Hacker News Reader

Comment by eterps | Hacker News Reader

I would love to be able to import the Project Oberon source code into an LLM, gain a deep understanding of it, and subsequently generate smaller subsets of the system (or variants like a CLI based system). Additionally, I'm interested in the potential to convert it into other programming languages, such as Nim, Go or Odin.

It seems that at the moment LLMs are not able to grok Oberon source code (although they are able to generate it).

View on HN

>gain a deep understanding of it

have you considered reading the book that explains the entire system?

> gain a deep understanding of it .. generate smaller subsets of the system

You can use the OberonViewer for this purpose with the original source code, or the Oberon IDE with a version of the Project Oberon System which runs with SDL on all platforms, see https://github.com/rochus-keller/oberon/#binary-versions and https://github.com/rochus-keller/OberonSystem/tree/FFI

IMHO, there's no good substitute for doing the work yourself. There are many strategies for reading source code, one of them may be a better use of time than fussing with an LLM.

> It seems that at the moment LLMs are not able to grok Oberon source code (although they are able to generate it).

This seems unlikely. The "knowledge" LLMs have is inherently bidirectional, so I suspect it's likely that one of the following is true:

- any model than can generate it can also grok it (though you may need to know exactly how to ask)

- that what it's generating isn't actually valid Oberon

- that it can only generate it for a very narrow subset of things (eg specific examples it's been trained on).

Reply on news.ycombinator.com