The keyboard, the audiobook, the subtitle: built for disabled people first. None of them closed the gap.

The machine you are reading this on begins with a blind woman. Some time between about 1801 and 1808, the Italian inventor Pellegrino Turri built one of the first working typewriters so that his friend, the Countess Carolina Fantoni da Fivizzano, who had lost her sight, could write letters without a scribe. He invented carbon paper to give it ink. The popular version has the two as secret lovers; that is a novel, published in 2010. The record, the letters that survive in the state archive at Reggio Emilia, has her addressing him simply as "my dear friend". The typewriter was assistive technology first. The keyboard under your fingers is its grandchild.

It is a pattern, not a one-off. Alexander Graham Bell was working on hearing when he arrived at the telephone in 1876. The audiobook began in 1931 as the American "Books for the Blind". Television captions were built in the 1970s for Deaf viewers, tried out first on a cookery programme, and now a large share of the country watches with subtitles on and the sound off. Texting grew out of a system built so Deaf people could use a phone line. Designers have a name for this: the curb-cut effect, after the dropped kerbs cut for wheelchair users that turned out to help everyone with a pram, a suitcase or a delivery trolley. Build for the person shut out, and the whole street walks through.

So when the pitch says AI built for disabled users will spill outward and improve everyone's working life, grant it. That is exactly what the record shows, and it will very likely happen again. But notice what the same record does not show. The typewriter became everyone's keyboard. It did not make blind Britons employed at the same rate as sighted ones, and two centuries on they still are not. Spilling outward to everyone is not the same thing as closing the gap for the people a tool was built for.

There is a second, quieter pattern worth naming, because it is the one we are living in. Every wave of technology has been sold as the one that would finally bring disabled people into work. The personal computer was. The internet was. Working from home was. Each was real, each helped somebody, and across the whole run the employment gap has barely moved off thirty per cent. The tools kept their promises to individuals. The promise about the labour market was never quite theirs to keep.

This is Disability Pride Month, and the British half of the story sits in the Science Museum, which keeps the typewriter's history precisely because accessible design turned out to be good for everyone. That is worth celebrating. It is also worth holding both halves at once. The long record says the new tools will help, and will reach far beyond the people they were made for. It does not say they will close the gap, because the gap was never only about whether the machine could be worked. The same wave of AI our own desks run on (models built by Anthropic, whose industry we cover) will open doors. Whether Britain walks disabled people through them is an older question, and the machine has never been the whole answer to it.