🔍 Read the full analysis: The Potential Of 'System One' AI To Change How We Interact With Technology on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
TypeSafe announced Jev, a new AI model aligned with Kahneman’s ‘System One’ thinking, focusing on decision automation rather than text generation. It aims to transform enterprise automation by providing fast, structured, and probabilistic decisions, challenging traditional large language models.
TypeSafe has introduced Jev, a new AI model that departs from traditional text-generating large language models by focusing on decision-making within software systems. Launched on September 15, 2026, Jev is designed to provide structured, typed responses with associated probabilities, enabling faster and more reliable automation. This development could significantly alter how enterprises implement AI for operational decisions, reducing reliance on human judgment and costly reasoning models.
Jev is described by TypeSafe as a ‘System One’ model, inspired by Daniel Kahneman’s psychology, which emphasizes quick, intuitive decision-making. Unlike conventional language models that produce free-form text, Jev returns structured, typed answers—such as ‘team: billing, confidence: 0.94’—that software can directly act upon, eliminating the need for parsing or interpretation. Built with a focus on automation, Jev handles questions of choices, scores, and yes/no probabilities, making it behave more like a function call than a conversational agent.
The model was developed by Diogo Almeida, a co-inventor of RLHF and InstructGPT, who argues that the reinforcement learning techniques used in large language models introduce issues like mode dropping and overconfidence. Instead, Jev employs a training method called Reinforcement Learning for Calibrated Decisions (RLCD), aimed at producing more reliable, probabilistic outputs suitable for automation. TypeSafe claims Jev can answer in 70 to 500 milliseconds at a cost of approximately $0.042 per million tokens, asserting it is significantly faster and cheaper than traditional models—by factors of hundreds.
While TypeSafe markets Jev as having ‘zero hallucinations,’ this claim is narrow; it refers only to output formatting errors, not decision accuracy. Independent benchmarks show Jev’s accuracy varies, with some tests indicating it performs well on narrow questions but less so on broader, more complex tasks. The company acknowledges that accuracy depends on how judgments are broken into smaller questions and combined, emphasizing that Jev is a tool to assist, not replace, human judgment entirely.
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 for Enterprise Automation
The advent of Jev represents a potential shift in enterprise AI, moving from text generation to decision automation. By providing structured, probabilistic responses that software can directly act on, Jev could significantly reduce costs and increase speed in operational workflows. Its design aims to eliminate errors caused by output formatting or hallucinations common in traditional language models, improving reliability in critical applications. This approach could lead to broader adoption of AI in areas like customer support, fraud detection, and process management, where fast, accurate decisions are essential.
Furthermore, Jev challenges the assumption that large language models are necessary for all AI tasks, proposing instead a specialized class of models optimized for decisions. If successful, this could reshape AI development priorities and deployment strategies across industries, emphasizing structured decision-making over conversational AI.
enterprise decision automation AI tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Evolution of AI Models and Enterprise Needs
Over the past three years, the AI industry has focused heavily on improving large language models (LLMs) with better reasoning, longer context, and more sophisticated text generation. Companies like OpenAI and Anthropic have led the way, but these models often require human oversight due to issues like hallucinations and overconfidence. In enterprise settings, LLMs have been used primarily for customer service, content creation, and coding assistance, but their limitations have spurred interest in specialized AI systems.
TypeSafe’s Jev is part of a broader trend toward developing AI models tailored for specific tasks, particularly decision-making within software systems. Its emphasis on structured, typed responses with associated probabilities aligns with industry needs for automation that is both fast and reliable. The development also builds on the criticism of reinforcement learning techniques like RLHF, which, despite improving conversational abilities, have introduced issues with reliability and overconfidence in AI outputs.
Jev’s launch follows a series of experiments and benchmarks that highlight the potential and limitations of decision-focused AI, with some independent tests showing promising results but also underscoring the importance of careful implementation and testing.
“Reinforcement learning techniques used in traditional LLMs create issues like mode dropping and overconfidence. Jev’s approach offers a more calibrated, reliable decision-making process.”
— Diogo Almeida, co-inventor of RLHF
Limitations and Challenges in Decision Accuracy
While Jev shows promise, its accuracy varies depending on task complexity and input quality. Benchmarks indicate it performs well on narrow, well-defined questions but less reliably on broader or ambiguous ones. The company acknowledges that accuracy depends heavily on how judgments are broken into smaller questions and combined. Independent tests have shown lower accuracy in some scenarios, such as phishing detection, where Jev scored around 62.6%, compared to larger models like Claude Haiku 4.5 at over 81%. These results suggest that Jev’s decision-making is still evolving, and its reliability in critical applications remains to be fully validated.
Additionally, the claim of ‘zero hallucinations’ is limited to output formatting errors; decision correctness still depends on how well the model is trained and how questions are structured. More extensive real-world testing and validation are needed to establish its robustness across diverse enterprise use cases.
Next Steps for Adoption and Validation
TypeSafe plans to expand testing of Jev in various enterprise environments, focusing on automating decision processes in customer support, fraud detection, and workflow management. Further independent benchmarking and real-world case studies will help validate its accuracy and reliability. The company also intends to refine its training methods and improve calibration to enhance decision quality.
Industry observers will watch how Jev performs at scale and whether its structured decision approach can replace or complement existing AI solutions. As adoption grows, integration with existing enterprise systems and validation through real-world metrics will determine its ultimate impact on automation strategies.
In the near term, expect more pilot projects and potential pilot programs that test Jev’s capabilities in high-stakes environments, alongside ongoing research into improving decision calibration and error rates.
Key Questions
How does Jev differ from traditional large language models?
Jev produces structured, typed responses with associated probabilities instead of free-form text, making it more suitable for automation within software systems. It acts more like a function call than a conversational agent.
Can Jev replace human judgment entirely?
No, Jev is designed to assist decision-making by providing reliable, calibrated responses. Human oversight remains important, especially for complex or high-stakes decisions.
What are the main limitations of Jev currently?
Its accuracy varies depending on task complexity and input quality. It performs well on narrow questions but less reliably on broader or ambiguous tasks. Validation in real-world applications is ongoing.
Will Jev be integrated into existing enterprise workflows?
Yes, TypeSafe intends to pilot Jev in various enterprise environments, aiming for integration into automation pipelines, customer support, and decision systems as validation progresses.
What is the significance of Jev’s speed and cost advantages?
Jev’s ability to answer in milliseconds at a fraction of a cent per query could dramatically reduce automation costs and increase throughput, enabling more decisions to be automated at lower expense.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
