OpenClaw

“I use it for many use cases. Took new job, so it's finding me a home, finding title attorneys, helping me craft negotiation strategy, build offer, research new industry, analyze interview transcripts and build 90 day new job onboarding plan... around house, I can tell it to move files from my home or g drive to my NAS and then scan plex library and have files ready to watch on plex in seconds... it has been trained to go into my split stack home network and upgrade various docker containers and ”

“…why would you ever want an AI that directly manipulates files? I’m absolutely flabbergasted. These people…they should understand the limitations of AI better than everyone else and how idiotic it is to have a LLM do anything but produce text. At least then you can identify the errors and correct them yourself. That’s…just what OpenClaw is? And people WANT that?! This is literally going to keep happening. And I think I am going to have zero sympathy whatsoever for the inevitable catastrophic out”

“I’m finding the discussions in here to be a breath of fresh air. After hearing the hype around this shit I wanted to see what it truly could do if it was given as much power as possible. I set up a virtual machine, gave it root access inside an isolated VLAN with its WAN traffic routed through a VPN. This shit burned through so many tokens trying to accomplish basic tasks, and then ultimately failing to deliver. Oh, and I forgot to mention: while it tried and failed to get shit done it installed”

“It accumulated tech debt at lightning speeds since they let a gabagillion agents loose on that code base without any human oversight. Lllms don't fear complexity like we humans do... so they just keep on piling on complexity every day... Now it is a total mess and it keeps growing. Go read the source code. It is a magnificent ball of tangled nonsense.”

“Thanks for your sharing! I’m actually stuck in the middle whether to spend the money on setting up Openclaw or not. There has been many negative comments on the tools recently. I am encouraged by your words: why choose to be overwhelmed by the tech development rather than experience it by yourself. So it rly worths it to keep up with trends and actually feel it on my own.”

“I’m finding the discussions in here to be a breath of fresh air. After hearing the hype around this shit I wanted to see what it truly could do if it was given as much power as possible. I set up a virtual machine, gave it root access inside an isolated VLAN with its WAN traffic routed through a VPN. This shit burned through so many tokens trying to accomplish basic tasks, and then ultimately failing to deliver. Oh, and I forgot to mention: while it tried and failed to get shit done it installed”

“Hey everyone, I've spent a lot of time digging into how people use OpenClaw, and there's a pattern that keeps showing up: most users are burning way more money than they need to. Here's why: 1. Model defaults are expensive Most people leave their agent on the default model (usually the most powerful and most expensive one). But 80%+ of everyday tasks — browsing, form filling, simple lookups - don't need the top-tier model. A cheaper model handles them just fine, often 10-20x cheaper. 2. There's ”

“I have started with Openclaw to work as an agent for my business operations. I purchased vps on hostinger. Started with linking Gemini API but it hit the limits pretty soon. Suggest what models should I interate for smooth operations”

“Yes, I'm using it and I like it. I find it just as weird that people have such strong opinions about software they're too scared to use themselves. I was a hater too at first because every use case sounded like stuff I could do with claude code anyway. Install it on an old laptop or something. You actually have to work to give it permissions to things. It's not going to drain your bank account if you run it on a freshly installed os. Don't give it access to things if you don't want to. I find th”

“Hard to install and get working, even with the help of semi-automated AI tools like Claude Code or Cowork. This is definitely not for the average user — it feels aimed almost entirely at developers or highly technical people, which locks most users out from being able to use it properly. It’s also extremely buggy and feels very insecure at times. A lot of the core functionality simply does not work reliably, especially Google Workspace authorization and tool calls, which frequently fail or brea”

“Every time I hear about OpenClaw it's either malware or something else like this. I already avoid AI in general as much as I can, but even if I was an AI bro I'd avoid this in particular like the plague.”

“Hey everyone! Wanted to share a wild week we just had and get your honest take. So my co-founder and I have been building an AI agent platform for enterprise clients — sandboxed execution, 97 LLM models, the works. B2B clients were happy, things were chill. Then OpenClaw blew up (157K GitHub stars!) and we started hearing the same story on every sales call: "We loved OpenClaw until someone's agent ran rm -rf on a shared volume." People want AI agents. They just don't trust them running loose on ”

“You have hands down the most engaging tutorial on this. If only you could have been my elementary school teacher in programming... I might have went on to study CS. Loving your vibe and editing. Instant sub. Also, there was copius amounts of coffee consumed during this video and I'm all for it 😂”

“Hard to install and get working, even with the help of semi-automated AI tools like Claude Code or Cowork. This is definitely not for the average user — it feels aimed almost entirely at developers or highly technical people, which locks most users out from being able to use it properly. It’s also extremely buggy and feels very insecure at times. A lot of the core functionality simply does not work reliably, especially Google Workspace authorization and tool calls, which frequently fail or brea”

“Try to setup one for my partner so she can have a persistent agent without having to do the Pi + tmux + VPN like I do. I thought openclaw built on pi would keep all the nice things pi have but with extra features. Wrong. It's a pile of crap. Convoluted setup. Random intermittent code error in their own tool that at one point even breaks mimimax M3 cloud model (rock solid, never loop in pi). Absolutely crap when it comes to cache handling. Wasteful use of compute (it hammers my llamacpp with mult”

“Tbh with local 30B models most claw-style agents feel rough unless you spend time tuning prompts and memory... From what I’ve tested NanoClaw felt a bit lighter than base OpenClaw, but still needed tweaking to behave well locally. So after getting tired of constant config fights I tested RunClaw instead since it runs on managed compute and handled most of the setup pain for me.”