After WWDC is always a good time to reflect on priorities. This year’s event for me was about refinements and focus.

On the technical side, I’m hoping to use Apple Foundation Models as a fallback when bigger models can’t be used. But since my Mac app is outside the App Store, I’ll be limited too, apparently without private cloud compute.

Congrats to Apple on a good week. The reaction I heard from folks seemed overwhelmingly positive. Onward! There’s a lot to work on.

Thinking about a point raised by Joanna Stern and Nilay Patel at The Talk Show last night, that regular people may not totally get the privacy of Apple’s private cloud compute. Will they trust it more than giving personal context to ChatGPT? All these years of Apple ads about privacy should help.

Excellent Stratechery update this morning about Fable and more broadly how Anthropic’s “only we can be trusted” messaging also helps their business:

It’s fascinating to observe: me, the rational business analyst, sees a hard-nosed but understandable business decision to cut off would-be competitors; Anthropic employees and advocates, the true believers, see a regrettable but understandable safety decision that ensures that responsible and thoughtful people — themselves, of course — will be the ones guiding our AGI future.

Bustling downtown San Jose.

OpenAI published an article about the company’s high-level plan for AI, written by Sam Altman and Jakub Pachocki. There are themes in it that will be familiar for anyone following OpenAI closely this year. This passage feels a little new:

Entirely automating everything is not the future we want. It would be unfulfilling, and it would be dangerous. AI should help people pursue their goals, not become untethered from them.

The Talk Show Live is tonight. I’ll be there. Not too late for tickets if you’re in the area. John Gruber blogs:

If you can make it in person, you should come. The California Theater is a beautiful big theater and tickets are still available.

Catching up on more of Apple’s new AI architecture. Finally have some clarity that the sort of default Apple Foundation Models will run on Apple servers. The most capable “Pro” model will run on Nvidia chips in Google Cloud. Seems like a reasonable way to split things up.

Myke Hurley blogged a very optimistic take about Siri AI:

I fully understand I may be in that WWDC Glow right now when it comes to Siri AI. I want to preface this before I say what I am about to say, which is that I imagine that after Siri AI ships later in the year, I will not have much need for any of the general-purpose chatbots like Gemini, ChatGPT, Claude, or others.

I don’t see this at all. But everyone is different, and Siri AI should be fairly popular.

I’ve posted an update to the Micro.blog 4.0 beta for Mac. You can download the new version here. There’s no “Check for Updates” for the beta. It improves bookmarking and also adds collecting anonymous machine stats for the first time, so we can have a better sense of what Macs people are using.

Haiku, Sonnet, Opus, Mythos, Fable. I’m not a fan of Anthropic’s business but I’m a fan of their names. So nice.