3 ms·
Feels like taking the average of all text on the internet, does result in barely OK output. Very few people are truly “average” in the numerical sense of the wo
by graypegg 3y ago
Feels like taking the average of all text on the internet, does result in barely OK output. Very few people are truly “average” in the numerical sense of the word, so you’re going to see better results from most devs compared to the coefficient soup that tries to find the commonalities between the best and worst devs.
Truth be told though, speaking as someone that still does not use LLM tools at work… “just OK” is totally viable for a lot of things. Prototypes yes, products expected to be around a while, maybe not.
- withinboredom 3y agoI recently had this discussion at work. There's a difference between writing software that will be seen again in a few months and writing software that probably will never be touched again (or at least touching it again will be pretty risky). Identifying which one you're working on is super important. When you're writing software that is touched fairly often, the "tribal knowledge" of how it works will likely live-on. You can be a little bit clever at times, you don't need to comment as heavily, and your variable names can be a bit wonky. When you're writing software that is hardly ever touched ... everything needs to be crystal clear. Write lots of comments, explaining the "why" things are the way they are. You want to avoid cleverness except in the name of performance (but then comment how you expect it to work so that when someone comes along to change it, they understand wtf you are doing and the constraints you used to define the code). It's a totally different ballgame. AI doesn't get this distinction, hell, most programmers don't either.