Early motor cars were built as horseless carriages. The same high seat, the same buggy springs, and on some models a socket to hold the whip. A whip for the horse that was no longer there. A new technology spends a while wearing the shape of the thing it replaces before it finds its own.
That image opens the manifesto of a company called TypeSafe, and it is aimed at the chat box. AI trained to be a helpful, articulate assistant makes sense if a person is sitting on the other side of it, and produces an AI that needs a human in the loop instead of one that runs quietly in the background. Their line on AGI is blunter: we already have general intelligence, and most software is no more intelligent than it was. Intelligence today is like databases before SQL, powerful but bespoke every single time. The manifesto closes on the line that gives this piece its title: we're building prod, not God.
On 15 September they shipped what the argument had been building towards. It is called Jev, and it does something strange for 2026: it refuses to write anything.
No paragraphs. No summary. No code. You ask it a question and it returns an answer with a number attached saying how sure it is. That is the entire range of what it will say, and the reaction from developers has been warmer than a launch like this usually gets.
TypeSafe calls the category System One models: a class of AI model built to make fast, structured decisions that software can use directly. You hand one a state, it returns typed answers and probabilities. That is the whole contract.
The direction is what makes it interesting. For two years every lab has improved the same object in the same direction: longer context, deeper reasoning, more modalities, better prose. This one goes sideways, and gets there by taking something away. Give up writing text and you can work out every answer at once instead of one token after another, guarantee the shape of what comes back, and drop cost and latency by two orders of magnitude. Subtraction as the feature is rare in this industry.
How it works
You send a block of text — an email, a support ticket, a page of a contract, the state of a running program — and a question with the answers already listed. Is this command dangerous: yes or no. Which team owns this: platform, billing, or security. How urgent is this, on a scale of one to five. Back comes the answer, plus a number saying how confident it is. You can ask a hundred such questions about the same text at once, and they are all answered together rather than one after another.
A chat model does this too, badly. You ask it to classify something, it writes a sentence explaining its reasoning, wraps the answer in JSON, and you parse the result and hope the shape is right. Most of the time it is. Occasionally it invents a category you never offered, or returns prose where you expected a number, and somewhere three layers down your pipeline something breaks at two in the morning.
A chat model gives you a paragraph and asks you to trust it. This one gives you a number and tells you how much to.
Jev cannot do that, because the shape of the answer is fixed before the model ever runs. The name comes from Kahneman: System 1 is the fast, intuitive mode of thinking, as against the slow deliberate kind. The useful way to hold it in your head is the one TypeSafe uses itself, in a footnote, slightly embarrassed about it: a smart if-statement. Ordinary code cannot decide whether a customer sounds angry. This can, in about a tenth of a second, and then ordinary code decides what to do about it.
What it costs
A frontier chat model charges somewhere between twenty cents and ten dollars per million words of input, and roughly five times that for what it writes back. Jev charges four and a bit cents per million words in, and nothing at all for the answer, because the answer is a number rather than an essay. A round trip takes between 70 and 500 milliseconds instead of somewhere between three seconds and five minutes.
Asking this model a question costs about what a database query costs, and takes about as long.
That is what has developers interested, and the early reports are specific. Vercel's Pranit Sharma told TechCrunch they swapped out an OpenAI model that screens potentially dangerous commands and got answers five to eighteen times faster, more accurately. Mike Taylor at Every ran 777 judgments across 37 documents in under seven tenths of a second, for roughly a quarter of a cent. Armin Ronacher, who wrote Flask, pointed at the use case that probably matters most: deciding, in real time, whether a job is hard enough to deserve an expensive model at all.
Why it is named after a Victorian economist
In 1865 William Stanley Jevons noticed something about coal that everyone else had backwards. Steam engines were getting more efficient. Each one burned less coal to do the same work. The obvious conclusion was that Britain would therefore burn less coal.
Britain burned vastly more. Cheaper steam power meant steam power turned up in places nobody had bothered to put it before, and the new demand swamped the saving many times over.
TypeSafe named the model Jev after him, which tells you how they see this. They are explicit that Jev is not smarter than the big models, and that on the tasks it handles it is roughly comparable. The claim is about cost: two orders of magnitude cheaper and faster at those tasks. What that changes is how many things become worth doing at all.
When intelligence cost dollars and seconds, you spent it at the edges of your system. At cents and milliseconds, you can afford to put it in the middle.
Every piece of software you run is full of decisions currently made badly by brittle rules, or made by a person, or simply not made. Which of these five thousand documents should a human look at first. Is this supplier email a complaint or a query. Does this transaction pattern look like the one we saw in March. Those are not questions worth a three-second, three-cent model call each. At a tenth of a second and a fraction of a cent, they might be.
Their own demo is a bot playing Doom by asking the model roughly ten questions a second, at about seven dollars an hour. They cheerfully admit a normal non-AI bot would play better. The rate is what matters here.
Now the cold water
The founder is Diogo Almeida, who spent four years at OpenAI and co-authored the InstructGPT work that ChatGPT was built on. The press has him as "a ChatGPT inventor", which is shorthand; he was one contributor among a large credited team. DCVC led a $40 million seed and Forbes reported a $200 million valuation, on the word of one person close to the deal.
So the pedigree is real and so is the money. The evidence is thinner than the enthusiasm.
"It cannot hallucinate" is true in a narrower sense than it sounds. Jev cannot return an answer that is not on the list you gave it — that is a mathematical property of how the output is constructed, not a behavioural improvement. It can still pick the wrong answer off that list, and sound confident doing it. TypeSafe's own documentation is clear about this: calibration is a property of a large group of predictions, not a promise about any single one. If the model says 80% a thousand times, roughly 800 should be right. Which 800, it cannot tell you.
The headline benchmarks are the company's own. The "193.6× faster, 444.6× cheaper" on the homepage comes from four workflow tests TypeSafe designed, built by TypeSafe's own staff, scored against the average answers of two rival frontier models rather than against verified ground truth. To their considerable credit they say all of this themselves, in a nuance box under each claim, and add that those figures are probably at the optimistic end. It is more honesty than most launches manage, and it is still not independent evidence.
Nobody outside the company knows what it is. There is no paper, no weights, no parameter count, no training recipe. Almeida is tight-lipped about the internals, outside observers suspect an open-weight model underneath, and the company says only that it is transformer-based and trained on synthetic data it generated itself. The new training method has a name, RLCD, and very little else in public.
Every headline number in this launch was produced by the company the numbers flatter. TypeSafe says so itself, in a box under each claim, which is rarer than it should be.
It is visibly jagged. TypeSafe publishes a list of what it is bad at, which is a genuinely unusual thing to publish: counting, arithmetic, dates, numerical precision, questions requiring several hops of indirection. Irrelevant context in the input makes it worse. Adversarial text can steer it. Ask it two related questions and the probabilities are not guaranteed to be arithmetically consistent with each other.
And in the one head-to-head an outsider has published, the tradeoff shows up exactly where you would expect: Taylor's test found Jev about 580 times cheaper and 25 times faster than Claude Fable 5.1 — and it caught six of seven planted errors, where Fable caught all seven.
What this means from here
It runs in America. There is no European or Swiss hosting option. TypeSafe's privacy policy says European data goes to the United States for processing. The data processing agreement carries the standard EU clauses, the UK addendum and a Swiss provision letting Swiss data subjects sue in Swiss courts, which is not the same thing as a Zurich region. Client data, patient data, anything under banking secrecy: that is a conversation with your legal team before it is a conversation with your engineers.
It also thinks in English, by the company's own account. The Bern consultancy apexAI has early access and is testing it for document classification and routing, and says plainly that it cannot yet draw reliable conclusions for German business text. In a country that works in four languages and does business in a fifth, test it on your own correspondence before believing anyone's benchmark.
Somebody was going to sell the decision without the essay.
Whether or not this particular company wins, the shape of the argument holds and other people will build it. A great deal of what we currently call "using AI" is a chat model doing a job that never needed prose, classifying and routing and scoring and flagging, and being paid for a paragraph it then throws away.