🔍 Read the full analysis: The New Era Of AI: Moving Beyond Sentence Generation With 'System One' on ThorstenMeyerAI.com
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
TypeSafe announced Jev, a decision-focused AI model that outputs typed answers with probabilities, moving beyond conventional text generation. This development aims to improve automation speed, reliability, and cost-efficiency in enterprise AI applications.
TypeSafe has launched Jev, a groundbreaking AI model that produces structured, typed decisions instead of natural language text, marking a new era in enterprise AI. This development aims to improve decision automation within software systems, reducing reliance on traditional large language models (LLMs) that generate prose. The release, announced on September 15, 2026, signals a shift toward decision-focused AI that emphasizes speed, cost-efficiency, and reliability for enterprise workflows.
Jev is built on the concept of ‘System One’ models, inspired by Daniel Kahneman’s psychology, which emphasizes quick, intuitive judgments. Unlike typical LLMs, Jev responds to structured questions by returning typed decisions with associated probabilities, enabling software to act directly on its outputs. Its design aims to eliminate errors caused by output formatting issues, often called ‘hallucinations,’ by restricting responses to fixed schemas. The model can handle questions of three types: choices, scores, and yes/no probabilities, making it more like a function call than a conversational agent.
Funded with $40 million led by DCVC and developed by Diogo Almeida—co-inventor of RLHF and InstructGPT—Jev is positioned as a more reliable, faster, and cheaper alternative to traditional LLMs. It claims response times between 70 and 500 milliseconds at a cost of approximately $0.042 per million input tokens, which TypeSafe states is roughly 194 times faster and 445 times cheaper than comparable workflows. The approach challenges the prevailing assumption that every enterprise problem requires text generation, instead focusing on decision automation within software systems.
TypeSafe emphasizes that Jev’s accuracy depends heavily on how users structure their questions. Benchmarks show Jev achieving around 67.8% agreement with reference models in some tests, but independent evaluations indicate lower accuracy—about 62.6% on a phishing email classification task—highlighting that performance is still improving and requires careful question design. The model’s confidence estimates are sometimes over- or under-confident, underscoring that reliability remains a work in progress.
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
Transforming Enterprise Automation with Typed Decisions
The launch of Jev represents a significant shift in how AI can be integrated into enterprise workflows. By producing structured, typed decisions instead of free-form text, Jev reduces the risks of output errors and hallucinations that have plagued traditional LLMs. Its speed and cost advantages could enable automation of many routine decisions—such as support ticket routing, compliance checks, and operational alerts—that previously required human judgment or expensive reasoning models. This development could lower automation costs and increase system reliability, impacting how companies deploy AI in daily operations.
Moreover, Jev’s approach questions the long-held belief that natural language generation is necessary for most decision-making tasks. Instead, it suggests that decision-oriented models, which produce actionable data directly, may be more suitable for enterprise needs, especially where speed and accuracy are critical. If widely adopted, this could accelerate the shift from conversational AI to decision automation, fundamentally changing enterprise AI strategies and tools.
As an affiliate, we earn on qualifying purchases.
The Rise of Decision-Focused AI and Industry Shift
Over the past three years, the AI industry has largely centered around large language models (LLMs) like GPT, which generate human-like text for a variety of applications. These models promised improvements in reasoning, understanding, and context length, but also introduced issues such as hallucinations, overconfidence, and high costs. Many enterprises rely on LLMs for tasks like customer support, content creation, and code generation, often with human oversight to mitigate errors.
In recent months, industry leaders and researchers have begun questioning whether text generation is always necessary. The development of Jev by TypeSafe signals a pivot toward models that deliver structured, decision-oriented outputs. This shift is partly driven by the need for faster, cheaper, and more reliable automation solutions, especially in enterprise environments where errors can be costly. The model’s emphasis on schema conformity and probabilistic outputs reflects a broader trend toward more predictable and controllable AI systems.
Diogo Almeida’s background—co-inventor of RLHF and InstructGPT—gives Jev a pedigree rooted in improving AI reliability. His company’s critique of RLHF techniques, which often lead to issues like mode dropping and overconfidence, underscores the move toward alternative training methods like Reinforcement Learning for Calibrated Decisions (RLCD). This focus on calibration and structured responses aims to address core weaknesses of traditional LLMs, making decision automation more dependable.
“Jev isn’t about generating words; it’s about producing reliable, structured decisions that software can act on directly.”
— Diogo Almeida, CEO of TypeSafe
As an affiliate, we earn on qualifying purchases.
Performance and Reliability Challenges Still Unresolved
While Jev shows promising speed and cost benefits, its accuracy and reliability are still under evaluation. Benchmark results indicate that its agreement with reference models varies, with some independent tests showing lower accuracy (around 62.6%) in real-world tasks like email classification. Its confidence estimates can be overconfident or underconfident depending on question types, raising questions about its readiness for critical enterprise applications. The extent to which Jev can consistently outperform traditional models across diverse tasks remains uncertain, and further testing is needed to establish its reliability in production environments.
As an affiliate, we earn on qualifying purchases.
Next Steps: Broader Testing and Adoption in Enterprise Workflows
TypeSafe plans to expand testing of Jev across different industries and decision tasks to better understand its strengths and limitations. As the model matures, more enterprise clients are expected to pilot its integration into automation pipelines, especially for routine, high-volume decisions where speed and cost are critical. Future updates may include improved calibration of confidence estimates and expanded question types. The company also aims to refine its training methods to boost accuracy and reliability, addressing current performance gaps. Widespread adoption will depend on how well Jev can demonstrate consistent, dependable decision-making in real-world scenarios.
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, enabling direct software actions and reducing errors related to output formatting.
Can Jev replace human judgment entirely?
While Jev aims to automate routine decisions quickly and reliably, its accuracy still depends on question design and context. It is not yet clear if it can fully replace human judgment in complex or nuanced cases.
What are the main limitations of Jev currently?
Performance varies across tasks, with some independent tests showing lower accuracy than expected. Its confidence estimates can be over- or under-confident, and it still requires careful validation before deployment in critical systems.
How might Jev impact enterprise AI strategies?
If proven effective, Jev could shift focus from conversational AI to decision automation, enabling faster, cheaper, and more reliable enterprise workflows, especially for high-volume routine decisions.
When will Jev be available for broader enterprise use?
TypeSafe plans to expand testing and pilot programs in the coming months, with wider deployment contingent on further validation of its accuracy and reliability in diverse real-world scenarios.
Source: ThorstenMeyerAI.com
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
