|
They cynic in me wants to point out how for many (most?) of the programmers out there, they aren't working on novel problems that need a lot of involved thinking or using that gray matter to come up with something innovative. Instead, your average Joe will be writing some Java code with Spring Boot (or .NET with ASP.NET, or Ruby with Ruby on Rails, or PHP with Laravel, or Python with Django) where how to do things is already largely established and they only need to answer the "what?" question - to codify a bunch of boring business rules into a technical solution that runs. That work isn't entirely different from tasks where codegen shines - like writing DB migrations from entities, or entity mappings from some existing DB schema, or writing client code based on an OpenAPI spec, or writing an OpenAPI spec based on your endpoint definitions. The less people have to do that manually, the better the results seem to be (e.g. a bunch of deterministic code that will do everything correctly based on the inputs, given how well known the problem space is). You could also say the same for having something like autocomplete, or automated code suggestions, or even all of those higher abstraction level frameworks there in the first place - where you don't start solving a problem with "First, I'll write something that binds to a port and takes incoming HTTP requests..." but rather "I'll look up how you add a required constraint for a business object that's mapped to a DB table so validations fail before I even try to save something bad. If I'm really lucky, I'll be able to return a user friendly error message too!" There, it doesn't really matter whether the official docs will tell you that, a colleague, StackOverflow, ChatGPT, or anything else... you're not doing that much creative work, but something a bit more blue collar. The end result of all this is programming becoming something closer to data entry with ample amounts of code to just glue other solutions together. There, the risks posed by LLMs are far lower and the utility far greater - because a lot of the code is similarly structured and quite boring. |