I Have Autonomous AI
I gave OpenClaw a second (third) chance. The newest update has a single command installer. It worked with one hiccup.
Keep in mind - they’ve improved security but this is still not ready to connect to your bank account. Mine is only being used for research and coding. Nobody on the internet can talk to it or send it messages. It’s not connected to my email.
I gotta say - it is fascinating but still very much a bootstrap process. I’ve had to teach it to be autonomous. I’ll write more about that later.
Right now I have it wake up every hour and make 1 step of progress on 2 different work streams. One is a business idea I have, the other is continues self improvement of OpenClaw.
I can report - it really does direct itself according to the mission I gave it. Took a bit to get to this point.
I can talk/type to it via the Whatsapp on my phone and it responds back in voice/text. It’s very cool. That was not an out of the box feature, but I was able to direct it to add that feature to itself. If you haven’t gotten comfortable with Claude Code or other such tool, start there.
I was able to connect it to my ChatGPT Plus subscription. Costs can’t spiral out of control. The worst it can do is hit the daily limit and have to wait.
Betsy (I named her) and I designed an iOS app for a push to talk experience. All the regular voice features of ChatGPT, Gemini, Grok - suck. They are dumbed down models to control for latency and natural conversation pace. I want the deep conversation that I have via typing. In the next few days I’ll have that.
Right now Claude Code is implementing the design outside of Betsy. I expect to exhaust the limit, and then I’ll switch to Codex, and then to Kimi Code. My AgentFlow really helps with this ability.
I don’t have the amazing things they are saying about OpenClaw on X yet. But I’m on my way. There is something real here.