But that's exactly why this "bubble" is so hard to predict, isn't it? The dot com bubble was simple. All companies were jumping the gun to make websites. Most were not useful, but the Internet was useful so it survived. LLMs are useful, but how much of its current use is actually valuable? If 95% is useful while 5% is memes, we're fine. But if 95% is useless, the industry will collapse hard. Who can tell what is useful when LLM companies sell the service to everyone?