logoalt Hacker News

keeda • yesterday at 6:03 PM • 1 reply • view on HN

The problem is, the low hanging fruit, the stuff it's good at, is 90% of all software. Maybe more.

And it's getting better at the other 10% too. Two years ago ChatGPT struggled to help me with race conditions in a C++ LD_PRELOAD library. It was a side project so I dropped it. Last week Codex churned away for 10 minutes and gave me a working version with tests.

Replies

overgard • yesterday at 9:19 PM

I think that typescript is a language uniquely suited to LLMs though:

  - It's garbage collected, so variable lifetimes don't need to be traced
  - It's structurally typed, so LLMs can get away with duplicating types as long as the shape fits. 
  - The type system has an escape hatch (any or unknown)
  - It produces nice stack traces
  - The industry has more or less settled styling issues (ie, most typescript looks pretty uniform stylistically).
  - There is an insane amount of open source code to train on
  - Even "compiled" code is somewhat easy(er) to deobfuscate and read (because you're compiling JS to JS)
Contrast that with C/C++:

  - Memory management is important, and tricky
  - Segfaults give you hardly anything to work with
  - There are like a thousand different coding styles
  - Nobody can agree on the proper subset of the language to use (ie, exceptions allowed or not allowed, macros, etc.)
  - Security issues are very much magnified (and they're   already a huge problem in vibecoded typescript)
  - The use cases are a lot more diverse. IE, if you're using typescript you're probably either writing a web page or a server (maybe a command line app). (I'm lumping electron in here, because it's still a web page and a server). C is used for operating systems, games, large hero apps, anything CPU or memory constrained, etc.
I'm not sure I agree that typescript is "90% of all software". I think it's 90% of what people on hacker news use. I think devs in different domains always overestimate the importance of their specific domain and underestimate the importance of other domains.