🔍 Read the full analysis: How Jev's Insights Shape The Future Of 'System One' AI Technology on ThorstenMeyerAI.com
Get movie nights 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 that produces structured, typed decisions instead of text. This approach challenges traditional language models in enterprise automation. The development could significantly impact how companies automate decisions, emphasizing speed and reliability.
TypeSafe has introduced Jev, a novel AI model that produces structured, typed decisions rather than free-form text, signaling a significant shift in enterprise AI development. This model aims to streamline decision-making processes within software systems, emphasizing speed, cost-efficiency, and reliability. The launch of Jev, backed by $40 million in funding, marks a departure from traditional large language models (LLMs) that generate text, instead focusing on decision automation that can be directly acted upon by software.
Jev is built on the concept of System One Models, inspired by Daniel Kahneman’s psychological framework distinguishing fast, intuitive thinking from slow, deliberate reasoning. Unlike conventional LLMs such as GPT or Claude, Jev handles structured questions—choices, scores, and yes/no probabilities—and returns typed answers with calibrated confidence levels. This allows software to act directly on decisions without parsing lengthy text responses, reducing errors caused by output formatting or hallucinations.
Developed by Diogo Almeida, a co-inventor of RLHF at OpenAI, Jev employs a training method called Reinforcement Learning for Calibrated Decisions (RLCD), designed to improve decision accuracy and reliability. TypeSafe claims Jev can answer in as little as 70 milliseconds at a cost of approximately $0.042 per million input tokens, making it significantly faster and cheaper than traditional LLM workflows. The model is marketed as producing decisions, not words, positioning it more like a software function than a conversational agent.
TypeSafe’s initial benchmarks show Jev achieving around 67.8% agreement with leading models in internal tests, with independent tests indicating lower accuracy on complex tasks like email phishing detection, scoring about 62.6%. The company acknowledges that accuracy depends heavily on how questions are structured and that the model’s reliability will improve with targeted fine-tuning. Importantly, Jev’s claim of ‘zero hallucinations’ refers narrowly to output formatting errors, not decision correctness, which remains subject to development 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 AI Automation
The introduction of Jev signals a potential paradigm shift in how enterprises approach AI-driven decision-making. By producing structured, typed outputs with confidence levels, Jev reduces reliance on human interpretation and parsing of text, enabling faster, more reliable automation of routine decisions. This could expand the scope of automation in customer support, fraud detection, and operational management, where quick, accurate decisions are critical. Additionally, Jev’s lower cost and faster response times may make decision automation more economically feasible at scale, challenging the dominance of traditional reasoning models in enterprise settings.
Furthermore, Jev’s approach questions the assumption that large language models are necessary for all AI tasks. Instead, it advocates for specialized models optimized for decision-making within software, which could lead to a reevaluation of AI architecture strategies across industries. If proven effective, Jev and similar models might replace or augment existing systems, emphasizing structured decision outputs over conversational text, and potentially reducing issues like overconfidence and hallucinations associated with LLMs.
enterprise decision automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on AI Evolution and Decision Models
Over the past three years, the AI industry has been dominated by large language models promising improved reasoning, longer context handling, and better code generation. Companies like OpenAI and Anthropic have led the charge with models such as GPT-4 and Claude, which generate human-like text for diverse applications. However, these models have faced criticism over issues like hallucinations, overconfidence, and the need for human oversight, especially in enterprise automation.
In response, some researchers and companies are exploring alternative AI architectures focused on decision-making rather than text generation. Daniel Kahneman’s psychological framework of System 1 and System 2 thinking has influenced this shift, with System 1 representing quick, intuitive judgments suitable for automation. TypeSafe’s Jev embodies this approach by providing structured, decision-oriented outputs, aiming to reduce errors and improve reliability in automated workflows.
Jev’s development builds on Almeida’s previous work with RLHF, but the company argues that reinforcement learning from human feedback is less effective for automation tasks. Instead, RLCD offers more calibrated, probabilistic decision outputs, aligning with the needs of enterprise software automation. The model’s launch reflects a broader industry trend toward specialized, task-specific AI models designed for operational efficiency rather than general conversational capabilities.
“Jev shifts the focus from generating words to producing decisions that software can act on directly, making AI more practical for enterprise automation.”
— Diogo Almeida, TypeSafe
structured decision-making AI tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Accuracy and Reliability of Jev’s Decisions
While Jev demonstrates promising speed and cost advantages, its accuracy benchmarks indicate room for improvement. Internal tests show a 67.8% agreement rate with leading models, and independent tests report about 62.6% accuracy on complex tasks like email phishing detection. The company admits that accuracy is highly dependent on question design and fine-tuning, and that the model’s decision-making reliability is still evolving. It is not yet clear how Jev will perform across diverse real-world applications or whether it can consistently replace human judgment in critical processes.
Additionally, the claim of ‘zero hallucinations’ pertains only to output formatting errors and does not address potential decision errors, which remain a concern. The true measure of Jev’s effectiveness will depend on ongoing validation, user feedback, and improvements in calibration and training methods.
As an affiliate, we earn on qualifying purchases.
Next Steps for Jev and Enterprise Adoption
TypeSafe is expected to continue refining Jev through targeted fine-tuning and expanded benchmarking. The company plans to release more detailed performance metrics and case studies demonstrating Jev’s application in various enterprise workflows. Meanwhile, early adopters are likely to pilot Jev in controlled environments to evaluate its decision accuracy and integration capabilities.
Industry observers will watch for independent evaluations and real-world deployments to assess whether Jev can meet the reliability standards required for critical decision-making. The broader AI community may also explore hybrid approaches combining Jev’s decision models with traditional reasoning systems to optimize performance and trustworthiness.
Overall, the next phase will involve validating Jev’s practical benefits and addressing remaining challenges related to accuracy, calibration, and scope of application.
As an affiliate, we earn on qualifying purchases.
Key Questions
How does Jev differ from traditional language models?
Jev produces structured, typed decision outputs with confidence levels, rather than free-form text. This allows software to act directly on decisions without parsing language, reducing errors and increasing speed.
What are the main advantages of Jev for enterprise automation?
Jev offers faster response times (as low as 70 milliseconds), lower costs ($0.042 per million tokens), and structured decisions that can be directly integrated into automated workflows, reducing reliance on human interpretation.
What are the current limitations of Jev?
Its accuracy, while promising, still varies depending on question design and fine-tuning. Independent tests show lower performance on complex tasks, and decision correctness remains an area for ongoing improvement.
Will Jev replace human judgment entirely?
Not immediately. Jev aims to automate routine, well-defined decisions, but critical or nuanced judgments will likely still require human oversight until reliability is further validated.
What industries are most likely to benefit from Jev?
Industries with high-volume, repetitive decision-making processes such as customer support, fraud detection, supply chain management, and operational automation are prime candidates for early adoption.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
