Hacker News Reader

The problem I have with LLMs is that one can never be sure that it will give you the best possible solution. In fact in coding very often it will give you a working but also outdated solution. And this is futile. Because in coding even the best possible solution nowadays gets old very quickly. But if you use LLMs your code will be outdated from the start. That is nothing I would pay for.

View on HN

You have to look at it as a contractor. If you tell a contractor to "build me X" then you might get anything back with a high probability of getting something common but outdated. You have to write a specification for it with all of the constraints and preferences you have. Works well if you know the domain, if you're using it for something you don't know much about then you have to do more of the legwork yourself but at least it will give you a starting point that can inform your research.

With coding, getting a good enough solution quickly is usually more valuable than getting the perfect solution eventually. And as you say, things get outdated quickly anyway. I openai pay for speeding up my work. Instead of obsessing over something for an afternoon, I let it stub out some code, generate some tests and then let it fill in the blanks in under an hour. Time is money. The value of artisanally personally crafted code is very limited. And it's shelf life is short.

Reply on news.ycombinator.com