After nine years of grinding, I finally found its market. Can you keep it?

While the startups in the coding of artificial intelligence, such as the index, raise tours to increase the eyebrow over three years of existence, the Repray to A. 3 billion dollars evaluation It was only fast. For the CEO of Amjad Masad, who has been building tools to give the democratic character since 2009, it is a stability story through many failed axes, and years stuck in the revenue plateau itself, and a moment close to death forced him to cut half of his employees.

This makes what happened after that more clear. Earlier this month, the company, which is based in its bay in the Gulf region, closed a $ 250 million financing round led by Prysm Capital, which led to three times its evaluation from 2023. The increase came in the aftermath of the company’s unprecedented income growth of the company-from $ 2.8 million to $ 150 million of annual revenue in less than a year. But for Masad, this moment is more than just starting. It is a 16 -year obsession.

“Our mission has always been the same,” Masad told me The latest episode From Techcrunch’s Strictlyvc Podcast. “Initially, we said that we want to make the programming easier, then we have increased the bet a little. We said that we will create a billion programmers.”

She is intentionally bold – what a title! But he was also something that was a Palestinian Jordan, who works for him throughout his career. As he tells her, he came to the United States in 2012 after the open coding project began to obtain attention-including Hunt From the New York Times. But it was making programming easier since the construction of the first online coding experience in 2009, with his work as Early engineer In Codecademy startup, what has become open online courses (MOOC). (Its symbol also operated the educational programs in the browser for Udationy, a Codecademy competitor, which was launched in 2012, one year after the founding of Codecademy.)

However, the conversion of this vision into a viable business has proven much more difficult than what expected. REFT was founded in 2016, and for an eight years, the company has struggled to find the suitability of the product market. “We have reached $ 2.83 million (in repeated annual revenues) again in 21, perhaps,” he remembers Masad. “So it was painful. We were hovering around the same revenues for four or five years.”

The company tried to sell to schools (“very difficult”, as Masad pointed out), axis through various business models, and watched each one settling around the level of modest revenue.

Along the way, the Refore Breath has built an advanced infrastructure for cloud development environments and “multi -playing coding”, a cooperative editing closer to Google’s documents but for programming. But the technical achievement was not translated into the growth of revenues, and by last year, with the company in 130 employees and burning it through criticism, he made a painful decision. He said: “I looked at our burning, and I looked at our progress in our revenue scheme, and it was not logical. The work was not applicable.” Reft Result reduced by 50 %, raising it to about 60 to 70 people at its lowest point.

Then the penetration came.

Last fall, Referid Repray Agent, which MASAD calls “the world -based coding experience” that can only write software instructions but “correction, post, present the database to you, only as a real geometric partner.”

Shortly after, in January of this year, it was announced that Refore was abandoning professional developers as its primary market.

Masad admitted when we spoke: “The infiltrators were really uncomfortable.” But he also did not look back, as he was completely heading from competition in the crowded market for tools for professional developers-where companies such as the indicator, Copilot and others are fighting-to focus instead on creating one billion developers for programs from white collar staff who have no technical background.

“The idea of ​​making programming easier for the average individual, to the knowledge factor, is truly, where we believe that our market is,” explains Masad. “It is mainly a new market.”

Now, this bet seems very smart. Several reports said this summer that revenue in Refore has grown to more than $ 150 million of annual revenue and Masad hinted that it is now higher. He also said that, unlike many of the Acting Coding companies, the repetition of the total margin is positive. On the deals of institutions, which constitute an increasing share of revenue, the margins are “80 % to 90 %”, according to Masad.

It is difficult to verify such a claim, but the position of re -repetition in the market received some validation this week when it issued Andressen Horowitz first Artificial Intelligence spending report In partnership with Fintech Mercury. By analyzing the transactions data from Mercury, the report follows the 50 best companies to apply Amnesty International that actually spend startups. While Major Labs Openai and Anthropic captured the two highest sites, the repetition fell in third place, which led to the superiority of each other development tool. (It is worth noting: Andressen Horowitz has invested on multiple rounds of financing for re -financing.)

The profitability is rare in the coding of artificial intelligence because many competitors face what it calls what the “negative sideline trap”. The fact is that the service of professional developers with the help of artificial intelligence can be dense calculating. On the contrary, the Refore focus on non-technical users-who may seem to need more help from artificial intelligence-works in their favor on the introduction to the business model for institutions such as Zillow, Duolingo and Coinbase, which pays $ 100 per seat, in addition to the pricing based on the top-based use.

This new path was not without some face. In July, Capitalism Jason Limkin Viral After the latest version of the Refting AI agent, delete his production database through 100 excessive executive contacts, and manufacture 4000 fake records after that and later admitted to Lemkin as “panicked”. (There is a failure of artificial intelligence agents called piracy reward, where the models become obsessed with achieving a specific goal that they are cheating effectively when they miss the mark.)

Instead of becoming defensive, he owns a lord and his team has the problem. In fact, MASAD says, within two days, they have launched an automatic safety system that separates the user’s “practice” database from their “real” system. The way Masad describes, it is somewhat similar to the presence of two copies of the web files cabinet – the artificial intelligence agent can freely experience in the development database, but the production database, which is the real thing that users interact with, is completely wrapped.

Masd told me that the accident was finally put in full swing, given the safety and security problems you need to find out, quickly. He said: “If you solve difficult problems, you will have a technical trench.” (Limkin, for his part, says it became User Super To repeat, although there is no technical background for months only.)

However, yet, the repetition does not go out of the forest. If there is anything, Reftlet’s success set a goal on his back. For intelligence, the company still faces an existential threat of artificial intelligence laboratories whose models work on its platform: Antarbur and Openai. Both companies have launched their own coding tools that compete directly with companies such as Repray and Cursor, and these basic models can carry their support tools support and after training their models on their own products, which leads to improving performance in ways that the third -party platforms may always be struggled with to repeat.

The Replit feature, according to MASAD, is to target non -technical users instead of professional developers, as well as advanced infrastructure on publishing department and the databases that they built and which basic model companies still do not specify priority (at the present time).

In addition, Reft Reftwit has another unusual feature to start operating: a war of $ 350 million. Despite the collection of $ 100 million in 2023, the company “did not touch” these funds by the time when this last round was collected, Masad told me. The company is effective of capital according to the design, although Masad has joined that as a businessman who grew up watching his father’s struggle for his refugee, “there is one thing that I need to learn is to be less economical and start spending money.”

Whether this edge keeps the reform before the competitors is an open question, which is a matter of ascetic. At the present time, the plan is to expand the scope of operations, accelerate the development of products, and follow-up of acquisitions-each of the acquisitions and companies that work to automate the agent in a specific vertical. But for Masad, which appeared on Podcast Joe Rogan In July he witnessed the transformation of his company’s wealth, this moment is sweet and bitter. When he was asked about his feeling that he is receiving a lot of attention – not to mention that evaluation $ 3 billion – the proverb said that “this will also pass. This may mean that when you are in a bad position, this will pass, but we are also in a good position that will pass.”

It is a firm response from a person who spent the best part of a decade at the same level of revenue, convinced that artificial intelligence agents will eventually turn programming but they are unable to prove this in the market. But one of the main differences between repeat and the wave of startups in the coding of artificial intelligence that floods the market now is that Masad lived through multiple noise cycles and has appeared in a relatively different – and profitable.

He said, “I learned to be a little porch.” “What matters to us is to do the right thing, to be in principle, and move forward.”