I’m not optimistic about the future of cities.
As just one example, the hostility to housing development from nimbys, the grifting of unions, the short-term cycles of elections vs the desire of younger people to live “in the heart of the action” breeds resentment and anger.
Anger creates opportunities for simpletons to rise to power...people who can’t solve problems but trade in recycled bandaid ideas.
The masses don’t want to be told that a capitalist market can solve their problems. They don’t want to hear that it may take years.
Or that maybe they just need to move to a place where they can earn $60k and live well and that place isn’t Soho.
Instead they want someone to say they can just make it all better with a snap. They fail. And the anger breeds more deeply.
This cycle reinforces itself. Social, legacy, all types of media feed on anger. The spiral downward accelerates. Anger needs a boogie man.
There is a bottom. And it’s not pretty. You don’t want to be in a city as it gets to the depths of this hole.
I’m optimistic about smaller cities, towns. And the possibilities of geographic distribution of opportunity that technology provides.
Your milleu shapes you. AI and Zoom can’t replace the power of sitting with people who inspire great thought. More and more I don’t think those people will be in SoHo or SF.
I’m in a three person group chat that trades ridiculous memes multiple times/day.
One guy, we’ve talked live twice in five years. Once when he called me to ask me to be a groomsman in his wedding. The second at his wedding.
A “who’s cah are we gonna take” guy. Few.
Tim Apple is, perhaps, the greatest supply chain orchestrator of our generation. Hardware is a supply chain problem. Apple started designing things because it *could.*
Steve/Jony Apple designed things downstream of what customers needed. Could wasn't in their vocabulary.
It's been a lot of fun trying the Duo out for a week, but it feels like I've gotten to use an extraordinarily polished prototype — not a finished product that I'd want to buy and carry.
daringfireball.net/2026/10/t…
Why have I bought into the Muskonomy? This is why. At 9pm on a Friday night I got in my wife’s car to pick up her Target order. I made it one house down before the car started throwing flat tire warnings. I pulled over and sure enough. It was flat. I very slowly drove it right back into our driveway, hopped out, and filed a claim through the Tesla app with my Tesla insurance. Within minutes a tow truck was dispatched to pick the car up and take it to the nearest Tesla service center. Since they’re open Saturday’s, the car will be ready for pickup tomorrow with a repaired or new tire. On any other car this would be a nightmare. Trying to find a tire shop over Saturday and Sunday, make sure my tire is in stock, pay a towing company to come get it… it would consume the weekend. Not with my @tesla though. A few taps of the screen and all I have to do it pick it up at the end. Wow.
Prediction: Intuit in 10 years doesn't survive if they don't control the data source. They will try to acquire something like (or specifically) Yodlee or Plaid then claim to need higher security standards than currently available, which will require developer verification, ultimately metering calls to maintain revenue as the picks and shovels provider to the swarm of alternatives (software and or agents) that will ultimately supplant QB which is a widely used and wildly despised ledger, basically.
Much of what we call AI slop isn't based on bad code. It's slop because of speed.
Most good ideas don't come quickly. They simmer over time, iterate in conversations, blossom in the shower, walks in the woods.
The extended time building something is an opportunity to let new ideas come.
I think this is much of what we think of as AI slop, because the ideas are not fully baked.
Just because we can build things quickly doesn't mean we need to ship them immediately.
One of the biggest barriers to autonomous agent operations is expiring sessions.
I've long been critical of overzealous infosec people who think their systems are so precious that they expire sessions in unreasonably short time.
This has ripple effects on efficient operations. For example, our regional bank expires auth through Plaid every 10 days. This means tools like Ramp and Netsuite lose connectivity to a dozen accounts, each of which has to be re-auth'd individually.
Chase seems to do it every 30 days, which is at least somewhat better. And Chase auths at the login level rather than the individual account level.
Aside from banks, ridiculously short expiration windows basically mean background automations break. You can't schedule things or perform agent-based tasks easily, especially on headless hosts, without re-authing, as in the case of Netsuite, every 3 days.















