Marc Andreessen — AI, crypto, 1000 Elon Musks, regrets, vulnerabilities, & managerial revolution

Asked whether he has been tempted in the last 14 years to found a company other than a16z, Marc Andreessen says the firm itself was that company. He and his partner Ben started it in 2009 and it let them exercise their entrepreneurial ambitions. It now has over 500 people, small for a tech company but big for a venture firm, and in his words it "has let us get all those urges out."

Pressed on whether any product needs to exist, he says he asks himself what he would do if he were 23 again, and he always has ideas. But founding is a commitment that changes your life, and he cites Sean Parker's description of it.

Starting a company is like chewing glass. Eventually, you start to like the taste of your own blood.

He tells anyone who asks whether to start a company that the answer is no, because it is a gigantic, emotional, irrational decision. He calls the process romanticized but not romantic, and since he has done it multiple times he does not revisit it, at least for now.

Patel suggests a 38th pitch in a day must feel similar. Andreessen disagrees: people are wired to handle stress differently, and he is fine with it, even seeks it out, but past a certain level it degrades his life. The distinction he draws is responsibility. A founder carries everything that goes wrong and is up at four in the morning, while an investor has a layer of buffer. He helps portfolio companies through crises, but it is not his company, so the stress is more diffused and easier to deal with. This is his description of his own temperament, offered without supporting evidence.

On blogging, Andreessen says the original blog ran from 2007 to 2009, after which the new firm soaked up his time. He now writes and posts intermittently, partly because he likes to experiment with new formats in a world made fundamentally different by social media.

Asked for today's biggest bottleneck, comparable to deployment before AWS, he answers: the impact of AI on coding. He sees a real possibility that basically every application category is upended within five years. In the old model there was a database, a front end, forms and known interaction models, and the industry had a good shared understanding of how humans and machines communicate across the windowing, mobile and web eras.

He expects future applications to be more of a dialogue between human and machine, written or spoken, with the human guiding and getting real-time feedback in a loop. The tools are at the very front end and an entirely new stack must be built. His wording is hedged throughout ("might," "potentially"), so this is a forecast and not a measured result.

On why a16z has no dedicated AI fund, he says the firm treats all of this as software, its core business, so the core venture fund is the core software fund. AI is the next turn on software and the most integral thing the firm does. Separate funds are for areas structurally different in how industries work.

Patel lists outcomes (Netscape sold for $10 billion, Chrome at roughly 2.7 billion users, Opsware sold for about $1.7 billion, AWS near $100 billion in yearly revenue) and asks whether these companies would have dominated their markets as independent startups. Andreessen declines to speculate. He spends virtually no time on the past because he cannot change it, and says people who revisit old decisions are less effective because they mire themselves in what-ifs and counterfactuals. He offers no estimate for any of the companies named.

His reasoning is that everything is straightforward on paper while reality is messy. Every startup charts a path-dependent course through a complex adaptive system, which he links to the earlier fascination with chaos theory. With 8 billion people interacting and starting companies that intersect in markets, random elements are everywhere, so the same scenario from point A can end at point B one time and point Z another. His advice to founders tempted to revisit every decision is to take the skills they have and do something new.

Asked whether venture capitalists are part of the managerial elite, Andreessen summarizes James Burnham as describing two kinds of capitalism. In bourgeois capitalism the owner often puts their name on the door, often owns 100% of the business and runs it, so ownership, control and operation sit with one person. Patel's one-line interjection about Andreessen Horowitz plays on this image.

In managerial capitalism, ownership and management separate. A public company has dispersed shareholders who pay no attention and cannot run it, and a professional managerial class takes over, ending up in control even if it owns only about 1%. Andreessen says Burnham held this was inevitable, good or bad, because modern organizations become too big, complex and technical to run without professional managers. He believes this is what played out and that the transition is now basically complete.

The advantage is that managers are often very good at running things at scale, as in health care and education. The cost is that, in his account, they lack the training, background, personality, temperament and incentives to build new things, since a manager's number one job is not to upset the applecart or introduce risk.