> So LLMs finally hit the wall. For a long time, more data, bigger models, and more compute to drive them worked

We can't say whether there is a wall, since we don't have anymore data to train on.