Sterner Stuff

Marc Andreessen says it’s time to build. I agree.
But why haven’t we been building? What went wrong?

1. I’ve been re-watching Cosmos (the original 1980 series) and it strikes me how limited our scientific ambitions have become compared to the supposedly stagnant 1970s, let alone the techno-optimistic and war-driven prior decades.

2. Voyager cost a billion dollars, inflation-adjusted. It’s paid back that amount dozens, maybe hundreds of times over: through direct technical advances, by producing decades of research output, and by inspiring countless young scientists and engineers.

3. Today, a billion dollars gets you… what? A single funding round based on community-adjusted EBITDA? 2% of a bailout for the airline industry? 500 yards of a new subway line in Manhattan? Break me a fucking give.

4. Not all investments have to generate cash on cash! If Voyager were venture-backed, it would be one of the greatest investments in the history of venture capital.

5. There’s an influential 2012 Paul Graham essay on frighteningly ambitious startup ideas. It’s a terrific essay as far as it goes, but I wish it went further!

6. For example, #1 on his list was “a new search engine”. Sure, replacing Google is hard, because Google has smart people and deep pockets and excellent tech and strong network effects. And if you succeed, you’ll almost certainly make pots of money.

7. But is it really ambitious? Not in my book. It’s evolutionary, not revolutionary. (The original Google was revolutionary). Ambition should be made of sterner stuff.

8. I see so many startups that seem ambitious at first glance, but they’re just retreading old ideas. Nuclear power. Supersonic passenger planes. Satellite networks. Electric cars. These were all done 50 years ago!

9. This is not a critique of Silicon Valley. This is an exhortation to do even more! Impractical, outrageous, insane ideas? Bring them on! We don’t have enough startups exploring immortality, or the matrix, or interplanetary exploration, or large-scale bio-engineering.

10. But at least startups are trying. Outside of the startup / tech / software / venture ecosystem, most large organizations aren’t even trying to build anything ambitious.

11. Vaccines for polio and measles and smallpox saved millions. The green revolution fed billions. The supply chain revolution improved the standard of living for the whole world. Where are the vaccines and seeds and container ships of today?

12. Our cities, our healthcare, our education, our factories, our transportation: none of them are transformationally improved from 50 years ago. And that’s inexcusable.

13. Marc Andreessen puts it bluntly: “The problem is not technology. The problem is not money. The problem is desire. The problem is inertia. The problem is regulatory capture. And the problem is will.”

14. When large numbers of independent agents exhibit the same patterns of sub-optimal behaviour, it’s worth investigating the underlying mechanisms and incentives that drive this behaviour. Why don’t we have the desire any more?

15. Why can’t we build?

16. One set of explanations involves measurement error. As William Gibson wrote, “The future is already here, it’s just not very evenly distributed”. Maybe we’re not looking in the right places for ambition and innovation.

17. China can build a hospital in a week, and transform their entire society in a month (modulo data veracity as always).

18. Globally, more people have been lifted out of poverty in the last couple of decades than in all of human history, and technology (of every type – global trade, engineering, agriculture, medicine, logistics) gets much of the credit.

19. So maybe bemoaning a lack of ambition is an overly western-centric point of view?

20. It’s also clear that there’s been a ton of innovation in software and information.

21. I walk around with a supercomputer, an infinite reference library, a universal communications device, and unlimited music, movies and books in my pocket. I can have almost any item in the entire world shipped to my doorstep in a day, with one click.

22. That’s … actually revolutionary. Revolutions don’t have to be in macro engineering like rockets and skyscrapers. But thanks to the hedonic treadmill, we get used to everyday micro revolutions, and perhaps we unfairly discount them.

23. Another mostly benign explanation is that the seeming slowdown in ambition is just part of the natural cycle of technology adoption,

24. Carlota Perez created the definitive framework for thinking about this. In her work, she posits that new technology stacks go through clearly identifiable phases: installation (genesis & initial excitement) followed by deployment (expansion & maturity).

25. She provides five historical case studies: the industrial revolution (1770-1830), steam & rail (1830-1880), steel & electricity (1880-1920), automobiles & oil (1910-1970), information & telecom (1970-present).

26. We’re currently moving from the installation phase to the deployment phase of our current tech stack.

27. This transition is usually accompanied by widespread cheap adoption of the stack, mass market expansion, new funding models, and the formation of oligopolies – all of which we’re seeing happen today.

28. Notably, this step does not typically involve new technological innovation, intense exploration or frenzied competition: hence the perception of a slowdown in ambition.

29. Uber is a great illustration. It’s a truly fascinating company: a creative business model unlocking a massive source of value (underused cars and drivers), creating customer behaviours that pose novel challenges to regulators and incumbents alike, with software coordinating it all.

30. But Uber isn’t a tech creator; it’s a tech deployer. Uber was built on top of a number of hardware and software developments that all came together at the right time: smartphones, messaging, payments, GPS, street maps, cloud compute and more.

31. Knitting these together at scale is operationally complex and incredibly impressive; it generates immense surplus value (and wealth), but it’s deployment, not creation.

32. We see this pattern again and again. The most successful startups of the 2010s are Stripe, Airbnb, Shopify and Uber – all deployment plays. (I’m not quite sure what kind of play WeWork was).

33. This is not a bad thing! Ubiquitous and pervasive deployment of the current information + communication + software tech stack will lead to a substantial rise in the real standard of living worldwide.

34. But in the long run, this deployment means that our current stack is on the road to being played out. If we truly desire transformative changes, we’ll need to find a new stack.

35. A third, less benign set of explanations invokes the sheer power of software and money in the modern world.

36. At the cutting edge of software, Jeff Hammerbacher wrote: “The best minds of my generation are thinking about how to make people click ads”. It feels so wasteful.

37. Finance is similar: there’s a tremendous amount of ingenuity and innovation in that field, but to what end? Making a few rich people marginally richer and a few other rich people marginally poorer?

38. Together, finance and software have hoovered up lots of talent and lots of capital. They’re the most influential industries in the world right now. As go finance and software, so goes the productive frontier of our economy.

39. And for both good and bad, both of these industries are examples of what Venkatesh Rao identifies as soft technology. Which means they offer a very specific kind of operating leverage.

40. This is key. The leverage offered by software at web scale, and by money at institutional scale, means that even small improvements or advantages can lead to immense returns for practitioners.

41. It’s easier (and safer) to aim for those small improvements than to swing for the fences – which is why smart and self-interested individuals choose that option.

42. As a result we seem to have traded step-function change for marginal efficiencies.

43. Maybe big innovation requires charismatic iconoclasts with a take-no-prisoners attitude: think Edison, Jobs, or Musk. Single-minded visionaries, not risk-aware optimizers. And maybe we’ve stopped producing them.

44. The same incentives that discourage individual risk-taking apply at a company level. As we move from installation to deployment, companies become ever more risk-averse. And investors love it.

45. It’s easier to take proven technologies and bring them to new markets than it is to invent new technologies. (Investors call this “executing a repeatable playbook”.)

46. It’s easier to win an audience through acquisitions and consolidation than it is to win customer hearts and minds. (Investors call this “increasing economies of scale”.)

47. It’s easier to grow revenue via oligopolies and customer capture than it is to engage in open competition. (Investors calls this “fostering network effects”.)

48. It’s easier to boost reported EPS using buybacks and leverage than it is to actually improve earnings quality. (Investors call this “running an efficient balance sheet”.)

49. All of this makes perfect sense! Why should a company do anything the hard way when an easier alternative exists? Why should investors take risk if they don’t have to?

50. Perezian deployment coupled with the power of software and money leads to a world that looks awfully like the one we live in.

51. Large companies growing ever larger, protected by ever-stronger network effects, a lack of appetite for fundamental research, and a never-ending quest for optimization.

52. This segues into our fourth set of explanations: systemic failures.

53. Peter Thiel points out that much of the progress we’ve had in recent decades has been in the field of bits, not atoms. And perhaps not coincidentally, bits are far less regulated than atoms.

54. More generally, costs are higher and progress is slower in fields that tend to have lots of regulations – healthcare, education, finance and housing being the most obvious examples.

55. Francis Fukuyama talks about internal clientelism: the idea that the US government is in thrall to interest groups associated with both right (big business, oligopolies, crony capitalism, rampant financialization) and left (NIMBYs, rent-seeking unions, public sector ossification).

56. Government capture leads to regulatory capture, which leads to stagnation.

57. Tellingly, the Perezian framework predicts increasing regulation of bits as the software stack becomes ubiquitous. This pattern is true of every tech as it approaches maturity; it’s not a cause for optimism.

58. Then there’s demographics. The centre of political power in the US has for a long time been the baby boomers.

59. Fifty years ago they may have been rebels. Today, they largely favour the status quo – which in turn favours them, via the tax code, the health, education and retirement systems, zoning regulations and more.

60. Society wide, there’s a smugness, a complacency, a lack of imagination and compassion and courage, an unwillingness to accept criticism, and a pervasive selfishness that is disheartening and I have to say tragic.

61. As someone who grew up in a poor country, I will never understand how a society so incredibly rich can still have poverty, and homelessness, and malnutrition, and poor education, and preventable disease.

62. What’s worse, endless infighting over the resources that could mitigate these ills means that there is no bandwidth left to actually build.

63. What unites all of these systemic themes is rent-seeking over productivity. Squabbling over slices from a fixed pie rather than trying to make the pie bigger.

64. Ben Thompson puts it vividly: “There are countries that act, and there are countries that talk.” Talking (and policing talk) means fighting for your slice of the pie; not making the pie bigger.

65. When rent-seeking is the goal, all your efforts become positional.

66. There’s another troubling possibility. It may be that – atrophied by years of risk aversion and rent-seeking and financialization – we’ve simply forgotten how to build.

67. Institutional knowledge, once lost, is hard to recover. The cure for scurvy was known, forgotten, rediscovered and forgotten again, before Szent-Györgyi isolated Vitamin C. We still don’t know exactly how to make Roman concrete or Damascus steel.

68. While modern historians have shown that the Dark Ages were not as tenebrous as previously thought, it remains the case that much of Western Europe was materially worse off in 700 CE than in 70. And the loss of institutional knowledge has a lot to do with that.

69. This, I think, is what Patrick Collison is getting at when he talks about speed. What starts as the inability to do things fast ends as the inability to do things at all.

70. Fortunately, there may be a way out. Here’s Andreessen again: “The things we build in huge quantities, like computers and TVs, drop rapidly in price. The things we don’t, like housing, schools, and hospitals, skyrocket in price”.

71. The best way to repair an atrophying muscle is to exercise it. The more you build, the better a builder you become; and the more you can build. Building well becomes a self-fulfilling prophecy.

72. This is the central thesis of the Foundation trilogy: a dedication to science and technology can counter an incipient Dark Age. In the long run, humanist optimism defeats greedy obscurantism. (Asimov is one of my heroes.)

73. Measurement, distribution, deployment, incentives, risk-aversion, financialization, regulations, capture, leverage, positioning, knowledge, atrophy.

74. The truth of why we can’t build is probably some combination of all of the above.

75. Change won’t be easy. We need to attack all of these impediments at once; fight a battle on many fronts.

76. Create better incentives for risk-taking. Fight regulatory capture. End the zero-sum positional games and the rent-seeking and the jockeying for status. Distribute opportunities. Guard against lost knowledge. And build, build, build.

77. The best time to build was yesterday; the next best time is today.

Further Reading