This Is the Year Apple Has to Deliver on Siri

Mark Gurman delivered more Siri news this week, and I’m left with the same feeling I’ve had for over a year now: equal parts hope and frustration.

Here’s the picture as it currently stands. Apple is planning two separate Siri overhauls, releasing months apart.

The Spring Update: iOS 26.4

The first update arrives with iOS 26.4, expected around March or April. This is the non-chatbot version built on a custom Google Gemini model running on Apple’s Private Cloud Compute servers. The goal here seems straightforward: finally cash all those checks Apple wrote at WWDC 2024.

Remember those promises? Siri that understands personal context. Siri that can find the book recommendation your mom texted you. Siri that works across apps instead of being confined to one at a time. Features that were supposed to ship with iOS 18, then got pushed to “later,” then pushed again to 2026.

The Fall Overhaul: iOS 27

Then, just a few months later at WWDC 2026, Apple plans to announce an entirely different approach. This one is codenamed “Campos,” and it’s a full chatbot experience. Think Claude or ChatGPT, but baked directly into your iPhone, iPad, and Mac. Voice and text inputs. Persistent conversations you can return to. The works.

Craig Federighi has previously expressed skepticism about chatbot interfaces, preferring AI that’s “integrated into everything you do, not a bolt-on chatbot on the side.” But competitive pressure from OpenAI and others seems to have changed the calculus.

Why Two Versions?

This is where I get frustrated. Releasing two fundamentally different versions of Siri months apart doesn’t inspire confidence. The first version sounds like something they cobbled together just to say they kept their promises. I’d almost prefer they skipped it entirely and focused all their energy on the chatbot.

Why This Matters

I’ve been critical of Siri over the last decade. Every year Apple makes promises it can’t keep. Every WWDC brings demos of features that arrive late, broken, or not at all.

And yet I continue to believe that a smart model on our Apple devices, with access to our local data, where everything stays local and private or runs through Private Cloud Compute, could be one of the best implementations of AI we’ve seen.

Think about who this could help. I spent 30 years practicing law. I know firsthand how many professionals are locked out of these AI tools because the privacy story isn’t good enough. Apple could change that.

And for the rest of us? We’re not particularly excited about sharing our personal information with giant AI companies either. A truly private assistant that actually knows your life without selling it to advertisers? That’s the dream.

Apple is uniquely positioned to deliver this. They have the hardware. They have the ecosystem integration. They have the privacy infrastructure. They have over 2 billion devices that could benefit.

But they have yet to prove they can actually ship it.

Where I Am Right Now

I currently use Siri where I can, but that’s very limited. I get far more use out of Claude than I do Siri at this point. (Claude’s recent Cowork feature is shockingly impressive.) That’s not where I want to be. I want the assistant built into my devices to be the one I reach for first.

The Bottom Line

All of this feels like it’s coming to a boiling point. We’ve all been patient with Apple for years now. It’s time for them to prove whether or not they can pull this off.

Let’s hope that in six months, Apple has finally answered the call.

The Lab Report for January 23, 2026

In this week’s episode: A few quick Apple updates, AirTag is a great addition to your luggage, some headlines from the rumor mill (iPhone and MacBook Pro stuff), and I share my handy Apple Mail “Send From” Labs video. Plus, a Joshua Redman track to take you into the weekend.
… This is a post for MacSparky Labs Members only. Care to join? If you’re already a member, you can log in here.

The M5 Pro and Max Are Going to Be Monsters for Local AI

Back in November, Apple quietly published a research article about the Neural Accelerators in the M5 chip. The numbers are wild.

The base M5 MacBook Pro already delivers up to 4x faster time-to-first-token compared to the M4 when running large language models through MLX. Image generation with FLUX is 3.8x faster. This is on the base chip with 24GB of unified memory.

Think about what happens when the M5 Pro and M5 Max show up with more memory bandwidth and more Neural Accelerators. And eventually the M5 Ultra in the Mac Studio.

Right now, people serious about running local AI often look at expensive PC builds with dedicated GPUs. The M5 generation might change that math entirely. A well-configured M5 Max MacBook Pro or Mac Studio could become the machine for people who want to run models locally, privately, on their own hardware.

Apple’s unified memory architecture was always a theoretical advantage for AI workloads. With the M5’s Neural Accelerators, that advantage is becoming very real. If you’re interested in local AI and you’re on an M3 or earlier, I’d wait for these announcements before buying anything.

File Renaming with Claude Cowork

Anthropic has recently released a Claude update that lets it do actual work for you in the background. In this example I have it take a folder full of scans and rename them with the internal date and document description based on Claude’s review of the file content.
… This is a post for MacSparky Labs Members only. Care to join? If you’re already a member, you can log in here.

Three Weeks with Apple Wallet Car Key

Three weeks ago, Rivian started supporting Apple Wallet car keys. I set it up immediately. (I made a video on it in the MacSparky Labs)

Before this, unlocking my car required pulling out my phone and waiting for a Bluetooth connection. It worked, but it wasn’t great. Sometimes I’d be standing at the door like an idiot while my phone figured things out.

One out of fifteen times, it just didn’t work, and I had to pull out the credit-card style key from my wallet and tap it against the door.

The Wallet car key is different. I walk up, and when I’m about five feet away, the car unlocks. Every damn time. Three weeks in, it hasn’t failed once.

I still carry the credit card backup key in my wallet. Old habits. But if this streak continues for another few weeks, that backup is coming out. One less thing to carry.

I hope this goes wide. Every car manufacturer should be on board with this.

OmniOutliner 6 Is Here

I think I first bought OmniOutliner at version 2. I remember it came in shrink wrap at a computer store.

Regardless, I’ve been using the application for a long time. I’ve used it to outline countless contracts, depositions, trial examinations, MacSparky videos, books, and podcasts ever since. In my opinion, it is the best-in-class outliner for the Mac.

And now version 6 just dropped.

OmniOutliner 6 brings some welcome changes. The interface got a modernization pass, with Liquid Glass design elements that look great on current Apple hardware. There are also new Dynamic Themes that adjust automatically based on your system settings.

The big news for me is Omni Links. These let you create connections between documents and specific rows within outlines. If you’re the kind of person who builds complex, interconnected projects (and if you’re using OmniOutliner seriously, you probably are), this is going to matter.

The Omni Group also added support for querying Apple Intelligence through their Omni Automation system. I’m still exploring what this means in practice, but the ability to bring AI queries into your outlining workflow has potential.

For the first time, OmniOutliner is now truly universal across Mac, iPad, iPhone, and Apple Vision Pro. Same features everywhere. That’s been a long time coming.

If you’ve been on the fence about OmniOutliner, or if you’ve let your outlining practice slip, version 6 is a good reason to take another look. The Omni Group continues to make serious software for people who want to think clearly and organize their ideas.

You can learn more at omnigroup.com/OmniOutliner.