The Model Wars are Someone Else’s Problem
I was sipping my morning coffee, scrolling through AI news , ducking and sliding past the AI slop - looking for an interesting story for the second episode of the alfred frontier and came across this “Apple planning to open Siri to run any AI service via their App Store as part of iOS27.”.
First thing I thought was - genius....
They won’t spend a dime on training , raising billions , and join an AI race they’ve already been lapped in. Yet Siri will have access to all the best models and become the main interface for these amazing models in apple made devices. Not only that, now all the major models will compete for the 2+ billion active apple device users , giving apple the leverage. For your everyday user with not much interest in the building of AI….Apple gets the credit.
Absolute cinema!
They get the credit not for being smarter than OpenAI or Anthropic. But because they understood the model isn’t the product. The interface, the context, the workflow around it - that’s the product. We’re in the middle of a model arms race. Blink and another lab drops something faster, cheaper, better at reasoning, better at code.
Now lets apply this idea on a micro level.
The instinct most of have had is to chase (don’t believe me just check out most AI influencers content , you’ll probably find 40 videos about how this latest model is gonna change everything) , swap in the new model. Rebuild the prompt. Ship again. I’ve done it. Most people I know building in AI have done it. But now I realise if your product lives or dies by which model you’re using, you don’t have a product. You have a wrapper. And wrappers don’t compound.
Think about what Apple just did in reverse. Labs are spending billions trying to win the model race. Apple stepped back and asked -who controls the relationship with the user? Who owns the context, the interface, the moment of intent? That’s where the value concentrates. That’s where the defensibility is. The same logic applies at every level of building with AI. Eighteen months ago, we were obsessed with prompt engineering. Craft the perfect prompt, get the perfect output. That was the skill. Now those prompts are half-obsolete because the models changed. The work didn’t compound.
The people worth watching aren’t betting on a model. They’re betting on the system around it. Domain logic. Context architecture. How you route, what you expose, what you abstract away. That’s where the real design decisions live and right now it’s being skipped for the shiny thing. Model-agnostic as a design principle, not a retrofit. So when a better model drops, the system absorbs it. When a cheaper one handles 80% of tasks just as well, you route to it. No rewrite. No scramble. The model becomes a runtime dependency. Which means the model wars become someone else’s problem.
Apple’s a useful frame here because they made the outcome of that race largely irrelevant to their strategy. Whether that’s the right bet is still open. But it’s a more durable one than picking a winner.
So now that we have established the model wars are someone else’s problem. What problem are you solving that survives the next update?
Written by:
Sammi, Founder of Alfrd
Connect with me on Linkedin