NEW YORK, September 24 — A startup founded by a former OpenAI researcher has stepped out of stealth with a decision model that does something large language models cannot: give the same answer twice. TypeSafe AI emerged from stealth on Sept. 15 and released the model, called Jev, the same day, according to TechTarget.
Jev is built for structured decision-making inside software rather than text generation, and it can return only three kinds of output — nouls, a measure of how true something is; scores, a graded rubric; and choices, multiple-choice answers. That narrowness is the point.
Jev is less powerful than a full-fledged large language model, but it can make quick determinations that developers use to guide LLM decision-making.
A System 1 model, by design
TypeSafe’s founder is Diogo Almeida, a former OpenAI researcher who co-authored a 2022 paper that established reinforcement learning from human feedback as a key method for training LLMs. In a Sept. 15 blog post, Almeida described Jev as a System 1 model, borrowing a cognitive processing concept from Daniel Kahneman’s book “Thinking, Fast and Slow.” System 1 decisions are reflexive and immediate, said Ned Bellavance, founder and technical educator at Ned in the Cloud. “Is this shirt blue?
Is the pasta cooked? Am I on fire?” Bellavance said.
Large general-purpose reasoning models, by contrast, are System 2 — more verbose and non-deterministic. “System 2 decisions are deliberate and thoughtful,” Bellavance said. “What’s the best way to approach this situation?
How do I really feel about the color blue? What is life?” For developers, the practical difference is speed and cost.
Jev’s structured outputs could offer a faster, more efficient alternative for teams embedding resource-intensive general-purpose LLMs in their applications to guide decisions, Bellavance said. “Think of Jev interactions more like a function you’d add to a program or automation routine, and not something that would hold a conversation or write code for you,” he said. “If you need to classify, score, or get a yes/no decision on something, Jev is a good fit.”
Two problems enterprises keep hitting
Jev addresses two common enterprise AI problems, said Torsten Volk, an analyst at Omdia, a division of Informa TechTarget: “One, a lack of deterministic outcomes, and two, token cost.” Other popular repos on GitHub have targeted the cost problem caused by unnecessary LLM verbosity using custom Claude skills, such as ponytail and caveman.
But Jev’s appeal also stems from the fact that it uses a separate model that doesn’t require pre-training, as some other decision models do, said Laurie Voss, head of developer relations at Arize, an AI observability platform vendor, in an email to TechTarget. “For years, companies have been fine-tuning small models into decision models and providing decisions cheaply and quickly,” Voss said.
“The big jump is that” — the source material ends there. The launch has garnered early interest from developers facing AI cost and governance challenges. What to watch next: whether TypeSafe can turn that early developer curiosity into production deployments, and whether the decision-model approach becomes a standard layer beneath the large models that dominate the conversation.
