🔍 Read the full analysis: The Practical Advantages Of 'System One' AI In Future Technologies on ThorstenMeyerAI.com
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
TL;DR
TypeSafe’s ‘System One’ AI, exemplified by Jev, delivers structured decisions with high speed and low cost, challenging traditional language models for enterprise automation. Its real-world advantages are promising but require further validation.
On September 15, 2026, TypeSafe AI unveiled Jev, a new class of ‘System One’ AI model designed specifically for decision-making within software systems. Unlike traditional large language models (LLMs) that generate free-form text, Jev produces structured, typed decisions with associated probabilities, enabling faster, more reliable automation. This development marks a significant shift in enterprise AI, focusing on decision accuracy and efficiency rather than language generation, and it could reshape automation strategies across industries.
Jev is built on the concept of ‘System One’ thinking, inspired by Daniel Kahneman’s psychological model, which emphasizes quick, intuitive judgments. Developed by Diogo Almeida, co-inventor of RLHF and InstructGPT, Jev is optimized for automation tasks, providing structured outputs such as choices, scores, and yes/no probabilities within milliseconds. It answers structured questions directly, allowing software to act without parsing prose, thus reducing errors related to output formatting and interpretation.
TypeSafe claims Jev operates at speeds between 70 and 500 milliseconds and costs approximately $0.042 per million input tokens, claiming it is nearly 194 times faster and 445 times cheaper than comparable models like GPT-6 Astra and Claude Fable 5.1. The model’s design aims to replace costly reasoning processes inside enterprise workflows, where decisions are often simple but frequent, such as classifying tickets or routing tasks. Its approach challenges the assumption that all AI tasks require large, general-purpose language models, suggesting that decision-focused models can deliver superior performance for specific automation needs.
While Jev is marketed as having ‘zero hallucinations,’ this claim is narrowly defined: it cannot produce off-schema answers or malformed data structures. However, it does not inherently guarantee the correctness of decisions—only that outputs conform to predefined schemas. Independent tests show that Jev’s accuracy varies depending on question complexity and benchmarking methods, with some tests indicating lower performance compared to traditional models, especially on nuanced or ambiguous tasks. The model’s overconfidence in some cases and underconfidence in others highlight the need for careful deployment and validation.
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 Decision-Making
The introduction of Jev and similar ‘System One’ models signifies a potential paradigm shift in enterprise AI. By focusing on structured decisions, these models can drastically reduce latency and costs associated with automation, enabling a broader range of tasks to be automated at a fraction of previous expenses. This could lead to increased efficiency, lower operational costs, and more scalable AI integration across industries such as customer support, finance, and logistics.
Moreover, the shift away from language generation toward decision-oriented outputs addresses many reliability concerns associated with traditional LLMs, such as hallucinations and inconsistent reasoning. For organizations, this means more predictable and controllable AI behavior, reducing the need for human oversight in routine decision-making processes. However, the reliance on calibrated probabilities and structured outputs also introduces new challenges in ensuring accuracy and managing overconfidence, which must be addressed through rigorous testing and validation.
Overall, these models could accelerate the adoption of AI in mission-critical applications, fostering a new era of automation where speed, cost, and reliability are balanced with decision quality. Still, the technology’s maturity and real-world performance are evolving, and widespread deployment will depend on further validation and refinement.
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 Towards Decision-Focused Automation
Over the past three years, AI development has largely centered around improving large language models (LLMs) like GPT and Claude, emphasizing better reasoning, longer contexts, and more natural language generation. These models, however, often face issues like hallucinations, overconfidence, and high computational costs, limiting their reliability for enterprise automation. In response, industry innovators like TypeSafe are exploring alternatives that prioritize structured, decision-oriented outputs.
The concept of ‘System One’ AI draws from psychological models of quick, intuitive judgment and aims to embed this approach into software decision-making. The launch of Jev marks a notable milestone, as it is designed explicitly for automation tasks, producing typed decisions with associated confidence scores, rather than generating prose. The funding of $40 million from DCVC and the involvement of AI pioneers like Almeida underscore the strategic importance of this shift.
While traditional models remain dominant, early benchmarks and pilot implementations suggest that decision-focused models could complement or even replace certain classes of enterprise AI tasks, especially those requiring rapid, repeated judgments. The ongoing challenge is to validate the accuracy and reliability of these models in complex, real-world scenarios.
“Jev is designed to produce decisions, not words, making automation faster, cheaper, and more reliable.”
— Diogo Almeida, co-inventor of RLHF
structured decision-making software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Uncertainties in Model Accuracy and Reliability
While Jev shows promising speed and cost advantages, its accuracy varies depending on task complexity and benchmarking methods. Independent tests reveal lower performance on nuanced tasks, such as email phishing detection, compared to traditional models like Claude Haiku 4.5. The overconfidence observed in some outputs indicates that, despite schema compliance, decisions may still be incorrect, raising concerns about reliability in critical applications. The true effectiveness of Jev in diverse, real-world scenarios remains to be fully validated, and further testing is needed to establish its robustness across different domains and use cases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Validation of ‘System One’ AI
TypeSafe plans to continue refining Jev through ongoing testing and real-world deployments, aiming to improve accuracy and confidence calibration. Industry adoption will likely involve pilot projects in enterprise workflows such as customer support, finance, and logistics, where decision speed and cost are critical. Meanwhile, independent researchers and organizations will scrutinize its performance across varied tasks to assess reliability. The broader AI community will watch whether ‘System One’ models can complement or replace traditional LLMs in enterprise settings, potentially leading to a new class of decision-oriented AI tools.
As an affiliate, we earn on qualifying purchases.
Key Questions
How does ‘System One’ AI differ from traditional language models?
‘System One’ AI produces structured, typed decisions with probabilities, rather than generating free-form text, enabling faster and more reliable automation.
What are the main advantages of Jev for enterprise use?
Jev offers rapid decision-making (70-500 ms), low cost ($0.042 per million tokens), and schema-compliant outputs that reduce formatting errors, making automation more scalable and predictable.
Are there any limitations or risks associated with Jev?
Yes, while schema compliance is guaranteed, decision accuracy can vary, and overconfidence or underconfidence in probabilities requires careful validation before deployment in critical systems.
Will ‘System One’ models replace traditional LLMs entirely?
Not necessarily; they are likely to complement existing models, especially in tasks requiring fast, structured decisions, but widespread replacement depends on further validation and performance improvements.
What industries are most likely to benefit from ‘System One’ AI?
Industries with high-volume, routine decision-making needs, such as customer support, finance, logistics, and IT automation, are prime candidates for adopting decision-focused AI models like Jev.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
