07 · Manufacture Compound Growth · The Sam Altman Playbook

In his Stanford startup lectures, Altman’s rule of thumb is stark: about 5–7% week-over-week growth is good, 10% is exceptional, and a modest-looking weekly rate sustained for long enough compounds into absurd absolute outcomes. He likes to note that Airbnb’s founders taped their target growth curve to the bathroom mirror. Growth, in this frame, is both the goal and the diagnostic. If it isn’t moving, the work isn’t real.

ChatGPT became the extreme case: a curve that, once lit, ran almost on its own.

~100M Nov '23

~300M Dec '24

~500M Mar '25

~900M Feb '26

ChatGPT weekly active users (reported milestones). Separately, it had hit ~100M *monthly* users within about two months of its Nov 2022 launch, the fastest-growing consumer app on record at the time.

Doesn’t transfer as a universal law. A single growth metric is dangerous for businesses where the right early motion is not growth, such as deep tech, biotech and regulated markets, and premature scaling kills more startups than it saves. ChatGPT’s curve was also mostly demand Altman rode rather than made: it shipped as a low-key research preview that nobody, including OpenAI, expected to explode. Manufactured growth and a wave caught in time look identical on the chart.