Alexandr Wang, founder of Scale AI and current Chief AI Officer at Meta, reflected on the first half of his time at Y Combinator in an interview: he wrote idea documents every day, feeling heavy anxiety, and after reading Paul Graham's "How to Come Up with Startup Ideas," he used the approach of "living in the future and creating what is missing" to reverse engineer his thoughts.

His judgment at the time was that human computing power would be dynamically orchestrated like computer computing power. There was no interface on the internet that could call for human annotation and processing like calling a server. He spent a night searching for a domain name, found that scale.ai was available for registration, and bought it. He later said that this name was an unexpectedly good decision.

After the initial wave of popularity from Product Hunt, there was a 4 to 6 month wandering period. There were few customers, and growth was not explosive, making success uncertain. He put an Intercom bubble on the official website, and when visitors clicked in, he was the one replying. It took six months to get the first core customer who truly wanted to scale. He described the initial feeling of being behind as everyone around him being in a fog; the gap was not about who acted two weeks earlier, but who could test the direction thoroughly.

Scale later developed a data annotation pipeline. In 2025, Meta acquired a 49% stake for about $14.3 billion, and Wang left to head Meta's Super Intelligence Lab, with his personal share valued at about $5 billion. Under new CEO Jason Droege, Scale told Forbes that last year's revenue was nearly $1 billion, up from $870 million the previous year, with cumulative revenue of $2.5 billion. Meta committed to paying Scale at least $450 million in service fees annually for five years. Wang stated that Scale is a key partner in training Muse Spark.

Muse is the AI agent launched by Meta during Wang's tenure. According to Sensor Tower data, Muse was downloaded 2.8 million times in the U.S. and Canada in the 12 days before its launch. The early starting point for Scale was not this product, but annotation needs like cat detectors in YouTube videos.

This was the founder using time to gain the first customer, not a valuation deal. The buyer was the early visitor who clicked on the webpage bubble, and the seller was Wang, who personally replied. Money and demand flowed from scattered consultations to the first willing large customer six months later, and then into contracts for the annotation pipeline. The beneficiaries were the data infrastructure that turned manual calls into interfaces; the ones under pressure were projects that only piled ideas in documents without engaging real visitors.

Source: Public Information