Back in 2017, the night before the WWDC keynote, I wrote a short post about table stakes. I’d spent a year living on the iPad, and a lot of it still felt like swimming upstream. My argument was that fixing the iPad’s file management and multitasking wasn’t the impressive part of the story. It was the price of admission. As I put it back then, that’s the starting point, not the ending one.

In a few days, we’ll get another WWDC keynote, and I keep coming back to that phrase, table stakes. This time for Apple Intelligence.

Apple has spent two years telling us what their AI is going to be. Some of it shipped. A lot of it didn’t. So heading into June, I find I’m less interested in the next round of promises and more interested in the floor. What does Apple Intelligence have to get right just to be in the game? I keep landing on three things.

The first is that it has to be stable and reliable. We already lived through the cautionary tale here. Siri started with a good idea and a lot of goodwill, and then it spent a decade being the thing you stopped reaching for because you couldn’t trust it. An assistant you can’t trust stops being an assistant. You just quit using it. If Apple Intelligence is going to ask for a place in how I work, it has to be the kind of dependable I can build a habit on.

The second is that it has to actually understand what I say and act on what I mean. Not just transcribe my words. Understand my intent and then go do the thing. There’s a big gap between hearing the request and finishing the job, and almost everything useful lives on the far side of that gap. If I have to rephrase myself three times and then go finish the task by hand anyway, the magic is gone.

The third is the one I care about most, and it’s the one I’ve been talking about constantly in the Robot Assistant Field Guide meetups. Memory and context. I’ve come to believe that context matters more than the brain. The reason an Apple assistant could be compelling has nothing to do with raw smarts. It’s that Apple is sitting on the complete picture of your life already. Your mail, your messages, your calendar, your photos, your location, the whole thing, right there on the device. Nobody else has that. If Apple delivers on it, they win a game the other players can’t even enter.

But context only counts if the assistant can use it. It’s not enough to hold all that information. The harness around the model has to be good enough to reach into that context and actually take action with it. That’s the whole ballgame. The data without the ability to act on it is just a very private filing cabinet.

So that’s my floor. Reliable, understands what I mean, and knows me well enough to help. None of it is flashy. All of it has to be there before any of the flashy stuff means a thing.

I’m still hopeful. Apple is in a unique position. They have the data and the users to make some really transformational technology. The question is whether they have the necessary skills and resolve to reach for it. We’ll know a lot more on Monday.