How 'System One' AI Could Lead To More Useful AI Systems, According To Jev
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How 'System One' AI Could Lead To More Useful AI Systems, According To Jev on ThorstenMeyerAI.com

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

Get smart everyday buys 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 new AI model built on ‘System One’ principles, focusing on decision-making rather than text generation. It aims to improve automation speed, cost, and reliability by producing typed decisions with probabilities. The development challenges common assumptions about large language models in enterprise AI.

TypeSafe has unveiled Jev, a new AI model that departs from conventional text-generating large language models by producing structured, typed decisions with associated probabilities. This development, announced on September 15, 2026, aims to enhance enterprise automation by offering faster, more reliable decision-making tools that do not rely on natural language output. The innovation is significant because it challenges the prevailing assumption that all AI problems require language models, potentially redefining how AI is integrated into business workflows.

Jev is built on the concept of ‘System One’ thinking, inspired by psychologist Daniel Kahneman’s distinction between fast, intuitive judgments and slow, deliberate reasoning. Unlike traditional large language models (LLMs) that generate free-form text, Jev responds to structured questions with typed answers, including choices, scores, and yes/no probabilities, each with calibrated confidence levels. This approach eliminates issues like hallucinations and output formatting errors that can disrupt automated pipelines.

Developed by Diogo Almeida, a co-inventor of RLHF and InstructGPT, Jev is designed to address flaws associated with reinforcement learning from human feedback, such as overconfidence and mode dropping. The model is optimized for decision-making in software, providing responses in milliseconds at a fraction of the cost of typical LLMs—claimed to be 193.6 times faster and 444.6 times cheaper based on specific workflows. The model’s decision-centric design allows businesses to automate tasks previously reliant on human judgment or expensive reasoning models.

TypeSafe emphasizes that Jev’s outputs are schema-compliant and free from hallucinations, but notes that it does not guarantee decision correctness—only that it produces well-structured, reliable data. Initial benchmarks show promising results, with some tests indicating accuracy around 67-70%, though independent evaluations suggest more cautious expectations. The company acknowledges that accuracy depends heavily on how questions are structured and the quality of input data, emphasizing that Jev is a tool to assist, not replace, human judgment entirely.

At a glance
breakingWhen: announced September 15, 2026
The developmentTypeSafe launched Jev on September 15, 2026, a ‘System One’ AI model designed for decision automation, marking a significant shift from traditional language models.

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 for Enterprise AI Automation

The introduction of Jev represents a potential paradigm shift in enterprise AI, moving away from language-based systems toward decision-focused models. By producing structured, probabilistic answers, Jev could enable faster, more reliable automation of routine decisions, reducing costs and increasing throughput. This approach addresses longstanding issues with hallucinations and output variability in traditional LLMs, potentially improving trust and integration in critical workflows. If widely adopted, Jev could influence how AI models are designed for decision support, especially in sectors like customer service, IT operations, and compliance.

Amazon

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 Model Evolution

Over the past three years, major AI launches have centered around improving reasoning, context length, and code generation in large language models like GPT and Claude. These models excel at generating natural language but often struggle with reliability, hallucinations, and cost-efficiency, leading many companies to seek alternative approaches. In enterprise settings, the need for fast, accurate, and structured decision-making has grown, prompting research into models that produce actionable data rather than text.

TypeSafe’s Jev builds on this trend by leveraging a new training approach called Reinforcement Learning for Calibrated Decisions (RLCD), developed by Almeida, which aims to produce decisions with known confidence levels. This contrasts with the prevalent use of RLHF, which often results in overconfidence and mode collapse. The model’s launch signifies a shift toward specialized AI tools optimized for automation rather than broad language understanding.

Prior efforts have shown that while LLMs can perform well in natural language tasks, their reliability diminishes in decision-critical applications. Jev’s structured output format and focus on speed and cost-efficiency aim to fill this gap, offering a more predictable and scalable solution for enterprise use cases.

“Jev is designed to produce decisions, not words. This makes it more suitable for automation where reliability and speed are paramount.”

— Diogo Almeida, CEO of TypeSafe

Amazon

structured decision-making AI tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Uncertainties About Accuracy and Real-World Performance

While initial benchmarks suggest Jev performs well within controlled tests, independent evaluations reveal variability in accuracy, with some tests showing a success rate around 63-70%. The model’s reliance on question structuring and the quality of input data means its real-world effectiveness may vary across applications. Additionally, the claim of ‘zero hallucinations’ pertains only to output formatting, not decision correctness, leaving open questions about its reliability in critical tasks.

It is also unclear how Jev will scale across diverse industries or handle complex, ambiguous decisions that require nuanced judgment. TypeSafe acknowledges these limitations and emphasizes that Jev is a decision-support tool, not an infallible oracle.

Amazon

decision automation tools for business

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Adoption and Validation

Following its launch, the next phase involves broader testing across different enterprise environments to validate Jev’s performance and reliability. Industry partners and early adopters are expected to evaluate its effectiveness in automating routine decisions, such as support ticket routing, compliance checks, and operational alerts. TypeSafe plans to release updated versions incorporating user feedback and further refining the model’s calibration and accuracy.

Research and independent benchmarks will continue to assess Jev’s capabilities, with a focus on understanding its limitations and optimizing question design. The company also intends to explore integrations with existing automation tools and expand its decision types to cover more complex scenarios.

Overall, the focus will be on demonstrating Jev’s value as a reliable, cost-effective decision engine that complements human judgment rather than replacing it entirely.

Amazon

AI decision engine software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How does Jev differ from traditional language models?

Jev produces structured, typed decisions with probabilities, rather than free-form text, making it more suitable for automation and reducing errors related to output formatting.

Can Jev replace human decision-making?

Jev is designed as a decision-support tool that automates routine judgments. It is not intended to replace human judgment in complex or nuanced scenarios.

What are the main limitations of Jev?

Current limitations include variable accuracy depending on question design, reliance on input quality, and the fact that it does not guarantee decision correctness—only structured, calibrated outputs.

How might Jev impact enterprise automation?

Jev could significantly reduce costs and increase speed for routine decision-making tasks, enabling more scalable and reliable automation workflows across industries.

What is the future development plan for Jev?

TypeSafe plans to expand Jev’s decision types, improve calibration, validate performance through broader testing, and integrate with existing automation systems.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Two Channels: How the Pentagon Just Split Frontier-AI Procurement in Half

The Pentagon split its AI procurement into two distinct channels, placing Anthropic in a strategic, non-redundant lane while excluding it from the classified network. This segmentation impacts vendor relationships and national security strategies.

The bank account in the chat. How personal finance became an agentic on-ramp.

OpenAI introduces live account data integration in ChatGPT for Pro users, marking a move toward agentic consumer finance and reshaping fintech interactions.

Kill-Switch-Proof: How To Build So Washington Can’t Take Your AI Stack Down

A guide on how organizations can architect AI systems resistant to government shutdowns, emphasizing dependency mapping, abstraction layers, and open-weight models.

Stenvrik: News as Geography

Thorsten Meyer AI detailed Stenvrik, a closed-beta news product mapping about 1,700 live stories to 49 city hubs.