Yes. How else could the LLM even begin to do what it did otherwise. Not a criticism of you, just a place for a rant but I still find it incredible when people think they're doing something novel when really they're just reassembling existing work.
What really reinforced this for me was that I recently vibe coded a home assistant addon that, as it turns out, there was a github project that did almost the same thing. The initial implementation that came out of the LLM was essentially the same just with just some small tell in things like the logging that came from what I prompted it.
It was a very small project (maybe 500 LOC) so there's a greater chance of coincidence but it really felt like the office junior had just ripped off someone else's work and tried to hide it (badly) and really, that's exactly what they do
I had performed a quick GitHub search and not found any code. It certainly didn't have these decoy notes documented on its own. It only found them because I provided it a known Erik Satie piece, which has a specific number of notes, and it dug in to figure out what the excess notes were.
How else could the LLM even begin to do what it did otherwise.
Are we really still having the "LLMs can only regurgitate what was in their training data" argument? Where did the counterexample to the Jacobian conjecture come from?
What really reinforced this for me was that I recently vibe coded a home assistant addon that, as it turns out, there was a github project that did almost the same thing. The initial implementation that came out of the LLM was essentially the same just with just some small tell in things like the logging that came from what I prompted it.
It was a very small project (maybe 500 LOC) so there's a greater chance of coincidence but it really felt like the office junior had just ripped off someone else's work and tried to hide it (badly) and really, that's exactly what they do