Coinbase Founder: The Crazy Journey Of Building A $100 Billion Company: Brian Armstrong

Armstrong repeatedly started by building something small—his tutoring marketplace, then early Bitcoin apps—without perfect clarity. Shipping the wrong Bitcoin app instantly revealed why its architecture was flawed and what Coinbase should be instead. His maxim, borrowed from Paul Graham, is “action produces information”: launch, observe, learn, and iterate instead of endlessly theorizing ideas.

His tutoring marketplace stagnated while taking a 10% cut and managing payments; users quickly went around the platform and paid tutors directly. When he stripped out payments, turned it into a free directory, added optional $10/month ‘featured’ profiles, traffic doubled yearly and ultimately sold for ~$2M. At Coinbase, adding a simple Bitcoin ‘buy’ button—removing the friction of acquiring BTC—unlocked organic, compounding growth.

In Buenos Aires, Armstrong forced himself—using ideas from Seth Godin’s ‘The Dip’—to ask what he’d still want to be doing in 10 years even if it hadn’t yet ‘worked’. The only answer was tech entrepreneurship. Later, his commitment to economic freedom via crypto kept him from pivoting to trendy “blockchain for banks” when the market turned; he would have shut Coinbase down rather than build something he didn’t believe in.

An introvert who loves building and learning, Armstrong realized that traditional CEO expectations—12+ direct reports, heavy people management, constant crisis‑mode—drained him. He cut his direct reports to four, delegated energy‑draining responsibilities, and focused on product, technology, and strategy. This made Coinbase run better and allowed him to stay engaged a decade in, rather than join founders who exit and never want to run a company again.

To avoid becoming a ‘one‑product’ company, Coinbase uses a 70/20/10 model: 70% of resources on the core business, 20% on adjacent bets, and 10% on high‑risk ‘venture’ bets. This keeps a founder mindset alive, builds a portfolio of revenue streams, and reduces dependence on a single S‑curve. Armstrong underscores that tolerance for failure paired with recognition of good execution (even on wrong ideas) is essential to make this work.