Minor bummer that I wasn’t expecting when doing more local LLM work in my app: the mlx-swift package adds quite a bit to application size on disk. Micro.blog for Mac jumps from only 10 MB to about 50 MB. Still relatively small for 2026, though.

Reviewed my ChatGPT memories after OpenAI updated dreaming today:

In contrast to saved memories, dreaming leverages a background process that allows ChatGPT to learn from many conversations and synthesize ChatGPT’s memory state in order to always provide the freshest, most relevant context to your conversations. Dreaming also makes it easier for memory to include context that occurs naturally in conversation, without relying on explicit requests to remember something.

Claude has dreams too. Codex and OpenClaw use lots of Markdown files to update memory, which makes it fairly transparent.

Amazing idea: Cash App launches a wand for tap-and-pay. 🪄

The day after a Spurs loss, expect all of my blog posts to have a tinge of anger to them. 🤪

Apple has published another one of their reports about all the money that is even loosely related to the App Store. It’s misleading and maybe even dishonest:

Apple today announced the global App Store ecosystem facilitated over $1.4 trillion in developer billings and sales in 2025, according to a new study by economists at Analysis Group. For more than 90 percent of the billings and sales facilitated by the App Store ecosystem, developers did not pay any commission to Apple.

Is my Micro.blog revenue “facilitated” by the App Store? Ridiculous. App Store policy is a sunk boat.

Jason Snell has developed a tool for podcasters called Double Ender. He blogs about what’s possible now without coding:

Whatever you call it, whether it’s being a producer or product manager or something else that isn’t a programmer, creating good software in the AI era still requires the power of a human brain: being creative, solving problems, and making decisions.

I also agree with Jason that Apple should be rethinking Xcode for how developers actually work. I still use Xcode but I’m using significantly less of it. A streamlined Xcode with half the features would pair well with Codex.

Argh. Great game but that loss hurts. Feel like that offensive rebound by the Knicks with a couple minutes left just flipped it, otherwise that might be a Spurs win. Hope we can bounce back in game 2. 🏀

Counting down the hours… NBA finals, game 1. I really don’t know what to expect. 🏀

David Smith blogs about his Apple-inspired ideals for development and design: If you are always striving towards improving quality you will eventually end up with a surplus of user expectation. In my experience this surplus is incredibly precarious because once you are bett...

This interview with Sarah Friar is really good. OpenAI is juggling so many things, I get the impression that it would all fall apart if she wasn’t CFO. She says the new device will be announced later this year:

What Jony and team are really good at is bringing humanity to devices. And I don’t really know how to explain that well, but when you see it, you feel it. […] It feels very natural. But it feels very lovable.

Satya Nadella in the closing to Build:

There are really two stories people can tell about this moment. One is that technology concentrates power, reduces human agency, and leaves to society to absorb the consequences. The other is that we use this next wave to unlock opportunity for developers, scientists, enterprises, and every community. And our job is to make the second story true.

Always interesting when people in power talk about not further concentrating power. I believe he’s sincere, though. I worry more when those with total control over a platform (see: Apple) don’t acknowledge it.

Strong winds at the front of this storm. Lost a branch of the pecan tree. Looks like minimal damage to the roof, mostly just the gutter. Whew. 🌳

Microsoft announced Project Solara at Build today. They’ve got a couple prototypes, including a clever badge-style device, and thoughts on UI:

We are also investing in just-in-time UI: the ability for an agent experience to adapt across devices and modalities without requiring developers to redesign everything for every new form factor. Today, that means semi-structured approaches like adaptive cards and known content types. Over time, it moves toward more dynamic and generative interfaces.

I enjoyed this blog post by Om Malik about how AI models will become invisible plumbing. Nice history lesson with fiber-optic cables too: A single strand of glass can carry only so much data at one wavelength – think of it as one lane of a highway. Wavelength Division Multi...

Following a report from Bloomberg on Uber, TechCrunch has a post about companies putting a cap on AI spending:

AI is getting expensive, and some companies are cutting back on usage in an attempt to moderate costs. That cohort includes Uber, which recently instituted internal usage caps as a way to cut down on its exorbitant AI spend.

I’m sure this is happening all over. That’s why I commented in my post yesterday that Anthropic’s revenue is not stable. It’s hard to cancel a subscription you value. It’s much easier to cut back on tokens.

🚂

Working more on Micro.blog 4.0 for Mac. Got a chance to rewrite all the old sidebar code, moving it from a very cluttered XIB to Obj-C code. Now feels more at home with other modern Mac apps.

Getting almost too late to turn this narrative around, but OpenAI tries to make the case for data centers in a new video about Abilene.

Demolition continues on 51st street.

Kath Korevec, who works at OpenAI, in a post on X:

Devs do a lot more than “just coding.” They write docs, design apps, design architecture, manage devops, work with data, etc. It’s no wonder codex is so good at “all the things.” It’s because we built it for developers.

Steve Troughton-Smith in a Mastodon post:

The best kind of WWDC is the WWDC that gives me a big grab-bag of tools to make my apps better along several axes, and doesn’t burden me with a huge amount of needless churn just to tread water.

With Nvidia announcing the RTX Spark today, I revisited my quick comment from last year about the DGX Spark. It appears the price went from $4k to $4700 since then. Mac Studio price hasn’t changed.

No pricing yet for the RTX. It has better specs, but you’ve gotta run Linux or Windows.

Of course hardly anyone needs a computer with 128 GB of RAM. I’m only interested in following this because some new things are possible when users have (let’s say) 24 GB. With memory so expensive right now, and cloud-based AI continuing to outpace everything else, local AI models will be somewhat limited. I think more apps could adopt a hybrid approach of using both cloud AI and local when possible.

A year ago I experimented with running local AI models inside Micro.blog for Mac. I did a bunch of work on this and then threw all of the code out. It wasn’t good enough.

I’ve now come back to it. With data center backlash, I think people want options that reduce our dependence on cloud models.

It’s wild that the corgi correctly predicted every game in the western conference finals. That’s a good dog. 🏀

Anthropic submits draft to go public:

Today, Anthropic, PBC confidentially submitted a draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission for a proposed initial public offering of our common stock. This gives us the option to go public after the SEC completes its review. The proposed initial public offering will depend on market conditions and other factors.

Something I learned over the weekend: Anthropic’s run-rate projections count token usage, not just subscriptions. Their revenue is probably not as stable as a normal SaaS company.

Reflecting on how WWDC has changed, I still like my blog post from 2014. I could update it for the modern era, but I’m not inspired to do so. In a way, the best of WWDC will always be captured in those great moments from a decade ago, or even farther back. Perhaps it’s part of getting older.

New Kickstarter campaign from Jason Snell and Myke Hurley to fund a podcast series covering Apple history:

Designed in California will consist of at least 30 episodes dropping over 12 months, looking at a wide range of Apple history, drawing from the best sources available.

It has been a while since we bumped the system requirements for Micro.blog for Mac. Current version runs on macOS 11 Big Sur. This feels like a good stopping point. The next version (4.0) will require macOS 14 Sonoma, so it’s more modern but still fine for the Tahoe haters.

Spider-Noir is off to a great start. I didn’t realize until we queued it up over the weekend that it was a TV series, not a movie. Watched most of it in black and white. 📺

Can’t believe WWDC is next week. Obviously the new Siri will the headline, but hope we get some other tweaks to Liquid Glass and general robustness.

Apple will improve Xcode too. Many developers barely touch Xcode now, though, so curious how Apple will keep it relevant, or if they even need to.