I’ve had openclaw running on an old macbook that was otherwise just laying around, using MiniMax M2.5 and GLM-5 for the past couple weeks. Beyond it breaking its own config and me asking it to help add features and fix bugs in some open source applications I use, I didn’t really have a great use for it. But today I told it the US started a new war, it went out and searched multiple sources to confirm, and then gave me a summary of what news sites around the world are saying.
I asked it to specifically look for West Asia and Middle East sources, translate, and summarize them and provide me with updates on anything big.
It’s doing that over Telegram now, though it is having a hard time finding any accessible Iranian sites.
That’s pretty freakin awesome. It’s also doing automatic Text To Speech, so I can just listen to the updates as they come in too.
You must log in or # to comment.
Not a great use imo, easy way to get yourself psyop’ed by getting a bunch of sources and summarizing them when you don’t know what info comes from what sources without checking the credibility of each of them individually
honestly all this AI crap truly is best used for a meaningless desk job
Are you for real? That’s some serious straw-grasping.
That a statistical model can get duped by even the lowest form of propaganda or missed context is not grasping at straws.
But we can’t say anything to AI slop enjoyers on hexbear anymore or else they have tantrums.
This is just going out and rounding up headlines from varied western asian news sources and providing a summary of what is being said along with the links.
That seems fine, I was envisioning a Google Search AI Overview type of thing
This is what it looks like when it sends me a message on telegram, and I can tell it do stop or do something different or whatever at any time. I’m not sure why it decided to do a casualty update like that, maybe because it’s mentioned in so many articles or something but I’m gonna have it stop doing that.
Here’s the last update it sent me a few minutes ago:
Monitoring active. Next update scheduled for ~21:21 PT (SIC) (about 30 minutes).
Sources I’m tracking:
• 🇹🇷 TRT World (Turkey) • 🇶🇦 Al Jazeera (Qatar) • 🇦🇪 Gulf News (UAE) • 🇮🇱 Times of Israel • 🇷🇺 RT/TASS (Russia) • 🇵🇰 Dawn (Pakistan) • 🌐 Reuters, AP News • 🇮🇷 Iranian state media (when accessible)
I’ll check all sources and compile updates every 30 minutes. The system will trigger me via heartbeat, or you can prompt me anytime for an immediate check.
Current status as of 12:51 PM PT:
• 201+ dead, 700+ injured in Iran • Iran firing 200+ missiles at Israel, 89 wounded • Strait of Hormuz closed • Khamenei’s status: unconfirmed (Israel claims dead, Iran says alive) • Gulf states hit by Iranian retaliation (1 dead in UAE) • Oman mediator says “this is not your war” to US
I’ll ping you with the next update in 30 minutes.
I tried getting the docker setup for openclaw running today and it just would not work. Endless permission issues it would seem. Im curious about it, but not curious enough to let it run outside of a sandbox.
it is a bit of a struggle getting it up and going the first time for sure. I just let it run free on my old macbook after i set up an Asahi linux partition on it. It doesn’t have access to anything since I’ve never used it to log into anything other than whatever I did to install openclaw.
It definitely likes to mess up its own config files when you asked it to change things though. I imagine with docker you’d have to do some port forwarding for sure. Not sure what other permissions it would require.
It was odd stuff like wanting to put the working folder and the openclaw folder in /root/ instead of my home Dir. But trying to use my home Dir it ran into permission issues making the folders. I had to edit the docker-setup.sh script a number of timed to get it working but eventually had a weird token mismatch issues trying to approve my device. I could just turn that all off I guess but I had already spent way to long getting it to that point.
hmm odd. Mine has all its configuration and whatnot in .openclaw in my home dir.
It’s probably docker that is causing the issues because it requires root to run just about anything.
I’d go with podman myself if I were going to do any sort of containerized setup since it doesn’t require root but can use dockerfiles.
Now that’s an interesting use case!
What are you gonna do if/when Anthropic tries to prevent oc access?
I’m not using their models so they have nothing to do with any of it. I’m using very capable and quite cheap Chinese models to do the stuff.
Local or a service?
apis… local models are just too crappy to do anything reliably
Any recommendations? I’m using local models and npu/tpu/whatever devices at the moment.
i use GLM-5 and MiniMax-M2.5 through their coding plans which are pretty cheap and you can do them by the month
i tried with local models but they just suck too badly. they’re too limited to get things right, much less be consistent
This is not a great idea to be honest, I hope you have that laptop on a separate subnet at the very least. I looked at the code for it before and it makes Piefed look like good code.
There is also the issue of trusting the hallucination machine to summarize correctly but GaveUp@hexbear.net mentioned it already.
its fine, dont worry about it
