Why Small Companies Can Produce So Much: Marginal Costs, Infrastructure, and 10x Engineers—Plus Robin on Labor-Driven Growth

Midjourney, a leader in AI-generated art Midjourney has begun offering a free video version. In other words, users can now use AI to make videos. This phenomenon, comparable to GPT and used by more than ten million people worldwide, has only eleven employees: one founder, eight R&D staff, one legal specialist, and one finance specialist. There are many similar cases. Amazon had only 158 employees when it went public. Instagram had only thirteen employees when it was acquired for US$1 billion. Why do so many impressive companies have so few people? Technology researcher Wei Xi has analyzed this question and identifies three reasons. First, these are technology companies. Compared with traditional businesses producing physical goods, their marginal product-development costs are very low. A physical-goods business must add labor for each additional item it produces. The larger the business, the more people it needs. In the technology world, however, adding a function may require only copying existing code. Serving 10,000 users rather than one may require only a few additional servers. Technology products can be replicated without adding labor costs. Second, technology infrastructure is becoming increasingly mature, so startups do not need to reinvent the wheel. OpenAI’s rise, for example, may appear to rest on its staff of more than 300 people. In fact, it also draws on training data from Wikipedia and other internet platforms, NVIDIA’s efficient graphics cards, Microsoft’s Azure cloud services, numerous papers published by Google, and more. Third, there is a 10x engineer effect in technology. If your talent is even a little stronger than that of your peers, your output may be several times greater. In other words, what matters in technology is not how many exceptional people a company has, but how exceptional its strongest person is. It is about the ceiling, not the average. For example, Jeffrey Dean, the current head of Google’s AI division, developed almost all of Google’s key early products virtually single-handedly. --- One more point can be added: all tremendously successful companies sit right at the critical point of a phase transition. This idea was proposed by Safi Bahcall. He earned a PhD in physics at Stanford, but joined McKinsey after graduating. He noticed that success in business closely resembles phase transitions in physics: a substance changes from one state to another. Water vapor condensing into droplets and droplets turning into ice are both phase transitions. In business, a phase transition means a change in the state of an organization or industry. Companies undergoing such transitions generally have the following characteristics. First, they are doing something many people are attempting but no one has accomplished. Many people work on AI, for example, but no one has achieved artificial general intelligence (AGI). Over time, the industry gets ever closer to the critical point of a phase transition. A unicorn could emerge one day. Second, their members are always divided into two camps, neither willing to yield to the other. Management’s job is to maintain a dynamic balance between the groups, keeping different views in constant friction. In other words, great products often emerge from the collision of ideas rather than from design alone. Third, the team must be especially resilient, and it cannot have too many people. Someone previously estimated that the critical headcount was exactly Dunbar’s number: 150 people. Beyond that, an organization’s capacity to innovate may decline. So how do you keep a team innovative? The most important thing is not designing an elaborate system, but managing the critical point: keeping the company on the boundary between steady, dependable execution and imaginative ideas. The longer it stays there, the greater its chances of innovation. #DedaoHeadlines20230616 ---- Robin: “If your revenue growth comes from headcount growth, or is highly correlated with it, the business you run is probably not fundamentally a technology company. It is a typical traditional business that needs more people to build scale.”