@davidgerard It looks like he’s talking about using LLMs for decompiling/reverse-engineering proprietary code (and using the output to write proper, maintainable code that replaces the proprietary blobs), rather than one-shotting slop that kinda works. From what I’ve seen, LLMs trained on codebases have vastly improved the ability to do this once prohibitively laborious activity, and there’s an arguable case for using them to enable adversarial interoperability, unlike vibecoding.

@davidgerard Right. Did they reverse engineer all of those COBOL systems yet?

@davidgerard Written, Interestingly and ironically, by the Director for Digital Autonomy

@davidgerard Yikes!

@davidgerard that's funny, the very first sentence on their homepage is

Software Freedom Conservancy is a nonprofit organization centered around ethical technology.