"Programming languages are dead." "Soon, you’ll just describe what you want in plain English."
We’ve all heard it. And sure, what LLMs can do is impressive - they can turn a simple prompt into functional code in seconds. But here’s the catch: whose prompt are we talking about? Mine? Yours?
It doesn't matter if the model understands us both. The real question is: do we understand each other? If I write a spec like Shakespeare and you write yours like James Joyce (or your mom, or whoever else), the AI might get it, but will we? Could you actually read a teammate’s "plain text" spec and catch a logical error? What about "Code reviewing" that becomes "Prose reviewing"?
It's great not having to focus on every semicolon in our C++ or Rust. We can stop thinking about syntax. But we still need a way to express semantics.
I wonder if we’ll ditch programming languages only to realize that we need an intermediate way to express our intent - not so the machines can understand us, but so we can. Will we end up just reinventing programming languages all over again?
No comments:
Post a Comment