Apple has been slow-rolling intelligence features out to its customers, but it still has not gone after one of the most interesting opportunities sitting right in front of it: using one of the biggest strengths of LLMs, on-the-fly coding, to make software itself more customizable. Apple has already spent years loosening the shell of the operating system. We got widgets, Lock Screen variants, Control Center customization, Home Screen tinting, desktop widgets, and more ways to rearrange the furniture around the apps. But the apps themselves are still mostly fixed objects. Mail is still Mail. Notes is still Notes. Calendar is still the layout Apple picked for everyone. Apple has become comfortable personalizing the perimeter while leaving the thing you actually spend your time inside largely untouched. That split used to feel normal. Now it feels strangely arbitrary.

That is especially true on the Mac, which used to feel more porous. Back in the day, between .nib files and Interface Builder, a moderately technical person could get surprisingly close to the structure of an app and even make something like iChat feel a little more their own. Most people never did that, obviously, but the existence of the possibility mattered. It made software feel shapeable. It suggested the interface was not a sealed artifact handed down from above. Over time Apple traded more of that openness for polish, consistency, and safety, which was a reasonable trade when changing an interface required real engineering labor and carried a decent chance of turning an app into a mess. The software became more stable, but it also became more fixed. That bargain made sense when variation was expensive.

It makes less sense when variation gets cheap. AI coding changes the cost structure in a way Apple has not fully absorbed yet. The new default answer to “I wish this app worked differently” is increasingly “fine, I’ll build my own.” It has become dramatically easier for people to reproduce familiar patterns, clone workflows, and stand up rough alternatives to software they already use. But that is exactly what makes Apple’s omission here feel so strange. If people can now generate rough variations of software so easily, why should customization begin with abandonment? Apple already owns the hard part: the base application, the performance work, the accessibility work, the sync, the privacy posture, and the trust that the thing more or less does what it says it does. The more interesting move would be to let people bend those apps directly rather than forcing them to start over every time Apple’s default priorities stop matching their own.

And I do not mean just theming. The interesting layer is deeper than accent colors and icon tinting. I mean layout, hierarchy, density, workflow, and light functional variation. Some people want Calendar to behave more like a serious planner. Some want Notes to feel more like a research workspace. Some want Music to stop pretending discovery is the center of the universe and act more like a library. Some want Contacts to hold a little more context than Apple currently thinks is necessary. These are not always “build an entirely new app” problems. A lot of them are “let me retune the existing app around the way I actually live” problems. That is why I do not think the future is everyone forking Notes. The better model is that Apple exposes safe seams inside its apps, then lets an LLM help people compose from those seams quickly. The user gets to reshape the experience without inheriting the burden of maintaining an entirely separate codebase.

Once you define the problem that way, the role of AI becomes much clearer. The model should not be blindly rewriting raw app code every time someone asks for a denser calendar, a three-column notes workspace, or a more library-first version of Music. It should be assembling from approved behaviors Apple deliberately chose to make adaptable: layout regions that can move, modules that can appear or disappear, metadata that can be surfaced, navigation patterns that can be swapped, lightweight bits of utility that stay within clear platform limits. In that world, AI is not the author of the app. It is the translator between human intent and a structure Apple has already decided is safe to bend.

That is also why the interface for all of this should not feel like coding. It should not be a hidden power-user layer or a lonely prompt box dangling off the side of the app. It should feel more like a no-code modifier sitting above the system itself. You click a part of the app, describe what you want to change, preview a few valid versions, keep one, refine it, or roll it back. Maybe there is a more advanced mode for people who want finer control, but the default interaction should feel closer to editing a Shortcut than building software from zero. This is where AI becomes genuinely interesting to me, not as a gimmicky assistant stapled onto the side of the OS, but as a way of collapsing the syntax tax between intention and interface.

And even that is only the first half of the idea. The real unlock is sharing. If someone finds a genuinely better way to adapt an app, that should not stay trapped on one machine. People should be able to pass those customized versions around the same way they share Shortcuts now, except with a much richer surface area. That is the point where this stops being a personalization feature and starts becoming an ecosystem. Apple’s default version would still exist for people who want the standard issue experience, but a whole layer of better fits could start to grow on top of it. You would get something closer to an interface culture than a settings panel.

This is what makes Apple such an obvious candidate to do it well: it already ships the foundation people would want to build on. It already has the design language, the platform rules, the trust, and the installed base. It could even extend the same model to third-party apps by letting developers expose their own approved seams rather than turning the whole ecosystem into a free-for-all. That is a much more interesting vision of intelligence than another summarization badge or another generated convenience feature. The next real step in OS customization is not more icon tinting or another place to move toggles around. It is giving people the power to reshape the apps themselves without requiring them to become full-time developers. AI coding finally makes that plausible. Apple’s opportunity is to make it safe, elegant, and social.