The performance industry has done extraordinary work. Lighthouse scores, image formats, edge networks, server components, partial hydration, prefetch heuristics. The page loads in 1.4 seconds instead of 3.8 and a generation of engineers can credibly point at the difference.
Meanwhile, the actual rate at which a user can get something done on the web has barely changed. The bottleneck moved. It used to be the network. Now it’s the interface itself.
What “interface bottleneck” means
Most sites are still designed for a single user persona: clear-eyed, fast-fingered, mouse-using, fluent reader, no cognitive load issues, default English, full color vision. Everything outside that persona pays a tax. Sometimes it’s a small one (five extra seconds to find the right button). Sometimes it’s the whole task: a checkout flow you can’t complete because the form labels never made it into the markup.
Add up that tax across the population that can’t use the default interface comfortably (roughly one in six adults globally, depending on which definition you accept) and it’s a massive amount of dropped completion that no amount of CDN improvement is going to fix.
What the fix actually looks like
- The page adapts to the user, not the other way around.
- Adjustments are immediate and reversible, not a settings panel and not a separate app.
- The same adjustment works across every site, because the user has the same needs across every site.
- The site owner is told what’s failing for users, not just what failed an automated checklist.
Why we build at the extension layer
Waiting for every site to fix itself doesn’t work. It hasn’t worked for twenty-five years of WCAG. The extension layer is the one place where the user gets agency: any site, any time, no permission from the publisher.
We still want the sites themselves to do better. That’s why the audit is free and the fix plans exist. But the user can’t wait. The bottleneck has been the interface for a long time. The opening is here for someone to ship the dial that turns it down.