Hacker News Reader

Comment by ironmagma | Hacker News Reader

Inevitably? Well, the promise of using something less efficient in terms of performance is that it will be more efficient in terms of development. Many times projects fail because they optimize too early and never built the features they needed to or couldn’t iterate fast enough to prove value and die. So if the native version is better but failed, it’s not so inevitable that it will get to that stage.

View on HN

Right, which is my point about LLM code assistants. If you did have two cases in the past: native but slow to add features so the project eventually dies vs. scripted but performance is bad enough it eventually needs to be rewritten. (Of course, this is a false dichotomy but I'm playing into your scenario).

Now we may have a new case: native but fast to add features using a code assist LLM.

If that new case is a true reflection of the near future (only time will tell) then it makes the case against the scripted solution. If (and only if) you could use a code assist LLM to match the feature efficiency of a scripting language while using a native language, it would seem reasonable to choose that as the starting point.

That’s an interesting idea. It’s amazing how far we’ve come without essentially any objective data on how much these various methodologies (e.g. using a scripting language) improve or worsen development time.

The adoption of AI Code Assistance I am sure will be driven similarly anecdotally, because who has the time or money to actually measure productivity techniques when you can just build a personal set of superstitions that work for you (personally) and sell it? Or put another way, what manager actually would spend money on basic science?

Reply on news.ycombinator.com