🔍 Read the full analysis: The Next Phase Of AI Development: Emphasizing 'System One' Over Sentence Generation on ThorstenMeyerAI.com
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
TL;DR
TypeSafe has launched Jev, a new AI model focused on decision-making rather than text generation. It leverages ‘System One’ thinking, promising faster, more cost-effective, and more reliable automation for enterprise tasks. The development signals a shift from traditional language models toward structured decision models.
TypeSafe has introduced Jev, a new AI model designed for decision automation that departs from traditional text-generating language models. This development, announced on September 15, 2026, represents a strategic shift toward structured, decision-oriented AI, aiming to improve speed, reliability, and cost-efficiency in enterprise workflows. The model’s focus on ‘System One’ decision-making could redefine how businesses automate routine judgments.
Jev is the first public release of TypeSafe’s ‘System One’ models, which are built to produce typed decisions with associated probabilities, rather than generating free-form text. Unlike large language models (LLMs) that produce prose, Jev accepts structured questions and returns typed responses such as choices, scores, or yes/no probabilities, enabling direct software action without parsing text.
The model is optimized for automation, delivering answers in 70 to 500 milliseconds at a cost of approximately $0.042 per million input tokens, claims TypeSafe. This performance is claimed to be hundreds of times faster and cheaper than traditional LLM workflows, making decision automation more accessible and economical for enterprise use. Jev is built using a new training approach called Reinforcement Learning for Calibrated Decisions (RLCD), which aims to address issues like overconfidence and mode dropping common in RLHF-based models.
TypeSafe emphasizes that Jev’s outputs are schema-compliant, eliminating errors from output formatting, but it does not guarantee correctness of the decision itself. Its accuracy has been benchmarked against frontier models, with results indicating that it performs well in narrow tasks but still requires careful question design. Independent testing shows Jev’s accuracy varies depending on question complexity and context, with some overconfidence in its probabilities.
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 AI Automation
The launch of Jev signals a potential paradigm shift in enterprise AI, moving away from text-based chatbots toward structured decision models that can be integrated directly into software workflows. This approach promises faster response times, lower costs, and more reliable automation, reducing the need for human oversight in routine judgments. If widely adopted, it could significantly impact industries that rely heavily on decision-making processes, such as customer support, finance, and logistics.
By focusing on ‘System One’ thinking—quick, intuitive judgments—TypeSafe aims to address limitations inherent in traditional LLMs, such as hallucinations, overconfidence, and formatting errors. This could lead to more dependable automation solutions, especially in high-stakes or high-volume environments where accuracy and speed are critical. However, the approach’s success depends on refining accuracy and understanding its limitations, as current benchmarks show room for improvement.
enterprise decision automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Shift from Text Generation to Structured Decisions
Over the past three years, major AI models have centered on improving reasoning, context length, and code generation, with companies betting on large language models (LLMs) to handle a wide array of tasks. However, these models have faced criticism for issues like hallucinations, overconfidence, and reliance on human oversight. In response, TypeSafe’s Jev introduces a different approach, emphasizing decisions over language, inspired by Daniel Kahneman’s ‘System One’ thinking—fast, intuitive judgments.
Jev’s development builds on the expertise of Diogo Almeida, co-inventor of RLHF and InstructGPT, who now advocates for alternative training methods better suited for automation. The model’s debut follows a wave of skepticism about the suitability of LLMs for enterprise decision-making, especially in scenarios demanding high reliability and speed. The company’s funding of $40 million from DCVC underscores investor confidence in this strategic shift.
While traditional models excel at generating human-like text, Jev’s structured approach aims to produce actionable, typed decisions directly consumable by software, reducing the need for parsing and interpretation. This shift reflects a broader industry debate about the future role of language models versus decision-focused AI systems.
“Our goal is to replace unreliable text generation with structured, calibrated decisions that software can act on directly.”
— Diogo Almeida, CTO of TypeSafe
AI decision-making tools for business
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unanswered Questions About Model Reliability
While Jev’s decision accuracy has shown promising results in narrow benchmarks, its performance in complex, real-world scenarios remains uncertain. The model’s overconfidence in some cases and underconfidence in others suggest that further refinement is necessary. Additionally, it is not yet clear how well Jev integrates with existing enterprise systems or how it handles ambiguous or conflicting inputs. Ongoing testing and real-world deployment will be critical to assess its reliability and scalability.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Validation
TypeSafe plans to expand testing of Jev across various enterprise environments, gathering data on its accuracy and robustness in real-world tasks. The company intends to refine its training methods and improve model calibration. Industry observers will watch for broader adoption, integration with existing workflows, and independent validation of its performance. Additionally, competitors may develop similar decision-focused models, intensifying the shift away from traditional LLMs.
automated enterprise workflow tools
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, instead of generating free-form text. This makes it more suitable for direct automation within software systems, reducing parsing and interpretation errors.
What are the main advantages of Jev for enterprise use?
Jev offers faster response times—answering in milliseconds—at a significantly lower cost, with more reliable decision-making that can be directly integrated into workflows without human intervention.
What are the current limitations of Jev?
Its accuracy varies depending on task complexity, and it can still be overconfident or underconfident in its probabilities. It also does not guarantee decision correctness, only schema compliance.
Will Jev replace existing AI models in the near future?
It is too early to tell. While Jev shows promise for specific decision tasks, widespread replacement of traditional LLMs will depend on further validation, integration success, and improvements in accuracy.
How does Jev handle ambiguous or conflicting inputs?
Current information suggests Jev responds with calibrated probabilities, but handling ambiguity remains an area for further development and testing.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
