People will definitely be using natural language (English) to write some categories of software, but it will look more like when higher languages were created to be compiled into assembly than what we currently refer to as vibe coding (which implies no specialized knowledge of software/systems design).

There were plenty of people who said higher level programming languages wouldn’t work for a variety of reasons (inefficient, loss of control, no understanding of what the computer is really doing). In the 90s, there were plenty of people who criticized interpreted languages (eg Classic ASP) for the same reasons.

The criticisms were accurate but the ecosystem routed around them because the trade off was too positive. Today very few programmers can explain how their program displays “Hello World” on the monitor, or reads keyboard input, for example. And it doesn’t matter. That software knowledge is no longer relevant in most scenarios, and programmers are better for it.

We can’t analyze these things as if the ecosystem is static (current memory/CPU constraints, current token economics, current workforce skills, etc.). We need to ask ourselves whether the ecosystem will react to new pressures and forces such that it will produce a world in 5 years where X or Y makes sense.