What Jev Got Right: Judgment as an Interface, Not a Paragraph
What Jev Got Right: Judgment as an Interface, Not a Paragraph For as long as software has been able to ask a model a question, almost every answer has come back as prose. Ask whether a support ticket is billing or account, and you get a paragraph. "This looks like a billing issue, though it could also be an account access problem — it depends on whether the charge was a renewal." A person reads…
Jev proved a key insight about artificial intelligence: judgment should be treated as an interface, not just a block of text. Traditionally, when a software program asks a language model a question, it receives a lengthy paragraph as an answer. However, Jev transformed this by delivering choices, scores, and probabilities instead. This change alters both shape and cost.
When the output is a set of values with probabilities, the surrounding system no longer functions as a pipeline with a language model in the middle. Instead, it becomes a pipeline with a component that has defined input and output contracts. This shift makes it much easier to test and measure the accuracy of the judgment.
The cost and time required to make these judgments dropped significantly. TypeSafe AI priced Jev at $0.042 per million input tokens, with processing times ranging from 70 to 500 milliseconds. At such a low cost, judgment is no longer a budget line but a negligible expense. Additionally, these decisions that were previously too expensive or time-consuming to make are now viable.
The ecosystem quickly adopted Jev, with a $40 million seed round from DCVC, led by its founder, Diogo Almeida, who co-authored InstructGPT and worked on RLHF at OpenAI. Within 24 hours of its launch, nearly 13% of Vercel's paid teams had already utilized the model, describing it as the fastest-adopted model they had seen. The adoption was so rapid that other gateways, orchestration frameworks, and observability tooling also integrated Jev into their systems.
However, while Jev solved the issue of expensive, frequent decisions, it introduced a new problem: the lack of a field to express the significance of some decisions over others. The pricing model treats all decisions equally, which makes it difficult to identify decisions that require more attention. Despite this limitation, Jev successfully made judgment cheap and scalable, solving a critical problem and earning its rightful place in the software ecosystem.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
