Look, we all want that. And we probably will get much of it. But priorities.
We know that Apple (like all the other serious SW companies) is using the top level models like Mythos for coding. But there are likely constraints, both in terms of token access and in terms of training people how to code in this new AI-assisted/augmented style.
Meaning the highest priority for everyone right now is racing to fix security issues. This is obviously working; in every domain we're seeing a massive acceleration in security bugs being fixed, though (for obvious reasons) mostly withut details. A few open source projects (like Firefox) have given some numbers, though even there, and again for obvious reasons, no details except to say that the security vulnerability detection capabilities are astonishingly impressive, that almost everything detected is a real problem, and frequently so complex no human likely would ever have seen the chain of flaws/bugs leading to the end result.
So we ARE seeing many bugs being fixed; just so far mostly bugs that you "probably don't care about" in the sense that the impact your life (unless you're some unlucky person targeted by eg malware).
There is probably something similar going on in HW, using things like Mythos or Gemini 3.5 first to look for security flaws or weird conditions that could lead to the HW locking up/rebooting/delivering an incorrect result.
Now obviously we'd like this to transition to more user-visible bugs. And on the HW side we'd like AI used to make the HW faster.
And I think we will be seeing this soon (like over the next year); it's just a question as I say of priorities and limited resources, both on the AI side and the human experience/learning how to use the tools side.
