Does AI actually have a wall, or do we just keep moving the wall?

I follow AI way too much, and I swear people have been saying it is about to hit a wall for years. Then Fable 5 comes out. Now GPT-5.6 Sol. Once again, models are doing longer tasks, using tools better, needing less babysitting, and getting more efficient. Yeah, launch posts are marketing. I am not blindly believing every claim, and I am definitely not saying AI will improve this fast forever. But what would an actual wall look like? Would new models feel basically the same for years? Would companies spend 10 times more compute for improvements nobody notices? Would the same obvious weaknesses survive every new release? At what point does a temporary slowdown become a real limit? Because so far, the wall seems to move every time we get close to it.