Alexandr Wang on Two Keys to Closing the Gap: Extreme Focus and Continuous Overdoing
In a 53-second interview, Alexandr Wang identifies two keys to closing the gap: extreme focus and continuous overdoing. The English subtitles state that if one is extremely focused, progress will be very fast; long-term investment of time and energy, along with consistently overdoing, will lead to even faster progress. He attributes his ability to achieve this to focus and overdoing, suggesting that others may simply not go that extra mile.
The circulated post summarizes this into three points, stating that in the industry, if one is required to walk 1 mile, they should walk 10 miles; if others deliver 100%, they should consistently deliver 200%; and every delivery should be treated as a life-or-death battle. The original piece does not mention these multiples or specifically name 99%. However, the verifiable written version is his October 2024 memo "Do Too Much": leaders never just do enough, but rather overdo; you are the upper limit of what anyone in the company cares about, and you must do more, care more, and try more than seems reasonable; too much is the right amount.
The memo normalizes what the outside world sees as excessive: what is called excessive optimism is merely optimism, excessive communication is just communication, excessive delivery is just delivery, micromanagement is just management, and ruthless prioritization is just prioritization. He writes that without Steve Jobs' obsession with details, there would be no Apple; without Elon Musk's near-fanatical execution, there would be no SpaceX or Tesla; he has never seen ordinary effort yield extraordinary results. Without overdoing, Scale would not be the company it is today.
The company's actions align with this statement. Scale AI was co-founded by him and Lucy Guo in 2016. Within six months of the rise of generative AI in 2022, he shifted most of the team to produce data for large models. In June 2025, Meta acquired approximately 49% equity for about $14.3 billion, valuing the company at around $29 billion, and he became Meta's Chief AI Officer. At that time, Forbes estimated his net worth at about $3.2 billion. The internal creed at Scale was written by him as increasing pace and ambition shapes reality.
This is about pricing attention to a set of management standards, not new financing. Those who share this buy an executable gap formula, while the original piece provides focus plus going the extra mile. The beneficiaries are teams that write excessive delivery into their recruitment and client standards, while suppliers who stop at 100% as per contract are under pressure. The 49% equity purchased by Meta represents this data capacity and the execution record of shifting the team within six months, not just an inspirational subtitle.
Source: Public Information
ABAB AI Insight
Wang's path is to exchange data services for entry tickets to model companies. He dropped out of MIT in 2016 to co-found Scale with Lucy Guo, initially labeling data for autonomous driving and government clients. When large models lacked training data in 2022, he redirected most of the workforce to generate training data within six months, expanding clients from automotive manufacturers to model labs like OpenAI and Meta. In 2025, Meta acquired about 49% non-voting shares for approximately $14.3 billion, and he joined Meta to manage AI. Overdoing here is not just a slogan; it transformed the company from a labeling workshop to a model fuel supplier.
The flow of money is that client budgets go into labeling, which is then capitalized by platform capital. Scale does not train the largest foundational models; it sells the tedious output that others are unwilling to do. The upper limit theory in the memo sets overtime and direction changes as the company's ceiling: had the 2022 shift followed the contractual pace, the window would have closed to other data vendors. Meta bought this validated switching speed, plus him.
Comparable examples include TSMC binding design companies with yield and delivery times, and Amazon initially binding retail customers with loss-making deliveries. Both are seen as excessive deliveries that later became default standards. Wang's current position is in a controlling phase: Scale's valuation is already anchored by Meta's 49%, and his personal role has shifted from startup CEO to AI head of a large platform, with excessive execution transitioning from external client acquisition to internal scheduling.
The essence is the transfer of pricing power. As model capabilities become concentrated in a few labs, the scarcity is not another intelligent speaker, but the capacity to deliver data weekly. The mechanism is that contracts only cover visible scopes, and going the extra mile becomes the ticket for the next exclusive order. Smart people supply excess, but those willing to do the tedious work are scarce, so the premium lies on execution records, not on IQ narratives.
ABAB News · Law of Cognition
- What others call excessive later becomes delivery.
- Contracts are written to 100%, but orders are left for 120%.
- IQ is in surplus, but those willing to do tedious work are scarce.