Not Separate but Equal. Separate but Better.

The Screen Was Never the Application

For forty years, we've treated accessibility as adaptation.

Build the visual interface. Then adapt it. Add labels. Fix keyboard navigation. Repair focus order. Hope the screen reader can reconstruct what the sighted user sees.

It was the only architecture we had.

But suppose the application's real interface isn't the screen.

Suppose it is the semantic model underneath.

Then the visual interface isn't the application.

It's one rendering of the application.

A screen-reader-native interface is another.

Neither is a copy. Neither is the original.

For decades we've judged accessibility by asking whether blind people can successfully use the sighted interface.

I think that's becoming the wrong test.

The test should be whether blind people have complete access to the application's capabilities.

If they do, accessibility has been achieved.

Not because we successfully adapted the visual interface.

Because we stopped pretending the visual interface was the application in the first place.

The Linux Foundation article,

https://aaif.io/blog/native-speakers-why-ais-most-powerful-users-are-blind/?utm_source=toptechtidbits&utm_medium=email&utm_campaign=06252026&utm_content=editorial

argues that accessibility requires both a complete semantic tool surface and a fully accessible rendered interface.

I think that's one architectural generation behind.

If the semantic surface is complete, then accessibility is a first-class, screen-reader-native rendering of that surface.

Not an adaptation. Not a reconstruction. Not a translation.

The visual interface remains essential for sighted users.

The screen-reader interface becomes essential for blind users.

Neither exists to imitate the other.

Both exist to express the same underlying application.

That isn't "separate but equal."

It's separate because the observers are different.

It's better because neither observer is forced to experience software through the other person's interface.