenarios is still under evaluation. Early tests show promising results but also highlight variability in accuracy depending on question type and context.

Will Jev replace large language models entirely?

Not necessarily; Jev is designed for specific decision-making tasks, especially routine judgments, complementing rather than replacing conversational and reasoning-focused models.

What are the main limitations of Jev?

While it eliminates formatting errors, it can still select incorrect answers within its schema, and its effectiveness depends heavily on training quality and workflow design.

Source: ThorstenMeyerAI.com

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

The Menu: What Ten Answers Reveal

A detailed examination of how ten jurisdictions respond to automation and AI, revealing patterns and political choices shaping the future of work and income.

Transform How You Learn: 5 Revolutionary Search Methods In AI

Google announced five new AI-powered search tools for education, including interactive visuals, practice quizzes, and Lens-based homework help, rolling out globally.

The Memento Constraint: Why Continual Learning Is the Trillion-Dollar Bottleneck Nobody Is Pricing

Exploring the ‘Memento’ challenge in AI, its impact on continual learning, and why solving it could reshape the trillion-dollar enterprise AI economy.

Minerva. The opposite path.

Italy’s Minerva project trained from scratch on 2.5 trillion tokens, yet scored only 4.9% on Italian academic tests, raising questions about native-language investment.