The AI hype is at a fever pitch with tech CEOs declaring coding as we know it is done, over, kaput, a solved problem, while everyone from Uncle Bob to Linus Torvalds and DHH is embracing it, albeit in a measured way, with much less fervor than AI CEOs.

What's an APL programmer to think? We are, by definition, contrarians. After all, we use a language that's over 60 years old and fundamentally unchanged over that time. We use strange symbols. We don't write loops. Our opinion of AI will not follow the crowd.

Is there anything more absurd than using natural language to explain a complex problem to a computer? This is what APL is for. Why would we take a concise, powerful, unambiguous notation designed specifically for humans and not for computers, and replace it with a verbose, vague, error-prone means of communication between man and machine? This is one big step backwards. Who in their right mind wants to spend all day having a discussion and argument with a laptop about what to code? We can think of nothing worse than having to essentially talk to the computer for a living.

Perhaps AI is for programmers who use languages designed for computers rather than people.

Boilerplate

We often hear programmers lauding AI because it relieves them of the onerous task of writing boilerplate code. There are two problems here. First, if your language requires boilerplate, that itself is a problem. Second, it is one of the fundamental jobs of a programmer to eliminate duplication and boilerplate. What kind of programmer writes boilerplate code more than once? Not a good or experienced one.

Documentation

Another task that programmers hope to rid themselves of is writing documentation. If documentation is viewed as something to be tacked on to the end of a project after the design and programming is done, then documentation is indeed a chore, and largely a worthless effort too. Documentation is design. Nothing shows the flaws in a design more than explaining it on paper to the person who is going to use it. Documentation, like coding itself, is an essay, an attempt to understand a problem. Writing documentation as you code allows you to write what should be, rather than what is. It allows you to improve the design rather than accept it as given.

Tests

Programmers also hate writing tests, and hope AI will relieve them of this job too. But test are just documentation written in code rather than natural language. Tests perform the same function as documentation: design.

UI

Then there is the UI, yet another dreaded task of the lofty software engineer. But putting a button on a screen is not hard. Positioning it, making it a nice color, giving it a drop-shadow, or nice hover effect, none of this is hard. What is hard is deciding that the button should exist at all. Or, if it should exist, what screen should it be on. Or what other controls should be near it. Or what the button should do when clicked. The difficulty of UI is in the design, not the implementation.

Craftmanship

A hand-made software program is not like a pair of hand-made shoes. Hand-made shoes can be sold once and used by one person at a time. Hand-made software can be copied infinitely for zero cost. Software craftsmanship is not like physical world craftsmanship. Better tools never remove the need for software craftsmanship, and in fact do the opposite. Software is design, not manufacturing.

DIY

The AI promoters want everyone to write their own, disposable programs. This is silly. Some people want to build their own computers, or build their own guitars, or build their own cars, or their own APL interpreters. These people are valuable and more power to them. But the vast majority of people want to use a computer, a guitar or car that has been very carefully crafted by people deeply immersed in the specific problem. The same holds true for software. Even programmers want to use software carefully crafted by other programmers.

The Scam

The genius of the AI salesmen is that they have designed a software product that is bug-free by definition. If it is wrong, it just says Sorry! My mistake. You are absolutely right. Let's fix that! and then proceedes to get it wrong again. And again. And again. The AI companies want you to buy into this massive, expensive dependency just to write a line of code. AI for coding is a solution to a problem that shouldn't exist.