After some bug fix juggling that wasn’t going anywhere, decided to throw a problem at GPT-5.6 Sol Ultra, which I had never used before because it churns through tokens quickly. It spun up a few subagents to handle different facets of the problem. Fascinating.

John Gruber isn’t buying the Claude watermarking either:

What happens if you copy my text and quote it in something you write, by hand? Now you’ve got a watermark in your prose, invisible characters or something, that implicates that your prose is AI-generated? Even though you didn’t use AI and just copy and pasted text from me? Madness.

Maybe the watermark won’t work on short sections of text, like a blockquote, because there aren’t enough words to detect a pattern (unless every word is shifted with the algorithm, which seems insane). But who knows, because how it really works is a secret.

Following up on my comment yesterday, I now have Codex → OpenClaw working, so I can chat with my bot in the same place I have other AI chats. Codex wrote a plugin for itself to handle the connection, plus some setup on the OpenClaw server. 🤖

I don’t like the Claude watermark announcement. For images, yes, watermarks are important. “A picture is worth a thousand words” and there’s too much of a chance for misinformation. For text, it’s more complicated, as essays will often be 95% human, with AI-assisted edits.

Great episode of Dithering this morning. I’ve been wondering about the Apple / Google partnership too, with Google falling behind. I think Siri AI will be a good release, but it’s not really in the same ballpark as the frontier models. Try GPT-Live sometime.

I’ve now read more of OpenAI’s motion to dismiss the Apple lawsuit. This part sums it up:

Instead of pleading facts showing misappropriation, Apple recasts the benign, lawful conduct of former employees—who in many of the cited instances were in fact trying to help Apple, at Apple’s request, by making sure their former colleagues’ work could continue without them—as theft of so-called trade secrets. The proffered “trade secrets” are not specific, protectable information at all; they are generic categories of the product-development process…

It supports my original hot take.

People worry about AI bots overloading servers, but in my experience the dumb requests for potential security holes are much worse. Bots trying to get into non-existent files like htpasswd, wp-admin.php, etc. Servers getting hammered today.

Another section from Mark Zuckerberg’s essay:

For example, in Richland Parish, Louisiana, where Meta is building a large data center, teachers received a $50,000 bonus this year because of the increased tax revenue from our investment. The superintendent told us that teachers are now moving there from across the country and he believes it will become one of the nation’s best school districts.

This sounds like Ben Thompson’s argument to just give people money. If you think of data centers as intelligence factories, paying teachers more is actually perfect.

Mark Zuckerberg has a long essay today, basically an expansion of the op-ed he wrote recently. On personal agents:

It will have strong privacy and security options so you can trust it to handle all of your personal content knowing that no one else can access your information, similar to how encryption works on WhatsApp. You’ll be able to interact with your agent through any device, including your glasses to keep you present in the moment with the people you care about.

Practically in the same breath: privacy and AI glasses! People just don’t trust Meta with this.