I don’t want to describe what I want to a computer in plain English. I want to express it in a formalised language more akin to mathematics than prose. Unambiguous. Logical. Deterministic.

Even if describing it in English worked perfectly, and had no other downsides, I’d still hate it. Because I wouldn’t be making it, I’d be product managing it, & I *never* wanted to be a PM.

The whole point was to directly make things. To feel the digital clay in my hands. Being a supervisor is not enough.

Some people say it’s just like moving from assembly to a higher level language, like the LLM is a compiler. But it’s absolutely not - compilers are still formal language and completely deterministic: you can trust them to produce a known output for a given input, every single time. With such a guarantee, it truly is just an abstraction, a productivity tool with directly analogous characteristics to the thing it is building on top of. You’re still in control, every action has a traceable outcome
LLMs are deliberately non-deterministic. So in that regard they are more like humans I suppose; semi-random and in need of coaxing. If you like a supervising role in conditions like that I guess you’re fine, but I don’t. I like to make things with tools that do exactly what I say, every time, like clockwork. Even if it takes longer than asking someone / something else to do it for me. Because the making was the point.

@sinbad you’re right, it’s not really programming anymore. It’s not as hand crafted.

Of course it depends on whether you enjoy programming or making your vision come true.

If I understand it correctly you’re proud of doing everything yourself and LLMs are diluting that.. so you choose not to use them.

For me, the programming part has become tedious a long time ago.

I have an idea and I want to realize it. LLMs give me more agency in that regard by making larger projects viable.