AI skills like prompt engineering will be obsolete by 2026, writer says

Why it matters: It changes how people should think about spending time learning AI tools. Rather than chasing short-lived techniques, it may be smarter to focus on the underlying job you want to do and keep pace with the tools instead of trying to be first.

Words to know: Prompt engineering: the practice of writing careful instructions to get a useful answer from an AI model.

Old habits that vanished

Charles Miller starts with a story from his youth. When he learned the C programming language, people insisted that one way to add one to a number was faster than another. The difference mattered in theory, but even by the mid-1990s, the most basic compilers already knew how to ignore the slow version when the result was not used. He says most older programmers still carry a dozen ingrained habits that modern computers actually make slower now ¹.

He is using that history to talk about something happening right now with AI. The lesson is simple: techniques that looked clever at one moment often become invisible because the tools improve around them.

Prompt tricks will disappear

The writer says the same pattern is already visible with prompt engineering. The careful phrasing, formatting tricks, and custom instructions that help people get better results from models today are the kind of things that get folded back into the products themselves. He expects that within months, the best ways of talking to AI will become the default experience, not a special skill anyone needs to learn ¹.

To make the point more sharply, he predicts that in five years some experienced people will still be adding long custom instructions like "you are an experienced software developer, make no mistakes" to their setups, even though the models will already handle that on their own ¹. He is not saying that learning about AI is pointless. You still need to do your job well today. But a skill that takes only a few weeks to pick up and becomes obsolete in a few months has no lasting value outside the moment it is useful .

Getting first is not the same as getting right

Miller also pushes back against the constant rush to adopt AI immediately. He recalls a CEO telling a crowded room eighteen months ago that anyone who did not jump on the AI train right now would be left behind ¹. He does not share that panic. He points out that companies like Google, Facebook, Apple and others have shown many times that doing a technology well matters more than being first to try it . His advice is to keep pace with the pack and be ready to move only when the moment is right, rather than chasing every new hype cycle.

What this means for ordinary people

The practical takeaway is not to abandon AI learning altogether, but to stop treating quick tricks as career advantages. The tools are improving faster than most training courses can keep up. If you are learning AI skills, focus on the actual work you want to do — writing, coding, designing, analysing — and let the AI part stay close enough to be useful without betting your career on techniques that will be automatic soon ¹.

This story reports on a trend for information only. It is not advice or encouragement to do any of it.

Key facts

Prompt engineering tricks are expected to be absorbed into AI products within months
The writer says AI skills that take weeks to learn and vanish in months have no lasting value
Getting a technology right matters more than being first, according to examples from Google, Apple and others
The advice is to keep pace with AI tools rather than chase short-lived techniques

Sources

“AI Skills” are worthless in 2026 — fishbowl.pastiche.org