There are different tools for different jobs.

Code is like words. On their own, it’s just gibberish on a page.

Programming is when you use code to solve a problem. Maybe you just need a script to run once and never again. That’s like firing off a quick text: short, disposable, gets the job done. Sometimes it doesn’t even require perfect spelling or even real words - as long as the message gets across.

Software engineering is a whole different game. Now you’re not just solving a problem — you’re thinking about time, scale, users, maintenance, money, team skills, and the next ten problems around the corner. That’s like writing The Lord of the Rings. Suddenly you need plots, characters, whole languages — consistent, coherent, and expandable for future stories.

It would be daft to write a text with Lord of the Rings-level rigour. And it would be a disaster to write Tolkien with text-message sloppiness. Different jobs need different approaches.

AI today? Brilliant at programming. Good prototypes. Smashes benchmarks. Likely to get better and better.

But, it’s currently not good at creating or contributing to complex software reliably. This will improve, for sure. But whether purely scaling AI model size and training is enough to solve the software development puzzle remains to be seen.

Let me know what you think below.