Way back when everyone wanted to buy development land, my friend Jeremiah Shamess of Colliers used to always tell me that the only way to do this was to either (1) pay the most or (2) believe in something that others don’t. This — making non-consensus bets — is something I like to talk about a lot on this blog, but what did that mean back then?
Well, when sites were seeing dozens of offers and the market was hyper-competitive, you really had to work to find any sort of overlooked value. Maybe it was an assembly. Maybe it was a density opportunity that others were missing. Or maybe it was a rail setback that the market felt would neuter the site, but that you had a solution for. Whatever the case, believing in something different was hard work.
Today, things are a lot different. The consensus bet would be to not buy development land in the first place, and the non-consensus bet would be to buy. But instead of having to believe in unique unlocks for a site, it’s obvious that the greater obstacle is believing that the market will be there to absorb your space. And if it is there, at what price?
Nobody really knows, and that’s what makes it non-consensus. But as always, non-consensus bets are where the greatest opportunities exist. That was true when the market was booming, and it remains true today.
Cover photo by Alfan Ziyyadan on Unsplash
The first time I ever used dial-up internet was sometime in the 1990s. Some of you will remember that a company called CompuServe used to mail out floppy disks with "10 free hours." And I still remember the feeling of amazement the first time I tried it. Suddenly, I could chat with people from around the world. Remember a/s/l? It was so enthralling that those 10 free hours certainly didn't last very long.
Fast forward to more recent times, and I got that exact same feeling of amazement when I started diving into crypto. The first time I created a wallet, logged into to a service (using only the wallet), and then transferred funds around, I thought to myself, "Wow, this is a fundamental shift in how the world works." A lightbulb went off. And I still feel this way about crypto, which is why I remain long ETH.
But now I'm also excited about AI (along with the rest of the world). With every new model release, it gets that much more impressive. Last week I wrote about Gemini 3 and, since then, I decided to cancel my ChatGPT subscription and move all my activity over to it. I'm sure that a better model will get released before we know it, but for right now I'm having a lot of fun creating just about everything.
Here's a cartoon isometric of Toronto that I prompted to include "landmarks" and the day's weather.
September 2025 was a milestone month for Waymo in California: It reached 1,000,000 paid driverless rides. This represents a year-over-year increase of ~182%, which is a pretty good sign that the technology works and that customers like it.
(Note: The dramatic falloff in rides in June 2025 was because of anti-ICE protests and vandalism taking place in Los Angeles and San Francisco. The company decided to temporarily suspend operations.)
This is still a small fraction of the traditional ride-hailing market, though. According to the California Public Utilities Commission, Uber and Lyft combined complete somewhere around 300-320 million passenger trips per year in the state. That averages out to roughly 25-27 million trips per month for context.
Way back when everyone wanted to buy development land, my friend Jeremiah Shamess of Colliers used to always tell me that the only way to do this was to either (1) pay the most or (2) believe in something that others don’t. This — making non-consensus bets — is something I like to talk about a lot on this blog, but what did that mean back then?
Well, when sites were seeing dozens of offers and the market was hyper-competitive, you really had to work to find any sort of overlooked value. Maybe it was an assembly. Maybe it was a density opportunity that others were missing. Or maybe it was a rail setback that the market felt would neuter the site, but that you had a solution for. Whatever the case, believing in something different was hard work.
Today, things are a lot different. The consensus bet would be to not buy development land in the first place, and the non-consensus bet would be to buy. But instead of having to believe in unique unlocks for a site, it’s obvious that the greater obstacle is believing that the market will be there to absorb your space. And if it is there, at what price?
Nobody really knows, and that’s what makes it non-consensus. But as always, non-consensus bets are where the greatest opportunities exist. That was true when the market was booming, and it remains true today.
Cover photo by Alfan Ziyyadan on Unsplash
The first time I ever used dial-up internet was sometime in the 1990s. Some of you will remember that a company called CompuServe used to mail out floppy disks with "10 free hours." And I still remember the feeling of amazement the first time I tried it. Suddenly, I could chat with people from around the world. Remember a/s/l? It was so enthralling that those 10 free hours certainly didn't last very long.
Fast forward to more recent times, and I got that exact same feeling of amazement when I started diving into crypto. The first time I created a wallet, logged into to a service (using only the wallet), and then transferred funds around, I thought to myself, "Wow, this is a fundamental shift in how the world works." A lightbulb went off. And I still feel this way about crypto, which is why I remain long ETH.
But now I'm also excited about AI (along with the rest of the world). With every new model release, it gets that much more impressive. Last week I wrote about Gemini 3 and, since then, I decided to cancel my ChatGPT subscription and move all my activity over to it. I'm sure that a better model will get released before we know it, but for right now I'm having a lot of fun creating just about everything.
Here's a cartoon isometric of Toronto that I prompted to include "landmarks" and the day's weather.
September 2025 was a milestone month for Waymo in California: It reached 1,000,000 paid driverless rides. This represents a year-over-year increase of ~182%, which is a pretty good sign that the technology works and that customers like it.
(Note: The dramatic falloff in rides in June 2025 was because of anti-ICE protests and vandalism taking place in Los Angeles and San Francisco. The company decided to temporarily suspend operations.)
This is still a small fraction of the traditional ride-hailing market, though. According to the California Public Utilities Commission, Uber and Lyft combined complete somewhere around 300-320 million passenger trips per year in the state. That averages out to roughly 25-27 million trips per month for context.
Here's a photo of a woman standing in the middle of a street in Tokyo wearing a trench coat and holding an umbrella. My prompt also asked it to make it look like a "grainy digital photo."
And here's a knolling shot (new word I just learned) featuring the gear of a global citizen (or globizen if you will). I prompted each of the objects, down to the white panel on the Blue Jays hat.
It's not perfect. Text remains an issue. If you look closely at the front of the passports or the text on the Fujifilm camera, you'll see that it's AI. But it's only a matter of time before this goes away. These kinds of images used to require a lot of time and effort. Now I can create them with one hand on my phone while I'm eating a bowl of cereal and having a morning coffee. There's zero marginal cost.
Thank goodness I've got more than 10 hours of usage.
Chart via Charlie Bilello
Here's a photo of a woman standing in the middle of a street in Tokyo wearing a trench coat and holding an umbrella. My prompt also asked it to make it look like a "grainy digital photo."
And here's a knolling shot (new word I just learned) featuring the gear of a global citizen (or globizen if you will). I prompted each of the objects, down to the white panel on the Blue Jays hat.
It's not perfect. Text remains an issue. If you look closely at the front of the passports or the text on the Fujifilm camera, you'll see that it's AI. But it's only a matter of time before this goes away. These kinds of images used to require a lot of time and effort. Now I can create them with one hand on my phone while I'm eating a bowl of cereal and having a morning coffee. There's zero marginal cost.
Thank goodness I've got more than 10 hours of usage.
Chart via Charlie Bilello
Share Dialog
Share Dialog
Share Dialog
Share Dialog
Share Dialog
Share Dialog
