This week a major AI lab put out an impressive new model and pulled it back within forty-eight hours. That, plus a vendor trying to charge a company thirty grand a year to access its own data, plus the oldest gut-check in business (what happens if you get hit by a bus?), all pointed at the same thing: ownership.

The barrier to building collapsed, you can stand almost anything up in an afternoon now. But the gold rush has everybody building fast on rented land: rented models that can be pulled, rented platforms that change the rules, rented access to your own data, whole operations sitting on one person's personal account. It all feels like ownership until the lease is up.

In this one I confess to doing the exact thing I warn clients about, make the case for the slower, unglamorous move of building where the client actually owns it, and close on the difference between a house built on sand and a house built on rock, and what I actually want to hand my kids.

AI in Action with Tal Swicegood is an honest take on what AI is really doing in real businesses. Find what I'm building at lvluplocal.com.