Here are some interesting charts from a16z showing that, despite its dominance today, tech still represents a smaller percentage of the US stock market than railroads did at the turn of the 20th century. One parallel you could draw from this is that "tech" as we know it today, may not be so dominant a hundred years from now.

But railroads continue to play a critical function in the modern economy. They are still the most cost-effective way to move heavy goods over long distances. A single freight train can carry the load of several hundred semi-trucks.

The more interesting parallel might be the one that a16z raises in its post: railroads both led to further economic growth and rewired the way businesses and organizations were structured.

Railroads were a new kind of business requiring massive scale and coordination, which led to new ways of thinking about "management." Perhaps not surprisingly, it was around this time (1881) that the world's first collegiate business school was formed at the University of Pennsylvania.

The parallel to AI today, as argued by Jack Dorsey and maybe others, is that it's going to similarly rewire how businesses are organized and what middle management does:

"Instead of absorb and route information, maintain alignment, pre-compute decisions, etc.—the kind of coordination that management typically is responsible for—in an AI business, humans move to the edges, to focus their judgment on customer contact and human interactions."

At least, this is the hypothesis.

But if it does prove to be true, let's consider what we often discuss on this blog, which is: what will it mean for our cities and built environment? Well, what I find interesting about the above quote is that it suggests AI will push humans further toward the things that we are uniquely suited to do: interacting with other humans and building meaningful relationships.

And if that is, in fact, what happens, then there's no more efficient place to be than in dense urban cities. Looking someone in the eyes, shaking their hand, and slurping ramen noodles together at a busy bar counter is not something that AI will be able to do for us.

Cover photo by Mike Beaumont on Unsplash

Charts from a16z

Fast, high-quality decisions and approvals are the lifeblood of organizations. And if you've ever worked in development or construction, you know that there are a lot of decisions and approvals — some small, some big — but all of which can delay and hurt a project. Ultimately, the objective is to achieve both high-quality and high-velocity decisions. But how?

Very broadly speaking, you want a bias toward action and progress. How this plays out might depend on the specific situation at hand, but here's one technique that we try to use whenever possible. I call it (as of 30 seconds ago) the "default-to-yes" principle. It works well for approvals and reviews, and it is very common in construction.

All you need are two things: (1) a date by which something needs to be reviewed or approved and (2) a default yes. A default yes means that if I don't hear from you by the deadline established by (1), I'm simply going to assume your answer is yes and move on. Consent is implied unless you object.

The opposite of this is a "default-to-no" approach, which means things get stuck until someone gets around to reviewing or approving the thing. That's far less optimal because there's no outer limit to how long something might take. With the default-to-yes approach, I know progress will happen no later than X days from now.

This is just one specific technique, and I'm not suggesting it will work for all decisions and approvals, but there's significant value in high velocity. And to achieve that, you want a deeply ingrained cultural bias toward action.

Cover photo by Rubén Bagüés on Unsplash

Back in 2011, Marc Andreessen wrote a widely cited blog post where he argued that "software is eating the world." In some ways, it feels like just yesterday that I first read it. But it has been 15 years, and boy, has the world changed. Now, the worry is that AI is eating software.

It has become significantly easier to write code, to the point that in the span of only two years, Google has gone from 0% of its new code being written by AI to now over 75% of it! But it's not just big companies. I know lots of non-technical people who wanted software that could do "X," and so they just vibe coded a solution. Done.

In fact, I've been experimenting and doing the same for several months now. It has become so easy that I feel an obligation to do it. But as we know, if everyone can do it, then it means there is no longer any value. The value will necessarily need to be created in other ways.

Earlier this week, we spoke about Uber and how being asset light — previously a hallmark of the gig economy — is potentially now a liability. Well, this is a broader theme. Josh Brown, CEO of Ritholtz Wealth Management, even coined a term for this: HALO. This stands for Heavy Assets, Low Obsolescence.

The general idea is that you now want physical stuff with a big moat that is immune to being disrupted by someone in their parents' basement using Claude Code. Hard assets are, arguably, where you want to be today. I guess that means real estate is back, baby!

Cover photo by Tim Mossholder on Unsplash