Finished reading: The Society of Unknowable Objects by Gareth Brown 📚

I think everyone is floored by this $1.5 billion settlement. Anthropic essentially won on fair use, so I just assumed the pirated books settlement was going to be a little more low-key than this. Hope the authors see most of the money. Never a dull moment in AI land!

Sarah Perez reporting on a Common Sense Media assessment of Gemini:

Common Sense also said that Gemini’s products for kids and teens ignored how younger users needed different guidance and information than older ones. As a result, both were labeled as “High Risk” in the overall rating, despite the filters added for safety.

This is part of what is probably a major shift underway at AI companies to worry less about helping create biological weapons and more about kids and psychological safety. Maybe futurists think too much about the sci-fi inevitabilities and not enough about today.

This is great. James Dempsey’s Liquid Glass song:

I made the lyrics as intelligible… as Liquid Glass controls are legible. Come on everyone, sing along! You don’t know the words, but neither do I. 🎶

Kagi News looks good! First they define the problem:

Driven by relentless ad monetization, news has become mental junk food - an endless stream of clickbait that destroys our ability to think deeply and clearly.

I’ve only watched a few clips of last night’s dinner with tech execs at the White House. Is it possible to accept the invitation and not grovel? This is my first draft at a comment: It is great to be here with so many leaders in the industry. <nodding to others around th...

Texas is adding state parks faster than I can visit them. Bear Creek:

Not only does the site provide access to the Frio River, Spring Creek, as well as its namesake Bear Creek, but there are several streams, canyons and ridges providing an idyllic background for hikers.

🏕️

There have been a lot of improvements to Micro.blog lately. If you don’t follow the @news account closely, you can also subscribe to the weekly email of posts from news.micro.blog, delivered every Monday.

Dusting off a little of the queued up work for JSON Feed. Thanks to Daniel Pietzsch, the JSON Feed Validator now has a JSON API, if any apps need to hook into validating JSON Feeds using it. There’s a new format=json URL parameter.

Seeing this sign after parking my car… Not so much worried as I am curious how bad it must be to order a custom sign for this.

Ever since first adding book reading to Micro.blog, I’ve tried to avoid recreating my own book database. So we lean on Google Books, Open Library, Goodreads, and even (for a while) paying for metadata. I think that approach has run its course. Going to add our own book curation and cover tools.

I know I go on and on too much about AI sometimes, but… OpenAI’s Codex continues to amaze me. You can have it just watch a GitHub repo for pull requests, then automatically check the new code and add a comment if anything looks wrong. 🤯 Great for a tiny team with not enough eyeballs on things.

As a VC-backed company, perhaps The Browser Company was always going to need to sell. My initial reaction is Atlassian seems a weird fit. But maybe not? From The Verge:

Miller is clear, even forceful, that Dia is not about to become just a wrapper for Atlassian apps, or shift to thinking primarily about IT managers and enterprise features. Dia is still for individual users. It’s just that now, it’s primarily for individual users at work.

I’m really enjoying Dia. I’ll probably keep using it unless or until it gets worse.

Kind of odd if someone has a link to schedule a meeting on their website, but they don’t share an email address or any other contact info. Is email a lost art? It still works. Simple, everywhere.

If you’re on the iOS 26 beta and use Micro.blog, you may want to try switching to our TestFlight beta. It is updated with very minimal Liquid Glass support. Nothing drastic, most of the app is going to stay flat and non-squishy.

Sounds like things are moving forward on a possible Gemini-powered Siri. Mark Gurman reports on an agreement between Google and Apple for testing, and a bunch more about Apple’s plans and new Siri architecture:

Apple is rebuilding Siri around three core components: a planner, the search systems for the web and devices, and a summarizer. The planner interprets voice or text input and decides how to respond; the search system scans the web or user data; and the summarizer pulls it all together into an answer.

Thought about burning an expedited review request with Apple. It has been many years since I asked for one. But also, reviews are consistently 24 hours, so don’t want to risk making things worse than just waiting.

I was pretty excited about the new Epilogue release until we released I had messed up signing in for new users. Bug fix submitted to Apple. 🙁

Oh wow, the smaller reMarkable tablet looks really good. I think they’ve got a winner on their hands. Might still be a bit too pricey for me, but I love the size and design of it.

Epilogue 2.0 for iOS out now. I’ve also submitted the Android version to Google for review, so that won’t be far behind. It seemed right to bump the version to 2.0 because of the new private notes support. The app won’t look very different if you don’t enable notes. With notes enabled on the profile screen, you’ll see a new “Add Note…” button when viewing a book you’re reading. Here you can add a note about your reading progress, and it will be saved to the Micro.blog backend, encrypted along with other Micro.blog notes. I blogged more about book notes in Micro.blog yesterday. Notes are available for all Micro.blog or Micro.blog Premium subscribers, but not Micro.one. If you need to upgrade from Micro.one, click on the Plans page on the web.

Stratechery covers the Google antitrust remedies, largely arguing that the judge did not go far enough. (“This is a waste of time.") I think sharing the search index is a big deal. When OpenAI builds their own search engine and Google drops to second place, we might view this decision differently.

Jeremy Cherfas blogging about the linkrot that will hit when Typepad shuts down:

Links are the foundation that supports the world wide web, and I take them seriously. I correct broken links when I come across them, archive my own stuff (and support archive.org), and generally try to get to the original behind anything I choose to link to.

This is why Micro.blog has a feature to automatically archive everything you link to. (Seriously! Here’s a video explaining it.) But I think Micro.blog is probably too new to contain a large number of Typepad links.

Good morning! Sunrise over MLK Boulevard. Too many trees and power lines in the way, but still was nice to see coming up over the horizon, on the way to coffee. ☕️

I read through a few of the remedy pages in the Google antitrust PDF. My expectations were extremely low for any sort of actual breakup. Sharing search index data with “qualified” competitors seems good, though. Kagi could use that, and of course all the AI companies will want it.

Even though I think AI is incredibly useful and in some cases even empowering, I’m open to the idea that kids should rarely use it, just as they should rarely use social media. When in doubt, limiting use to 18+ is okay. At the very least, I don’t think it’s a simple question.

I don’t really understand Tesla’s latest master plan (post on Twitter / X) and I’m not going to think too hard about it, but the included illustration bugs me, especially the couple walking next to their baby stroller while a robot pushes it. AI is fine. Humanoid robots are a mistake.

Molly White covers recent activity on betting markets:

Prediction market platforms Kalshi and Polymarket opened betting markets on President Donald Trump being “out as president” as social media platforms erupted over the long weekend with rumors that he had died.

What a crazy world we’re living in. Trump survived a bad case of Covid before vaccines. He survived getting shot at. I wouldn’t bet on tomorrow let alone end of the year. 🇺🇸

This is a good change from OpenAI:

We’ll soon begin to route some sensitive conversations—like when our system detects signs of acute distress—to a reasoning model, like GPT‑5-thinking, so it can provide more helpful and beneficial responses, regardless of which model a person first selected.

Essentially, if the model notices the user is having a mental health challenge, it should slow down and use the best reasoning model, just as it would for other hard problems. There will also be new parental controls.

It’s a shame this wasn’t in place earlier. I’m still glad it’s being prioritized now.

A few flowers left at Cherrywood. ☕️

When we added private, end-to-end encrypted notes to Micro.blog, I said that this would be a foundation for new features. The first feature was shared notes. Any private note can be published to your blog with a special, unguessable URL, convenient for sharing with other peo...