Designed for One, Used by Millions

Count how many people you see today watching video with the sound off and the captions on. In a train carriage it will be most of them. Captions were built for deaf and hard-of-hearing viewers, mandated after decades of advocacy that the industry resisted as expensive and niche, and they are now used overwhelmingly by hearing people in loud rooms, quiet offices and beds next to a sleeping partner.

This keeps happening. It happens so reliably that it has a name — the curb-cut effect, after the ramps cut into pavements for wheelchair users, which turned out to be indispensable to anyone with a pram, a suitcase, a delivery trolley, a bad knee or a skateboard. Build for the edge and the middle arrives uninvited.

The pattern, and why it is not a coincidence

There is a structural reason assistive design generalises. Designing for a user who cannot rely on a given channel forces you to make the underlying information explicit and available in more than one form. Explicit, multi-channel information is simply better engineering — it is more robust, more machine-readable, easier to translate, easier to search, and it degrades more gracefully.

Consider the lineage:

• Text-to-speech, built for blind users, is why anyone can listen to an article while driving.
• Speech recognition, developed substantially for people who could not type, became dictation, voice search and every voice assistant on the market.
• The typewriter was invented, by one account, so a blind countess could write legible letters.
• The telephone emerged from work on hearing and speech; its inventor's mother and wife were both deaf.
• Predictive text and autocomplete came from communication aids for people with motor impairments.
• High-contrast modes and text scaling, built for low vision, are now the settings everyone reaches for on a sunny day or at forty-five.
• Semantic document structure, required for screen readers, is the same structure that makes documents searchable, convertible and durable.

Accessibility is not a constraint applied to a finished design. It is a set of questions that make the design better before it is finished.

What changed in the last decade

Two things, both consequential. First, assistive capability moved from specialist hardware into the general-purpose device. A dedicated screen reader once cost more than a car; a capable one now ships free on every phone. Magnifiers, colour filters, switch control, real-time captioning, live transcription, object recognition and text extraction from photographs are all built in, maintained by the platform, and updated without anyone having to apply for funding.

This matters beyond price. Specialist hardware marks its user as disabled in public. A phone does not. A great deal of the value of mainstream assistive technology is that it is indistinguishable from what everyone else is carrying.

Second, machine perception got good enough to be useful for real navigation tasks. Live description of a scene, reading a menu aloud, identifying a banknote, recognising a face, finding the door — these are not novelties for someone who could not previously do them independently. Simultaneous live captioning of an in-person conversation has changed daily participation for a lot of hard-of-hearing people in a way that no amount of goodwill managed.

What it replaced

Dependency, mostly. That is the honest answer and it is worth stating without sentiment. The specific thing being replaced is having to ask another person for help with something you would rather do alone — reading your own post, choosing your own clothes, navigating your own route, knowing who just walked into the room. Independence is not a feature. It is the entire point.

It also replaced a great deal of segregation in design: separate entrances, separate versions, separate "accessible" websites that were always worse and always out of date. The modern position — one product that works for everyone — is both cheaper and less demeaning, which is a rare combination.

What it cost

Fragility of the mainstream. When accessibility rides on a consumer platform, it is subject to consumer product decisions. A redesign that breaks screen-reader navigation is an inconvenience for most users and an outage for some. The dependency is real and mostly invisible to the people making the decision.

Automation without verification. Automatic captions are a genuine advance and they are also frequently wrong, particularly with accents, technical vocabulary, crosstalk and proper nouns. Treating auto-generated output as compliance — ticking the box and moving on — produces something that looks accessible and is not. The same applies to automated image descriptions and generated alt text.

The overlay problem. A whole industry sells scripts that promise instant website accessibility. Disability advocates have been consistent and vocal that these frequently make things worse by interfering with the assistive software users already have configured. The demand for a one-line fix is understandable; the fix does not exist.

Uneven coverage. Cognitive and learning disabilities receive a small fraction of the attention that sensory ones do, despite affecting more people. Plain language, predictable navigation, reduced time pressure and forgiving error handling are cheap, benefit everyone, and are consistently the first things cut.

The design lesson

Every product decision picks an implicit range of users — how well they see, how steadily they can tap, how much they remember between screens, how much noise is around them, how much stress they are under. Most designs quietly assume the designer. Widening that assumption is the single highest-leverage quality practice available, and the history of this field suggests the beneficiaries will mostly be people who never thought of themselves as needing help at all.