Tonight I asked Siri to start a timer. What I actually wanted was a stopwatch. When it asked how long, I said, "Oh wait, I meant stopwatch." It responded, "How long."
That was the whole exchange. Siri heard a correction, ignored it, and continued down the only path it knew. It wasn't confused. It was incapable. And that's a meaningful distinction.
Most voice assistants have spent years getting better at the specific task: hear a command, execute it. Siri is actually quite good at that. Ask it to start a stopwatch, it starts a stopwatch. Ask it to set a timer for ten minutes, it sets a timer for ten minutes. The problem shows up the moment you step outside the clean path. The moment you change your mind mid-sentence, or correct yourself, or treat the conversation like an actual conversation. Siri doesn't have a model for that. It's a transaction processor, and it works great until you don't follow the transaction format.
This matters because the bar has moved.
People who use Claude, ChatGPT, or any of the current conversational AI tools on a regular basis have adjusted their mental model of what a conversation with software can be. You can correct yourself. You can say "actually, I meant something different." You can speak the way you think, which is rarely in clean, well-formed commands. The software figures out what you meant. That's the baseline now, at least for people who've spent time in these tools.
Siri is competing against that standard on every iPhone, and it isn't close.
I recognize I'm further ahead on this than most iPhone users. The frustration I felt tonight is the frustration of someone who lives in these tools, and most people haven't made that shift yet. But they're going to. Slowly and then quickly. And when they do, the gap between what Siri does and what they've come to expect is going to become very visible, on the same device they're holding.
Apple has been public about its Apple Intelligence initiative. The notification summaries, the writing tools, the various on-device features that have shipped over the past year. Some of it has been genuinely useful. Some of it has been embarrassing. What it hasn't been is conversationally capable in the way the rest of the field is moving. Apple knows this. When they launched the iPhone 16, they ran ads showing Siri doing things it couldn't actually do. They ran those ads for months, then quietly pulled them after acknowledging the features wouldn't arrive on the timeline they'd promised.
The optimistic read on Apple's position is that they've been deliberate. They own the silicon, the hardware, the OS, the privacy architecture. When they decide to move, they have real advantages. That's a fair argument and I've heard it made credibly. But the record doesn't support the deliberate framing anymore. The personalized Siri features Apple promised for 2025 didn't ship. The full conversational overhaul may not arrive until 2027. Bloomberg recently reported that Apple is planning to replace the current Siri interface entirely with a new chatbot, which is less a sign of progress than an acknowledgment that what exists can't be repaired incrementally.
Siri is the most visible face of that gap. It's not a back-end model people rarely interact with. It's the voice on your phone, the one hundreds of millions of people ask things every day. If it can't handle "wait, I meant stopwatch" in 2025, and the replacement isn't coming until 2027, the distance between where it is and where users will expect it to be is real and compounding.
Apple will get there eventually. The resources are there, the engineering talent is there, and the platform advantages are real. But 2027 is a long time in this market. The companies setting the conversational standard right now aren't standing still while Apple catches up. And for a product as embedded in daily life as the iPhone, the interface through which most people will ever interact with AI, falling two or three years behind isn't a temporary inconvenience. It's the kind of gap that changes habits.