Apple finally fixed Siri. So why does it feel anticlimactic?

Apple finally shipped the Siri it promised. The assistant understands your photos, emails, calendar, and screenshots. It finds the receipt you forgot you saved. It reads your driver's license number from a picture. It plays the correct podcast on the first try — a low bar, yes, but one Siri tripped over for a decade. You can talk to it naturally, type to it, revisit past conversations. It brainstorms recipes from whatever's rotting in your fridge. It splits a restaurant bill through the camera viewfinder. The feature list reads like the 2016 keynote dream, rendered real.

And the reaction is a collective shrug.

The problem isn't execution. The problem is timing. Apple dawdled while the category mutated. Google, OpenAI, Anthropic, and Microsoft didn't wait. They built agents that write code, compile research, operate browsers, control desktops, and chain tools together without human hand-holding. An AI that answers questions and fetches files is no longer a product — it's a baseline. Siri 2025 is a competent assistant in a market that has already promoted assistants to coworkers.

Apple knows this. The company's private cloud compute architecture, the sealed enclave that lets Siri touch your data without Apple seeing it, is a genuine technical achievement. It also functions as a ceiling. On-device processing and encrypted private compute protect privacy, but they limit model size, latency, and the ability to spin up the massive context windows that agentic workflows demand. Apple chose a constraint and called it a philosophy. Rivals chose capability and called it the future.

The gap shows in the details. Siri can launch apps and trigger shortcuts. It cannot, say, open a spreadsheet, analyze the data, build a chart, drop it into a presentation, and email the deck to your boss while you walk the dog. That workflow exists today — just not on iPhone. Apple's "App Intents" framework still requires developers to enumerate every action. The industry has moved toward function calling and tool use that models figure out at runtime. Apple's approach scales with engineering headcount. The rival approach scales with model intelligence.

There's also the voice problem. Apple touts adjustable expressivity and pacing sliders. That's polish on a modality that matters less every month. The dominant AI interaction pattern has shifted to text, to multimodal canvas interfaces, to side-by-side workspaces where the model shows its reasoning. Voice remains primary for driving and cooking — edge cases, not the center of gravity. Apple optimized the shrinking surface.

Yet Siri works. That matters. Millions of people will ask their phone for the photo of their insurance card, get it instantly, and never think about the infrastructure. They'll dictate an email while walking to the car and hit send without unlocking. They'll ask "when's that band playing near me" and get a correct answer sourced from the web, not a Wikipedia snippet. Utility compounds. Apple's installed base turns competent execution into mass adoption overnight. The feature doesn't need to be novel to be valuable.

But novel is what moves the stock. Novel is what attracts developers. Novel is what defines the platform for the next decade. Apple's AI strategy now resembles its Maps strategy: ship a solid product years late, iterate in public, eventually win through distribution. Maps worked because the bar was static. AI's bar rises monthly. A competent chatbot in 2024 was a moat. In 2025 it's a checkbox. In 2026 it'll be invisible infrastructure.

The anticlimax stems from a category error. We judged Siri against Apple's promises. The market judges it against the frontier. Those are different rulers. Apple met its own spec. The spec was obsolete at launch.

What comes next matters more than what arrived. Apple's silicon roadmap — M-series chips with increasingly massive neural engines — suggests the company intends to run larger models locally. The private cloud compute system can expand. App Intents can evolve into something closer to function calling. But each step requires Apple to relax a constraint it marketed as principle. The privacy enclosure that makes Siri trustworthy also makes it slow to evolve. Apple cannot have the agentic future and the absolute privacy past simultaneously. Something yields.

For now, Siri is fixed. That's not nothing. It's just not enough to feel like the future. The future already left the building.