Jev, explained simply

The version with no jargon

Think of the models you already know as writers. You ask a question, they write you an answer, and you read it. They are good at that and they charge for every word they produce.

Jev is not a writer. It is closer to a clerk with a clipboard and a set of boxes to tick. You say: here is a customer email, and here are three departments. Which one? It ticks a box. You say: here is a comment, is it abusive, yes or no? It ticks a box. It cannot write you a paragraph about the email even if you ask, because that is not what it does.

Because ticking a box is so much less work than writing, it is fast and cheap in a way that changes what is worth doing. One person here sorted 100,000 social posts in about twenty seconds for 67 cents. Another classified 500 emails for three and a half cents. Doing either of those with a writing model would take hours and cost real money, so in practice nobody would have done it at all.

What you give it, and what comes back

Two things go in. First, the text to judge — an email, a comment, a page, a row from a spreadsheet. Second, your question, with the answers you are willing to accept written out.

Three things come back. The answer it picked. How likely it thought each of your options was, as percentages. And a confidence score, which is the useful one: it says how sure Jev is overall. Most people use that number to build a rule like “if it is more than 80% sure, just do it; if not, show it to a human”.

You can also ask many questions at once about the same text, and they all come back together for almost no extra cost. Someone here asks 61 questions about a single draft post and gets every answer in about a second.

What it cannot do

• It cannot write. No drafts, no summaries, no code. A normal model still does all of that.
• It cannot invent an answer outside your list, which is a feature — but it also means you have to know the list in advance.
• It cannot hear audio or watch video. It can look at PNG and JPEG pictures, but only as a preview feature since Jev 1.14, which TypeSafe says is not ready for production.
• It is not a calculator. Sums, counting, comparing dates, matching text exactly — leave those to ordinary code, which gets them right every time.
• It can still be wrong. “Cannot make things up” means the answer will always be one of yours, not that it is always the right one.

It works next to the writing models, not against them

Almost nobody swaps a writing model for Jev. They use both. The writing model does the part that needs words — draft the reply, explain the bug, write the code. Jev does the small judgments around it, the ones that happen dozens of times per task.

Three of those come up again and again. Which model should handle this request: the cheap fast one, or the expensive careful one? Is this action safe to run on its own, or should a person look first? And: did that actually work, or is the assistant claiming it is finished while the tests still fail?

None of those need a sentence written. They need a box ticked, right now, for almost nothing. That is the gap Jev fills.

Why it is called Jev

After William Stanley Jevons, who noticed in 1865 that as steam engines got better at burning coal, Britain burned more coal rather than less, because suddenly it was worth using engines everywhere. The bet behind the name is the same: make a decision cheap enough and software will start making millions of them.

It was released on 15 September 2026 by TypeSafe AI, and people started building with it the same day. This site has been collecting what they made ever since — 855 projects so far.

Where to go next

If you want the technical version — what a call looks like, what it costs, how it scores against other models — read what is Jev. If you want to see what people do with it, the builds are grouped by job on Jev use cases. And if you want to feel it rather than read about it, the AI slop detector runs on Jev: paste a URL and it answers 35 questions about the page in about a quarter of a second.

And for the same explanation written for developers, both of these walk through a real call, the confidence thresholds and where the headline numbers come from.

Akshay 🚀

@akshay_pachaar

LLMs vs. Jev, clearly explained! TL;DR The key difference is not that Jev generates faster. Jev does not generate text at all. A traditional LLM receives context and produces an answer one token at a time. Even when the output is a small JSON object, every token depends on those generated before it. Jev receives the same context but evaluates predefined decisions directly. When those decisions are independent, it can evaluate all of them in parallel. Consider an agent handling a failed deployment. It may need to determine: → Whether the incident is urgent → Which team should handle it → Whether the proposed command is risky → Whether the task is complete An LLM generates a response containing these answers sequentially. The application then parses and validates it. With Jev, you define the questions and expected answer types upfront. It evaluates them together and returns typed answers with probabilities. Jev supports three decision primitives: 1. **Choice** selects from known options, such as engineering, billing, or sales. 2. **Score** places the input on an ordered scale, such as low, medium, or high risk. 3. **Noul** evaluates a yes-or-no condition and returns the probability that it is true. The probabilities matter as much as the selected answers. If engineering receives 91% probability and billing receives 9%, automatic routing may be reasonable. If the probabilities are 52% and 48%, the system can escalate, gather more context, or call a stronger model. This keeps control inside ordinary software. Code owns the thresholds and consequences. Jev supplies the semantic judgment that a normal `if` statement cannot derive from unstructured text. It works best when the possible answers are known, the decision depends on meaning, and a careful person could judge the input quickly. It is not designed for writing, summarization, code generation, arithmetic, or decisions requiring several dependent reasoning steps. Independent questions can run in parallel, but decisions that depend on earlier results must remain sequential. Jev also cannot return an option outside the declared schema, but it can still select the wrong valid option. Type safety prevents malformed outputs, not incorrect judgments. The clean mental model is this: LLMs generate new language when the answer space is open. Jev evaluates known paths when the answer space is bounded. I wrote the full breakdown explaining Jev and where it fits. The article is quoted below.

Guide·x.com

LLMs vs. Jev, clearly explained

Akshay Pachaar: Jev does not generate text at all. It answers Choice, Score and Noul questions in parallel, and your code owns the thresholds.

Akshay 🚀

@akshay_pachaar

Guide·x.com

Jev Clearly Explained

Akshay Pachaar’s X article on Jev as a millisecond decision layer, and where it sits next to an LLM.

Made with Jev is independent. The model itself is announced in TypeSafe’s launch thread.

Common questions

What does Jev mean?

It is a model name, not a word. TypeSafe AI named it after the Jevons paradox: when something gets cheap enough, people use far more of it, not less.

Is Jev a chatbot?

No. It never writes a sentence. You give it some text and a multiple-choice question about that text, and it gives you back one of your own answers.

What is Jev good for?

Decisions a program makes over and over: is this email urgent, which folder does this go in, is this comment abusive, which link should the agent click next. Thousands of them, for pennies.

Does Jev ever make things up?

Not in the usual sense. It can only answer with one of the options you wrote, so it cannot invent a department that does not exist or hand back a broken answer. It can still tick the wrong box, which is why it also tells you how sure it is.

Do I need to be a programmer to use it?

To use it directly, yes — it is an API that other software calls. You can use things built on it without writing any code, and some of them are free and open in a browser tab.