Recently, many people on Twitter have been experimenting with Openclaw alternatives. But since Claude Code's official Telegram connection solution came out, I've even used Openclaw less. I highly recommend it! 👍 Claude Code is truly powerful in multi-step continuous execution and directing other agents. Especially after it starts subagents by default, the main session is rarely blocked. You can give it a bunch of complex tasks at once, and it can clearly divide them and advance and implement them one by one. This means that through Telegram, it's really like directing a highly efficient employee who never sleeps: you can assign tasks anytime, anywhere, and most of the time, once they're completed, it's relatively hassle-free. For example, letting it design solutions on its own, run multiple iterations all night, and collaborate with Codex reviews all go very smoothly with minimal issues. It's a completely different level from the experience of Openclaw + GPT, which was often interrupted and required constant confirmation. Also, the Mac mini is indeed a great tool: it can truly run silently 24/7 without taking up much space. The only bottleneck now is basically just burning through tokens for money 😅