Why is ChatGPT for Mac So Good?→

Linked By Federico Viticci

Great post by Allen Pike on the importance of a great app experience for modern LLMs, which I recently wrote about. He opens with this line, which is a new axiom I’m going to reuse extensively:

A model is only as useful as its applications.

And on ChatGPT for Mac specifically:

The app does a good job of following the platform conventions on Mac. That means buttons, text fields, and menus behave as they do in other Mac apps. While ChatGPT is imperfect on both Mac and web, both platforms have the finish you would expect from a daily-use tool.

[…]

It’s easier to get a polished app with native APIs, but at a certain scale separate apps make it hard to rapidly iterate a complex enterprise product while keeping it in sync on each platform, while also meeting your service and customer obligations. So for a consumer-facing app like ChatGPT or the no-modifier Copilot, it’s easier to go native. For companies that are, at their core, selling to enterprises, you get Electron apps.

I don’t hate Electron as much as others in our community, but I can’t deny that ChatGPT is one of the nicest AI apps for Mac I’ve used. The other is the recently updated BoltAI. And they’re both native Mac apps.

Permalink

The AI App Experience Matters More Than Benchmarks Now

By Federico Viticci

Different experiences with app connectors in Claude, Perplexity, and ChatGPT.

I was catching up on different articles after the release of Claude Opus 4.5 earlier this week, and this part from Simon Willison’s blog post about it stood out to me:

I’m not saying the new model isn’t an improvement on Sonnet 4.5—but I can’t say with confidence that the challenges I posed it were able to identify a meaningful difference in capabilities between the two.

This represents a growing problem for me. My favorite moments in AI are when a new model gives me the ability to do something that simply wasn’t possible before. In the past these have felt a lot more obvious, but today it’s often very difficult to find concrete examples that differentiate the new generation of models from their predecessors.

This is something that I’ve felt every few weeks (with each new model release from the major AI labs) over the past year: if you’re really plugged into this ecosystem, it can be hard to spot meaningful differences between major models on a release-by-release basis. That’s not to say that real progress in intelligence, knowledge, or tool-calling isn’t being made: benchmarks and evaluations performed by established organizations tell a clear story. At the same time, it’s also worth keeping in mind that more companies these days may be optimizing their models for benchmarks to come out on top and, more importantly, that the vast majority of folks don’t have a suite of personal benchmarks to evaluate different models for their workflows. Simon Willison thinks that people who use AI for work should create personalized test suites, which is something I’m going to consider for prompts that I use frequently. I also feel like Ethan Mollick’s advice of picking a reasoning model and checking in every few months to reassess AI progress is probably the best strategy for most people who don’t want to tweak their AI workflows every other week.

“The iPad’s Software Problem Is Permanent”→

Linked By Federico Viticci

I love my iPad Pro, but, as you know, lately I’ve been wondering about what comes after iPadOS 26. We have much better multitasking now, and key workflow limitations such as file management, audio recording, and long-running background tasks have been addressed by Apple this year. But now that the user-facing system’s foundation has been “fixed”, what about the app ecosystem?

Over at Snazzy Labs, Quinn Nelson has been wondering the same, and I highly recommend watching his video:

Quinn makes a series of strong, cogent arguments with factual evidence that show how, despite multitasking and other iPadOS 26 improvements, using apps on an iPad Pro often falls short of what can be achieved with the same apps on a Mac. There is so much I could quote from this video, but I think his final thought sums it up best:

There are still days that I reach for my $750 MacBook Air because my $2,000 iPad Pro can’t do what I need it to. Seldom is the reverse true.

I’m so happy that Apple seems to be taking iPadOS more seriously than ever this year. But now I can’t help but wonder if the iPad’s problems run deeper than windowing when it comes to getting serious work done on it.

Permalink

I Finally Tested the M5 iPad Pro’s Neural-Accelerated AI, and the Hype Is Real

By Federico Viticci

The M5 iPad Pro.

The best kind of follow-up article isn’t one that clarifies a topic that someone got wrong (although I do love that, especially when that “someone” isn’t me); it’s one that provides more context to a story that was incomplete. My M5 iPad Pro review was an incomplete narrative. As you may recall, I was unable to test Apple’s promised claims of 3.5× improvements for local AI processing thanks to the new Neural Accelerators built into the M5’s GPU. It’s not that I didn’t believe Apple’s numbers. I simply couldn’t test them myself due to the early nature of the software and the timing of my embargo.

Well, I was finally able to test local AI performance with a pre-release version of MLX optimized for M5, and let me tell you: not only is the hype real, but the numbers I got from my extensive tests over the past two weeks actually exceed Apple’s claims.

iPadOS 26.2 Beta Restores Drag and Drop Gestures for Split View and Slide Over →

Linked By Federico Viticci

Following the comeback of Slide Over in iPadOS 26.1, Apple is continuing to iterate on iPadOS 26 multitasking by restoring functionalities that had been removed from the launch version of iPadOS 26.0 in September. Yesterday, in the third developer beta of iPadOS 26.2, the company brought back drag and drop gestures to put app windows directly in Split View and Slide Over without having to interact with additional menus. To understand how these old gestures work in the context of iPadOS 26, I recommend watching this video by Chris Lawley:

As you can see, the gestures are pretty much the same ones as iPadOS 18, but the interaction is slightly different insofar as the “pull indicator” for Slide Over (re-introduced in iPadOS 26.1) now serves two purposes. That indicator now acts both as a signal that you can drop a window to instantly tile it as one half of a Split View, and it’s also a drop target to enter Slide Over right away. The design is clever, if maybe a little too hard to discover…but that’s always been the case with multitasking gestures that aren’t exposed by a menu – which is exactly why Apple is now offering plenty of options in iPadOS 26 to discover different multitasking features in different menus.

I’m glad to see Apple quickly iterate on iPadOS 26 by finding ways to blend the old multitasking system with the platform’s new windowing engine. Based on the comments I received after publishing my iPadOS 26 review, enough people were missing the simplicity of Split View and Slide Over that I think Apple’s doing the right thing in making all these multitasking systems coexist with one another.

As I argued on last week’s episode of Connected, and as Myke and Jason also elaborated on this week’s episode of Upgrade, the problem with the iPad Pro now is that we have a great foundation with iPadOS 26 and very few third-party apps that take advantage of it beyond the usual names. I suspected as much months ago, when I explained why, in a world dominated by web apps, the iPad’s next problem was going to be its app ecosystem. The web services I use on a daily basis (Slack, Notion, Claude, Superhuman, Todoist – the list goes on) simply don’t make iPad apps of the same caliber as their desktop/web counterparts. So I find myself using Safari on the iPad to get my work done these days, but, for a variety of reasons and dozens of small papercuts, Safari for iPad simply isn’t as good as Safari on the Mac.

Given how the third-party app ecosystem story for iPad is outside of Apple’s control and how most companies aren’t incentivized to make excellent native iPad apps anymore, now that multitasking has been largely “fixed” in iPadOS 26.2, I hope Apple turns its attention to something they can control: making Safari for iPad truly desktop-class and not a baby version of Safari for Mac.

Permalink

Trying to Make Sense of the Rumored, Gemini-Powered Siri Overhaul

By Federico Viticci

Quite the scoop from Mark Gurman yesterday on what Apple is planning for major Siri improvements in 2026:

Apple Inc. is planning to pay about $1 billion a year for an ultrapowerful 1.2 trillion parameter artificial intelligence model developed by Alphabet Inc.’s Google that would help run its long-promised overhaul of the Siri voice assistant, according to people with knowledge of the matter.

There is a lot to unpack here and I have a lot of questions.

On MiniMax M2 and LLMs with Interleaved Thinking Steps

By Federico Viticci

MiniMax M2 with interleaved thinking steps and tools in TypingMind.

In addition to Kimi K2 (which I recently wrote about here) and GLM-4.6 (which will become an option on Cerebras in a few days, when I’ll play around with it), one of the more interesting open-source LLM releases out of China lately is MiniMax M2. This MoE model (230B parameters, 10B activated at any given time) claims to reach 90% of the performance of Sonnet 4.5…at 8% the cost. You can read more about the model here; Simon Willison blogged about it here; you can also test it with MLX on an Apple silicon Mac.

What I find especially interesting about M2 is that it’s the first model to support interleaved thinking steps in between responses and tool calls, which is something that Anthropic pioneered with Claude Sonnet 4 back in May. Here’s Skyler Miao, head of engineering at MiniMax, in a post on X (unfortunately, most of the open-source AI community is only active there):

As we work more closely with partners, we’ve been surprised how poorly community support interleaved thinking, which is crucial for long, complex agentic tasks. Sonnet 4 introduced it 5 months ago, but adoption is still limited.

We think it’s one of the most important features for agentic models: it makes great use of test-time compute.

The model can reason after each tool call, especially when tool outputs are unexpected. That’s often the hardest part of agentic jobs: you can’t predict what the env returns. With interleaved thinking, the model could reason after get tool outputs, and try to find out a better solution.

We’re now working with partners to enable interleaved thinking in M2 — and hopefully across all capable models.

I’ve been using Claude as my main “production” LLM for the past few months and, as I’ve shared before, I consider the fact that both Sonnet and Haiku think between steps an essential aspect of their agentic nature and integration with third-party apps.

That being said, I have been testing MiniMax M2 on TypingMind in addition to Kimi K2 for the past week and it is, indeed, impressive. I plugged MiniMax M2 into TypingMind using their Anthropic-compatible endpoint; out of the box, the model worked with interleaved thinking and the several plugins I’ve built for myself in TypingMind using Claude. I haven’t used M2 for any vibe-coding tasks yet, but for other research or tool-based queries (like adding notes to Notion and tasks to Todoist), M2 effectively felt like a version of Sonnet not made by Anthropic.

Right now, MiniMax M2 isn’t hosted on any of the fast inference providers; I’ve accessed it via the official MiniMax API endpoint, whose inference speed isn’t that different from Anthropic’s cloud. The possibility of MiniMax M2 on Cerebras or Groq is extremely fascinating, and I hope it’s in the cards for the near future.

AI Experiments: Fast Inference with Groq and Third-Party Tools with Kimi K2 in TypingMind

By Federico Viticci

Kimi K2, hosted on Groq, running in TypingMind with a custom plugin I made.

I’ll talk about this more in depth in Monday’s episode of AppStories (if you’re a Plus subscriber, it’ll be out on Sunday), but I wanted to post a quick note on the site to show off what I’ve been experimenting with this week. I started playing around with TypingMind, a web-based wrapper for all kinds of LLMs (from any provider you want to use), and, in the process, I’ve ended up recreating parts of my Claude setup with third-party apps…at a much, much higher speed. Here, let me show you with a video:

Kimi K2 hosted on Groq on the left.Replay

Max Weinbach on the M5’s Neural Accelerators →

Linked By Federico Viticci

In addition to the M5 iPad Pro, which I reviewed earlier today, I also received an M5 MacBook Pro review unit from Apple last week. I really wanted to write a companion piece to my iPad Pro story about MLX and the M5’s Neural Accelerators; sadly, I couldn’t get the latest MLX branch to work on the MacBook Pro either.

However, Max Weinbach at Creative Strategies did, and shared some impressive results with the M5 and its GPU’s Neural Accelerators:

These dedicated neural accelerators in each core lead to that 4x speedup of compute! In compute heavy parts of LLMs, like the pre-fill stage (the processing that happens during the time to first token) this should lead to massive speed-ups in performance! The decode, generating each token, should be accelerated by the memory bandwidth improvements of the SoC.

Now, I would have loved to show this off! Unfortunately, full support for the Neural Accelerators isn’t in MLX yet. There is preliminary support, though! There will be an update later this year with full support, but that doesn’t mean we can’t test now! Unfortunately, I don’t have an M4 Mac on me (traveling at the moment) but what I was able to do was compare M5 performance before and after tensor core optimization! We’re seeing between a 3x and 4x speedup in prefill performance!

Looking at Max’s benchmarks with Qwen3 8B and a ~20,000-token prompt, there is indeed a 3.65x speedup in tokens/sec in the prefill stage – jumping from 158.2 tok/s to a remarkable 578.7 tok/s. This is why I’m very excited about the future of MLX for local inference on M5, and why I’m also looking forward to M5 Pro/M5 Max chipsets in future Mac models.

Permalink