When you really start playing with openclaw, you'll realize: 1. You can't just give the little lobster 100u and expect it to earn you 1wu. 2. Even though you know it burns through tokens, you didn’t expect it to burn *that* many tokens. 3. The differences between models are huge—Claude is miles ahead, but it’s also ridiculously expensive. 4. Someone might tell you tokens are expensive, but no one tells you that APIs are also expensive, and even less that the official API is *crazy* expensive. 5. Even though your little lobster is loaded with skills, claiming to do image recognition, full data scraping, etc., in practice, it’s always limited. 6. And no one will tell you that whatever the little lobster can do, Claude can do too. After all the tinkering, when you finally get the little lobster to output daily reports and weather updates for you, you’ll hit a calm period. That’s when you realize you don’t actually need these features. The features you *do* need require expensive models to keep iterating, and they burn through a ton of tokens. The little lobster is really only suitable for people who have a clear idea they want to bring to life. Needs driven by public FOMO are all fake needs.