Hot take: new MacBook Pro with XDR looks good! I expected much worse on pricing. Much better display for +$400 not bad when you’re already in for $7,800! I personally don’t care about touch and I think it’s a dumb feature. But the display specs beyond touch are worth $400. Also the normal m5 laptop prices are untouched, so that’s good! I’d get the 4TB storage, 128GB memory with nano-texture if you’re an AI pro.

At a small @DeanOrnishMD event: “I think of modern medicine as a bunch of doctors around a sink with a leak and they are just constantly mopping up the water. Why don’t we stop the leak?”

Big news today from Apple: Keynote is getting an open, JSON-based file format (.keyx), plus a Python library to create Keynote files on any platform. Pages, Numbers and Keynote files today are closed and hard to modify unless you use AppleScript (limited and old) or Shortcuts (a huge pain), so you can’t really use AI to make documents in those formats. This was a HUGE pain inside Apple as well as outside, and most tools worked around it by making a PowerPoint file and importing it into Keynote, which wasn’t perfect. Creator Studio subscribers on the Mac also get MCP support in Keynote, Pages and Numbers, so you can bring your own AI agent to read, edit and automate all three. Both the new format and MCP support arrive in November. Bringing an AI-friendly format to Keynote (and I assume Pages and Numbers soon after) is a big positive change.

Apple announced the M6 14-inch MacBook Pro this morning, and I see the Mac laptop lineup is confusing to a lot of you. Here’s my quick guide, with an AI take on each. MacBook Neo ($699): a phone chip (A18 Pro) in a great entry-level laptop for couch surfing. For AI you get the built-in Apple Intelligence stuff like transcription, translation and proofreading, and cloud models. Only 8GB of memory, so no third-party on-device models on this one. MacBook Air with M5 ($1,299, up to 32GB): enough for something like to sift your email, run projects and track your calendar. Not enough for good content generation, so it won’t write your emails, fill out your spreadsheet or write code effectively. The catch: no fan, so push a local model hard and it throttles. 14-inch MacBook Pro with M6 ($1,999): the entry-level Pro, and it has a fan. At Apple we saw two core buyers. Students going into computer science (“Mom, the Air is for English majors”) whose parents don’t want to spend a lot. And companies, because if you hand a programmer an Air, they say “I’m a pro.” CIOs told me they loved putting a base Pro on a developer’s desk for a lot less than the M5 Pro and Max models. Its 32GB runs utility models, basic content work, light coding assistants and simple image gen. Apple isn’t lying when it says on-device AI, but it’s limited. Think of it as about what fits on a single 24GB 3090. M5 Pro (up to 64GB) and M5 Max (up to 128GB): if you’re serious about AI, I recommend the M5 Max with 128GB, nano-texture, upgraded processor and 4TB for $7,849. It’s an amazing portable machine that runs real coding and content models, light image gen, and even light video gen if you let it run in batches. As I said last week, put it in your cart TODAY, in case Apple pulls pricing shenanigans when the rumored M5 Pro/Max touchscreen MacBook Pros arrive tomorrow. Rumors say the current models stay, hopefully at the same price, but you never know. I’m not a fan of a thinner touchscreen Mac. Unless it goes to something crazy like 256GB memory, the 128GB M5 Max MacBook Pro stays my pick for serious portable AI.

My quick takeaways on the Anthropic S-1: 1. This is a historical moment 2. $6.6M in revenue per employee!! Hard to even comprehend how good that is but Anthropic is AI native as it gets so makes sense. 3. I always thought OpenAI would drop their S-1 first 4. Customers spend about 9x more after a year. Net Revenue Retention is approximately 900% as of Sep 30, 2026. Good SaaS companies brag about 120%. 5. Cowork is growing faster than Claude Code did. Goes to show you the demand for consumer products. Duo will probably rip. 6. 3 years ago this company did $41M in revenue. In the first six months of 2026, it did $16.3 billion. I don't think anyone has a mental model for that curve yet, I sure don’t. 7. Nearly half of revenue comes through cloud marketplaces. About 48% comes from customers buying through them. Big companies buy where their budget already lives, and the lesson for founders selling to enterprises is to get listed where your buyers already spend. 8. Their run rate formula is published: the last 28 days of usage times 13, plus subscriptions times 12, which i guess is the new ARR math for founders? 9. It looks like They're merging Cowork into Claude dot au which tells you the chat box era is ending and the "go do the work for me" era is here. 10. Thousands of Anthropic employees are about to become liquid at once, so expect the biggest wave of new AI startups and angel checks San Francisco has ever seen. Reminds me when Facebook went public but this is way bigger. Let’s go! 11. Buried in the filing, Anthropic says the gap between what Claude can do and how widely it's used is one of its biggest growth opportunities, so the company itself is telling you the models are ahead of the market and someone has to close that gap for dentists, landlords, and plumbers. 12. Whatever happens, exciting times to be building in AI.

Here are my priors for the next few months: 1. Cybersecurity will push the big labs back into hyper-growth mode, much like the coding agent wave did. 2. More GPUs will shift from training to inference. 3. Open-weight model prices will drop. This may push closed labs to lower their prices too. 4. GPU rental prices will keep rising. The middle of the market will get squeezed, even as the pie grows larger. You can read more about it below. Feedback is greatly welcomed.

Tonight I’m at AiOS SF talking AI on Apple devices, but... one more thing 😀 Years ago at Apple I set up a demo for a senior government official: real-time translation software for service members, running on an iPod touch strapped to a forearm. Before it started, I was alone in the room with Tim Cook. I don’t think he had ever said my name before. “So Todd, what are we showing today?” I WAS NOT READY FOR THAT. I told him all about the translation software. There was a pause that felt terrifying. Then: “That’s good, but remember to always show and explain the Apple product first.” The translation app was the cool part, but Apple sold iPod touch. That’s what I was there for. I never forgot that. I see the same thing in AI now. Companies talk about the model they’re using, and tooling, and hardware, and the customer is left wondering what the product actually does for them. That’s where I fit. After thousands of presentations to executives during 22 years at Apple, and after helping launch Apple Intelligence and AI messaging, I’m starting @DaywardAI to help companies show and explain their AI, and build things with it. A lot of it runs on hardware they own. Turning the technical into the understandable is my superpower. I’m already working with a few companies, private individuals, and artists to create amazing results from unreasonable requests, and I’m having a blast doing it! We are so so early on AI, the next few years are going to be INCREDIBLE.

I missed in the Microsoft announcement yesterday that the Microsoft Surface RTX dev box lacks the connectx-7 network interface you'll find in pretty much every other DGX spark box. This limits you to a single dev box, since clustering over 10Gb ethernet is 20x slower, less efficient, and not really practical. Given the number of great 2, 3, and even 4x spark recipes from @MiaAI_lab and others this is a pretty big downgrade that limits future expandability.

Agree with this @appleinsider article: companies aren't quitting Macs because Apple Intelligence ships installed by default. 1. Many large companies have already audited and approved it. 2. The rest have enough control through MDM and third-party tools to shut off what they don't want. Apple has no single off switch, but the granular controls satisfy most corporate needs. 3. The bigger issue isn't on-device AI. It's Private Cloud Compute. Plenty of large companies want none of their data going to an unmanaged cloud they don't have a contract with, however secure it may be. Apple also doesn't offer what many companies expect from an AI cloud, like a HIPAA BAA or FedRAMP authorization. You can't shut off PCC on its own, so if you want none of your data going to PCC, you have to shut off every feature that uses it. Apple doesn't document which features those are, but there's plenty of unofficial documentation out there. So companies that haven't approved it turn off Apple Intelligence features entirely. Most of my CIO-level conversations at Apple were about AI spend, stopping employees from shipping AI slop, and having observability into what AI was being used for. Many companies also have a public partnership with Microsoft, Anthropic or OpenAI and they want *all* AI to go through that partner (and through IT), not Apple.

Tony Fadell on on-device AI in @TechCrunch “Trust and safety are going to be paramount with anything that we turn over to some kind of intelligence,” Fadell said. “The only company I could see right now — maybe one other — is Apple who could do this. Apple has all the hardware, they have all the chips, they have all the pieces, but they don’t have all the AI stuff.” “When you’re hearing all of these things about, ‘Oh, data centers are going to conquer the world,’ … I don’t believe it,” he said. “I’ve seen this before. I’ve seen this play out in the internet days and what have you. We have so much power on the device now, and even more compute, and they’re still battery operated.” I'm not convinced Apple is set up to execute well on agentic AI, but I'm optimistic that Ternus is going to fix software engineering. The whole article is worth a read, @tfadell is a legend, of course.

I found Microsoft's announcements today to be pretty meh. They copied Spotlight (who copied Raycast) with a bunch of quick action features that basically nobody will use. Surface Laptop Ultra is a dud for AI. 1TB storage only on the 128GB preorder. You can't even config an M5 Max MacBook Pro with that little. And no options for > 2TB? This is for AI devs? No, no it is not. Price roughly the same as the MacBook Pro M5. $5,899 for the Surface Laptop Ultra with 128GB and 1TB, $6,699 for the MacBook Pro with 128GB and twice the storage at 2TB. You might be able to swap in a 4TB SSD in to the Ultra, but it looks like it might be a tiny 2230 drive and consumers can't get 4TB in that size. One advantage you get over macOS is full CUDA, but most CUDA devs would likely prefer to run a Linux system natively, not with all of Windows along with it. Surface RTX Spark Dev Box will never sell in volume. $5,999 for a 128GB Spark desktop is about $1,000 cheaper than NVIDIA current prices. But it runs Windows, so almost no one cares. The whole point of Spark is to run the same platform you're deploying to, nobody deploys AI to Windows servers. Sure, CUDA runs in WSL, but that's Linux inside Windows, so expect complexity and compatibility issues. I guess enterprise shops that run Windows might be interested, but they are famously cost-conscious and this is an expensive box. $5,999 is a lot of tokens. Shower thought: is Microsoft saying cloud AI is insecure, not private, and expensive? How are they threading *that* needle with their enterprise customers? 🤔 It is good validation for the value of on-device AI, and "hybrid intelligence" is good marketing from Nadella. Apple has been completely, totally silent on on-device AI for business since one private business event back in June. (Right around when I left Apple and stopped working on AI and business!) 😀