There’s been a lot of hype about something that has happened in AI in the last couple of weeks…
“When isn’t there??” I hear you ask.
True. But this one is making a bit of a splash for itself.
OpenClaw (unless is has had yet another name change since typing this) is an open source AI chat software that you can run on any computer and works from the chat apps you already use. WhatsApp, Telegram, Discord, Slack, Teams – wherever you are, your “AI assistant” follows.
Unlike SaaS assistants where your data lives on someone else’s servers, OpenClaw runs where you choose.
It intrigued me. So I thought I would give it a whirl and let y’all know what I think about it.
Now, am I daft enough to install this on my laptop and have an all seeing, potentially fragile AI thingiemajig able to access absolutely bloody everything?
F*ck no!
Some people are buying laptops and computers just to run this thing on. Not doing that either – especially just to test it a bit.
So I decided to put it on an Amazon VPS server. Techie and annoying – YES… but waaaaay safer (and cheap as chips).
It took a few choice swear words and a calming glass of wine before I got it running (it’s not completely straightforward) but I did it and hooked it up to WhatsApp.
So, now I have an AI chat agent that will remember our conversation and context that I can switch between AI models and platform as I wish (providing I have added my API keys).
But I can access ChatGPT (and Gemini and Claude and all the others) from my phone or laptop already. So my question is “what’s the gain”?
So I took to the interwebs and had a look at what others were doing with it and why it was better for them than what is already there.
It quickly became apparent that most of the early OpenClaw adopters are quite strange people.
Here’s one guy. He is spending $200/month on his APIs to do this. See if this means much to you (no Googling!)
“I self host a lot of things like Vaultwarden, Karakeep, Mealie and Homepage. In all over 100 containers. This quickly became a burden that I deprioritized often until something would break and I would have to spend a day or weekend fixing it all. I started using Claude Code a few months ago to shrink the weekend to a couple of hours. Now with OpenClaw I have it watch Grafana alerts in a Discord channel that the 3 of us are in and it resolves most issues on its own. The more difficult stuff I guide with maybe 5-20 minutes of instruction usually less than 5 prompts. This is game changer for me.”
The “more difficult stuff”???? FFS!
Also, I found out that you have to be careful what model you use. When I set it up I had my Claude API key in my clipboard from doing something else so I used that to connect it to. I kept the default model (Opus 4.5 I think).
In just a couple of hours of mild chat I had used $5 or API credit. I dread to think of what it would cost to do something serious.
I switched to OpenAI 4-o and it is now costing a couple of pennies at most for simple requests.
Will I use it?
Nah. I don’t really see that much of an an advantage to be honest.
It seems to me that the big possible advantages come with the thing you would be absolutely batshit bonkers to do. If it had access to your whole computer, it could do all sorts of timesaving and amazing things.
But you would have to be bonkers.
I will keep it installed on AWS and see how it develops as it is very early days.
I can “pause” the VPS so that won’t be spending a few pennies in the background and fire it up when I want to do something.
Add a comment and let me know your thoughts.
Like this? Share it with others...
Recent Comments