#39 We invented computers so we wouldn't have to calculate anymore. Then we spent fifty years typing.

We talk about ChatGPT, Claude, or Gemini every day, as if they were objects that fell from the sky in 2022. Let's take three minutes to step back. Computing was invented for a specific reason, and that reason was never to make humans type on a keyboard eight hours a day.

The original intention: to delegate repetitive work

Before it was a machine, the word "computer" referred to a profession. A computer was a person, often a woman, paid to perform calculations in an assembly line in astronomy, ballistics, or insurance offices. The entire history of the discipline stems from this: relieving humans of a mechanical task that occupies them without enriching them.

Charles Babbage designed his Difference Engine in the 1820s because published tables of logarithms and navigation were riddled with human error, before conceiving the more general Analytical Engine in 1834. Herman Hollerith invented punched-card data processing for the 1890 US census, which threatened to take more than ten years to process manually. ENIAC, in 1945, calculated artillery trajectories. In all three cases, the objective was the same: the machine took over the repetitive work, while the human retained the judgment.

Note what is missing from this genealogy. No one ever wrote that the goal was to create a job consisting of re-entering data into a form.

The visionaries described something other than what we have built

In 1960, J.C.R. Licklider published a text entitled " Man-Computer Symbiosis ." His thesis was clear: the target relationship is not that of a tool to be manipulated, but that of two partners who think together, with the human setting the goals and formulating the questions, and the machine doing the preparatory work. He added a now-legendary remark: in his own work, approximately 85% of his "thinking" time was actually spent on mechanical preparation tasks, not on thinking.

Two years later, Douglas Engelbart published " Augmenting Human Intellect ." In December 1968, he demonstrated the mouse, hypertext, videoconferencing, real-time collaborative editing, and windowing in San Francisco. All of this has existed for fifty-eight years. The demonstration has gone down in history as "The Mother of All Demos."

What Licklider and Engelbart described was a conversation. What the industry delivered was an input interface.

The keyboard is not a conquest, it is an inheritance

Look at the object in front of you. The layout of your keys dates back to Christopher Sholes's typewriter, patented in 1868. It wasn't optimized for the speed of human thought, but to prevent the mechanical hammers from jamming. In 2026, we're using an engineering compromise designed to solve a linkage problem that hasn't existed for a century.

The screen, the window, the icon, and the pointer all originated with the Xerox PARC in the early 1970s, popularized by the Macintosh in 1984. Since then, the interface design has remained virtually unchanged. Forty years of interface stability in a sector that presents itself as the fastest-growing in the economy is an anomaly that no one comments on.

The consequence has become so ingrained in our culture that it's practically invisible: to get anything from a machine, it's up to the human to learn the syntax. The drop-down menu, the folder structure, the spreadsheet formula, the programming language, the required form field—each time, the human translates their intention into a protocol designed for the machine. We call this "knowing how to use a computer." It's actually the price of a historical misunderstanding.

And now ?

The question that opens this series is therefore not "will AI change everything". It is more disturbing: what if the computing of the last fifty years has been a long technical parenthesis, and what if we are returning to what Licklider and Engelbart described from the outset?

If that's the case, then language models aren't a break from the past. They're the next layer of abstraction in a series that already includes five or six, and the real issue isn't the technology or the models, but the relationship between humans and machines. That will be the subject of the next article.

At Agence Alps, we approach this question from a very practical angle: which of your teams' tasks still involve data entry and formatting, and what would remain of their work if these tasks were eliminated? If you're preparing a digital roadmap for 2027, now is the perfect time to ask the right question .