In the last one month or so, OpenClaw went viral. Some people claimed that it was even more viral than ChatGPT. I was a little late to the Claude Code wave, so I wanted to wait this one out as well. When everyone was screaming about how crazy OpenClaw was, I was trying to understand why Claude Code became such a big hit - even with non-technical people.
My other reason for avoiding it was that it looked like a tool for ‘shallow work’. This is a term popularized by Cal Newport who I have been following for a long time. A chat with a bot on your phone is no way to do intense ‘deep work’, I reasoned.
But I kept seeing post after post, and very soon it started to be a matter of now or never for me. I decided I had to do it. I needed to know how to use OpenClaw safely - because anything connected to all your accounts obviously can do a lot of harm. The last push was probably Greg Isenberg’s interview of Kitze where the guest said something on the lines of ‘if you are not using Telegram then what are you doing’. I decided that I would start with one Telegram bot as my communication channel.
Like everything open-sourcee, OpenClaw has loads of clones. I picked TinyClaw because it looked like a nice subset. I don’t have time for monitoring my Claw a lot, so I chose this simple option. It was quite easy to get started with, and in no time I was chatting with my bot ‘Ramu’ which was connected to TinyClaw running on my Mac Mini.
For a while I didn’t know what to ask my bot to do. I tried to shop for men’s pants for a while, which caused my Claw to install Agent Browser and open up a screenshot on my Mac! I realized that this was not a good use case, even if I managed to get the screenshot sent in the chat. I kept chatting on-and-off with the bot while handling a new baby in the house and her older brother. My situation was close to this post, although not so severe.
Today in the morning I managed to update my Notion journal from my bot. And even though the text wasn’t added exactly where I wanted, it was a beginning. I was definitely developing new mental muscles.
I am convinced that this is going to be a game changer for so many scenarios. A lot of software needs to change as well, because they were never developed for bots to use. For example, the idea of user accounts on computers is for human users. Very soon, every human user might be using dozens of bots. Yes, there’s docker and there’s always the option to host on cloud servers. But AI assistants will be mighty useful inside local networks and computers. Not everything needs to go to the cloud. The OS of the future probably needs to be different. A lot of human-focused functionality can be removed - or kept optional. The possibilities are endless.
I’ll be playing with TinyClaw, and I also want to try out clones like ZeroClaw which incidentally is a friend’s project. Someday I might try out the OG OpenClaw as well. I am surely having fun with this, which is what Peter Steinberger - the creator of OpenClaw urges us to do.