With a team of roughly 20 people, its valuation has skyrocketed from 200 million to 10 billion in just one month: this "non-talking" AI has put right the direction that ChatGPT once led astray.

A company named TypeSafe has recently become the focus of Silicon Valley. Its product, Jev, is an AI that "does not talk" — it does not chat with you, but only helps you get work done. The company has only around 20 employees, but its valuation has soared from 200 million US dollars to 10 billion US dollars in less than a month.

Diogo Almeida, the founder of Jev, made a bold statement: "ChatGPT has led AI down the wrong path."

I. "ChatGPT has led AI down the wrong path": The Rebellion of a Non-Chatting AI

First, let's look at what Jev is. It is a closed-source model that only provides services through APIs, and it is built as an "agent" — that is, an AI that can autonomously execute tasks, except that it splits the execution rights into two parts: it does the work, but leaves the final decision to human users. Its most distinctive feature is that it does not talk. It will not chat with you back and forth like ChatGPT, it only does one thing — help you get things done.

The media has given it a nickname: "the most popular mute model". The tasks it performs are very specific: playing 50 games at the same time, booking a flight ticket in 7 seconds, sorting out 724 ads in 40 seconds. All these tasks have one thing in common: they are all "actions", not "conversations".

The stance of founder Diogo Almeida is almost a rebellion against the entire AI industry. His exact words are "ChatGPT has led AI down the wrong path" — in his view, the direction of AI should not be "chatting with you", but "doing things for you". Over the past two years, everyone has been competing to build models that are better at talking, writing poetry, and answering questions, but they have forgotten what users really want.

The explosive popularity of Jev precisely proves that this rebellion hits a real pain point. Users are already tired of "chatty AI". They do not lack a friend to chat with, what they lack is a tool that can actually get things done. Jev pulls AI back from "conversation" to "execution", which immediately taps into a gap ignored by tech giants.

The phrase "led down the wrong path" refers not only to the product direction, but also to the misallocation of resources across the entire industry. Over the past two years, major tech companies have poured tens of billions of dollars in computing power and top talents into making models better at talking — expanding larger context windows, enhancing writing capabilities, and creating more human-like conversations. But when users use these more conversational AI, they find that they still cannot truly help them complete a whole task. No matter how strong the conversation ability is, if you ask it to book a flight ticket, it still has to guide you step by step; if you ask it to reply to a work message, you still have to rewrite the suggestions it gives yourself. That is exactly the gap Jev spotted: what users need is not "great chat performance", but "task completion".

II. Never Press the Send Button for You: Restraint as Survival Wisdom

The most noteworthy part of Jev is not how capable it is, but how "restrained" it is.

It follows a distinct design principle: it helps you reply to WeChat messages, book flight tickets, and sort out ads, but the final "send button" and "confirm button" are always left for human users to press. It provides you with suggestions, drafts and solutions, but never makes the final decision for you.

This level of restraint is of great significance in the current wave of AI agents.

In the same week that Jev went viral, Muse, an AI agent developed by Meta, was banned 13 days after launch because it could automatically compare prices and place orders for users, which hurt Amazon's revenue, and 6 banks also issued warnings. Muse and Jev are almost two extremes: one executes automatically, the other never executes automatically. The result is that one is banned, and the other becomes a hit.

This is no coincidence. The conflict between AI agents and platform giants is essentially a conflict over "who owns the users" — AI agents helping users bypass platforms is equivalent to touching the core interests of platforms. Muse crossed this red line, while Jev wisely positioned itself as the "co-pilot": I can help you make judgments, but the decision-making power is in your hands.

The logic is actually not complicated. Once an AI can place orders and complete transactions on behalf of users, it will bypass the platforms and erode their revenue. Muse crossed this line, but Jev did not. By keeping the "send button" in the hands of users, Jev essentially tells the platforms: I will not steal your users. This choice leaves it a way out before the full break between AI agents and platforms.

Some media have accurately summarized Jev's product logic: it turns "judgment" into a software component. It does not sell "doing work for you", but "helping you make more accurate judgments". This positioning not only avoids the red line of platforms, but also evades the ethical controversy of "AI replacing humans", and exactly caters to the user psychology of "I trust you, but I do not feel safe giving you the decision-making power".

This positioning as a "judgment component" bypasses the two most awkward pitfalls of AI: the first is "AI hallucinations" — since AI cannot guarantee that all the suggestions it gives are correct, leaving the decision-making power to humans essentially passes the responsibility for "who is to blame if something goes wrong" back to the users themselves; the second is the fear of "AI replacing humans" — it clearly tells you "I will not make decisions for you", so you will not feel that it is going to take your job. These two pitfalls are exactly the two most troublesome problems for current AI products. Jev bypasses both of them with the simple design of "not pressing the send button".

III. Around 20 Employees, Valuation Rising from 200 Million to 10 Billion: AI Has Shattered the Traditional Logic of "Running a Company"

Finally, let's go back to the most striking number: around 20 employees, with a valuation of 10 billion US dollars.

Strictly speaking, its valuation rose from 200 million US dollars to 10 billion US dollars in less than a month, an increase of about 50 times. This growth rate is crazy in any industry; in the AI industry, it has become a landmark case.

It highlights the fundamental impact that AI has brought to the concept of "company". In the traditional software era, for a company to be valued at 10 billion US dollars, it usually requires tens of thousands of employees, decades of accumulation, and huge sales and customer service teams. But now, it can be achieved with around 20 employees, one product, and just one month.

Why? Because AI has completely shattered the cost structure of "running a company". Jev follows the path of "small profits but quick turnover" — low-cost subscriptions and mass adoption, instead of burning money for growth. Data shows that it has already occupied nearly 13% of paid teams on a certain payment gateway. This figure means that it does not need a sales team of thousands of people to negotiate with customers one by one, the product itself can spread, get paid, and grow in the developer community.

The path of small profits but quick turnover is actually another rebellion against the logic of "burning money for growth". Large model companies have burned tens of billions of dollars and are still struggling with "how to make profits"; Jev uses low subscription prices, relies on spontaneous spread in the developer community, and can achieve growth without burning money. There is a simple truth behind this: the marginal cost of AI tends to zero, so high volume is more in line with the economic laws of AI than high unit price. But this also means that Jev's moat does not lie in "how strong its technology is", but in "how fast it runs and how widely it spreads" — once the giants react and crush it with free services and a mature ecosystem, it remains the biggest suspense whether a small team of 20-something people can resist the pressure.

The fact that around 20 employees is an advantage rather than a disadvantage is because under the leverage of AI, the marginal output of each person has been amplified many times. In the past, writing code, building products, operating, and providing customer service required a whole department; now with AI, one person can be as productive as a whole team in the past. The organizational form of companies is evolving from "manpower tactics" to "small teams empowered by AI leverage".

But the opposing view is equally sharp. Can a team of around 20 people support a valuation of 10 billion US dollars? Valuation can be boosted by market hype and imagination, but delivery, operation and maintenance, customer service, security and compliance all require manpower. When Jev evolves from a "viral developer tool" to a "platform serving tens of thousands of paid teams", the 20-something team may be completely insufficient immediately. The flip side of small profits but quick turnover is that low-margin businesses also have thin risk resistance — a single security incident or product failure may make the 10 billion US dollar valuation shrink instantly.

The restraint of "never press the send button for you" is not only Jev's moat, but also its ceiling. Refusing automatic execution means that it can only stay as a "co-pilot" forever, and cannot charge the premium of "autonomous driving". This is a safe business choice, but it also limits Jev's upside potential — under the general trend that AI agents will eventually move towards "autonomous execution", will a product that voluntarily gives up execution rights be replaced by a more radical and adventurous competitor? There is no answer to this question for now.

Back to Jev. Around 20 employees, one month, valuation rising from 200 million to 10 billion, a "non-talking" AI — no matter whether it is called a "new paradigm" in the AI era or "another bubble", it at least proves one thing: when everyone is competing to build better chatbots, the one who focuses on "getting work done" wins.

And the question it raises for the entire industry is more memorable than its valuation: should the direction of AI be chatting with you, or doing things for you? We cannot tell which answer is correct right now. But the fact that Jev is favored by both users and capital in such a short period of time at least proves one thing: users are running out of patience with AI that "only knows how to chat".