Hey {{first_name | there}},

You know what's weird about AI in 2026?

We built machines that can write poetry, generate images, and hold conversations better than most humans. 

But the biggest breakthrough this month is a model that refuses to talk at all.

Let’s talk about it.

Last week, TypeSafe AI launched Jev. It's not a chatbot. It can't write your emails. It can't generate code. It literally cannot generate text at all. 

But what can it do? It’s a new type of AI model that can take decisions. Fast ones.

Here's what I mean:

You give it a piece of information (like a customer complaint or a support ticket). You ask it questions about that information like "Is this urgent?" or "What category does this fall under?" And it gives you a final decision to make. That's it.

No chatty explanations. No "Hey there! Let me help you with that!" Just: here's the decision, here's how confident I am in it, move on.

Just think about how most AI agents work right now:

An LLM decides what to do. A tool executes. The LLM evaluates. Then it decides again. Every single step requires a full model call, which costs money and takes time.

Jev replaces those middle steps. Instead of asking a Claude model to classify a ticket (which takes 3 seconds and costs money), Jev does it in 70 milliseconds for almost nothing.

The founder, Diogo Almeida, literally helped invent ChatGPT at OpenAI. He worked on the RLHF methods that made chat models work. And he left because he realized chat models alone wouldn't lead to automation.

His exact words: "Models have been superhuman at chat for years, so where is all the automation?" 

That question kept him up at night for four years. Jev is his answer.

Jev was built for AI agents. Not for you. Not for me. But for the machines that make decisions in software workflows.

It's what they call a "System One" model. The idea is that most decisions in automation don't need a full reasoning model. They need quick classification, routing, scoring.

And that's exactly what Jev does. But here's the question nobody's asking:

If AI agents are making decisions faster and cheaper than humans, where do we fit?

Let me give you a real example.

Imagine a customer support workflow. Before Jev:

- Ticket comes in
- LLM reads it (costs money, takes time)
- LLM classifies urgency (costs money, takes time)
- LLM routes to right team (costs money, takes time)
- Human reviews and decides next step

After Jev:

- Ticket comes in
- Jev classifies urgency in 70ms for almost free
- Jev routes to right team in 70ms for almost free
- Human reviews and decides next step

See what happened? The human still reviews. But the boring classification and routing decisions? Gone.

I would say, yes.

Yes, because the architecture is genuinely different. Jev isn't a smaller LLM pretending to be fast. It's a completely new type of model that processes questions in parallel instead of generating text token by token. The engineering is real.

Yes, because the cost reduction is massive and measurable. Vercel already reported 5 to 18x speed gains replacing OpenAI with Jev for safety classification. Cloudflare rushed to add it to their gateway.

But it has its limits too, don’t be fooled.

Jev can't think. It can't reason through complex problems. It can't handle anything that needs explanation or creativity. It's brilliant at narrow, repeated decisions. And terrible at everything else.

It also can't do arithmetic, treats dates as text, and has a 32K context limit. These aren't small limitations.

So where do humans stand?

Human roles are about to change so much.

When we use ChatGPT to classify support tickets, we're using a $10 billion reasoning engine to do what a simple if-else statement could do if we just understood the data better.

Jev acknowledges that most automation decisions don't need general intelligence. They need fast, consistent classification.

The human role shifts from doing the classification to defining the classification rules. From routing tickets to designing the routing logic. From making the small decisions to overseeing the system that makes those decisions.

I won’t call it replacement, but things are definitely evolving.

Well, either way, it was interesting to read about this AI model, and I’m trying a bunch of stuff with it. Reply to this and let me know if you’ve tried anything so far?

- Aashish

AI is moving fast and most people are falling behind. 

The Rundown AI keeps you ahead of the curve. 

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.