🔍 Read the full analysis: Why 'System One' AI Might Be The Most Practical Step Forward, According To Jev on ThorstenMeyerAI.com
Get hardware and tech essentials 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
TL;DR
TypeSafe announced Jev, a decision-oriented AI model based on ‘System One’ thinking, designed for fast, structured decisions in enterprise software. It challenges the dominance of large language models for automation tasks, emphasizing efficiency and reliability.
On September 15, 2026, TypeSafe AI announced Jev, a new class of decision-focused AI built for automation inside enterprise software. Unlike traditional large language models (LLMs), Jev produces structured, typed decisions with associated probabilities, rather than free-form text. This development marks a significant shift in AI design philosophy, emphasizing speed, reliability, and cost-efficiency for decision-making tasks.
Jev is based on the concept of ‘System One’ thinking, inspired by psychologist Daniel Kahneman, which refers to quick, intuitive judgments. The model is designed to handle structured questions—choices, scores, and yes/no probabilities—returning results that are directly actionable by software. It does not generate text; instead, it provides typed, schema-conformant decisions, reducing errors caused by output formatting or hallucinations common in traditional LLMs.
TypeSafe claims Jev can answer questions in 70 to 500 milliseconds at a cost of approximately $0.042 per million tokens, making it significantly faster and cheaper than existing models. The company argues that most enterprise decisions are System One judgments—such as classifying a ticket or assessing urgency—and that these can be automated more efficiently with Jev’s approach. The model is built using a new training method called Reinforcement Learning for Calibrated Decisions (RLCD), which aims to improve decision accuracy and confidence calibration, addressing issues like overconfidence and mode dropping associated with RLHF (Reinforcement Learning with Human Feedback).
Jev’s initial performance benchmarks show promising results in controlled tests, though accuracy varies depending on question complexity and input quality. Independent testing indicates a lower accuracy rate than some frontier models, highlighting that effective deployment requires careful question decomposition and calibration. Nonetheless, the model’s ability to produce structured, reliable decisions at high speed and low cost positions it as a practical alternative to traditional LLMs for many enterprise tasks.
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 and AI Design
The introduction of Jev represents a shift in enterprise AI strategy, emphasizing decision accuracy, speed, and cost-efficiency over free-form language generation. Its structured output reduces the need for human parsing and post-processing, potentially enabling more reliable automation of routine judgments. For organizations, this could mean faster workflows, lower operational costs, and fewer errors caused by hallucinated or malformed outputs. Moreover, Jev’s approach challenges the assumption that every problem must be solved with large language models, opening new avenues for AI deployment in decision-critical applications.
By focusing on typed decisions with calibrated probabilities, Jev aligns AI more closely with software functions, reducing the reliance on human oversight. This could accelerate AI adoption in areas like customer support, compliance, and operational management, where structured, fast, and reliable decisions are essential. However, the model’s accuracy depends heavily on how well questions are formulated and how decision boundaries are set, indicating that successful implementation requires careful design and ongoing calibration.
As an affiliate, we earn on qualifying purchases.
Background on AI Models and Enterprise Use Cases
Over the past three years, the AI industry has been dominated by large language models (LLMs) like GPT and Claude, which generate human-like text and have been widely adopted for chatbots, content creation, and reasoning tasks. These models, while versatile, suffer from issues such as hallucinations, overconfidence, and high costs, which limit their reliability for critical enterprise applications. To address these limitations, AI researchers and companies have explored alternatives, including smaller, more specialized models, and different training techniques like RLHF.
Jev’s development by TypeSafe builds on a growing recognition that many enterprise decisions are routine, structured, and do not require free-form language output. Instead, models that produce typed, schema-conformant decisions can be integrated directly into software workflows, reducing parsing errors and increasing trustworthiness. The model’s emphasis on decision calibration and speed reflects a broader industry shift toward AI that complements existing software systems rather than replacing human judgment entirely.
Since its funding of $40 million led by DCVC and the involvement of Diogo Almeida—co-inventor of RLHF—TypeSafe positions Jev as a practical step forward in enterprise AI, challenging the narrative that larger, more complex language models are always the best solution. Its focus on automation, speed, and structured output aims to fill a niche where traditional LLMs are too slow, expensive, or unreliable.
“Jev is designed to produce decisions, not words. We believe most enterprise judgments are System One—fast, intuitive, and decision-based—and should be handled as such.”
— Diogo Almeida, CEO of TypeSafe
As an affiliate, we earn on qualifying purchases.
Unconfirmed Aspects and Performance Limitations
While TypeSafe reports promising benchmarks, independent tests have shown lower accuracy rates, such as 62.6% on a phishing email classification task, compared to 81.3% for some frontier models. The true effectiveness of Jev in diverse real-world applications remains to be fully demonstrated. Additionally, the model’s reliance on question decomposition and calibration means that improper question design could limit its reliability. The long-term robustness and scalability of Jev’s approach are still under evaluation, and broader industry adoption will depend on further validation and real-world testing.
As an affiliate, we earn on qualifying purchases.
Next Steps for Jev and Enterprise Adoption
TypeSafe plans to expand testing and gather user feedback to refine Jev’s calibration and accuracy. The company is likely to release more case studies demonstrating its application across different industries, emphasizing how to best formulate questions and interpret outputs. Additionally, as competitors develop similar decision-focused models, Jev’s success will depend on its ability to demonstrate consistent reliability and cost savings at scale. Industry observers will be watching for broader deployments and independent validation studies over the coming months.
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, rather than free-form text, making it more suitable for automation and reducing errors caused by output formatting or hallucinations.
What are the main advantages of Jev for enterprise use?
Jev offers faster response times (70-500 ms), lower costs ($0.042 per million tokens), and more reliable, schema-conformant decisions that can be directly integrated into software workflows.
What are the current limitations of Jev?
Its accuracy varies depending on question design, and independent tests have shown lower performance than some frontier models. Proper calibration and question formulation are critical for effective deployment.
Will Jev replace large language models entirely?
Not necessarily; Jev targets specific decision-making tasks within enterprise software, complementing rather than replacing the broader language capabilities of models like GPT and Claude.
What is the future outlook for decision-focused AI like Jev?
Further validation, real-world testing, and industry adoption will determine its success. TypeSafe aims to refine Jev and expand its use cases, potentially transforming enterprise automation.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
