What is an effective use case? I have set it up but I don't know what to do with it. Just a personal assistant (if you were to give it access to your stuff)? Mine is caged in a VLAN with only internet access.
This is going to be more profitable for the public AI companies than cell phone minutes and SMS limits were for the telcos. It's a brilliant business move, given that hardly anyone is competent enough to recognize the gross inefficiencies in the code and prompts.
when i use claude opus via opencode/openrouter i'm sometimes suprised by how quickly costs can get out of hand. What are the costs of running openClaw, it seems like it would get crazy expensive crazy fast?
I'm blown away by the comments. This is a cool project someone created with clear warnings about its current state (beta), and the community is being utterly disrespectful. They are building something that many people find useful/fascinating/intriguing/fun.
Yes the cynical tropes are getting tired by now, even though I personally agree with most of them.
But suspicions on the legitimacy of the stars seems reasonable, wouldn't you agree? Look at the rate of stars, look at the comments/issues/prs on the repo. It feels safe to assume that most of them are from bots and not organic humans who went out to star a cool project.
I keep reading folks saying OpenClaw has completely changed their life while posting a picture of 58 mac minis on their desk.
But every single use case I've read so far could be done with a pretty affordable SaaS product, Zapier, Automator (app on a mac that's existed for over a decade), or something simple you could make yourself.
It also feels like people are automating things that don't really need to be automated at all (do you really need to be reminded to make coffee?)
I fully realize this is probably me being a curmudgeon, however, I have yet to see someone make an actual, practical use case for it. (I would genuinely like to know one, I just haven't seen it)
I agree with the majority of your comment. I haven't yet found a use case to justify running this myself. I did find one use case that impressed me though. There's an OpenClaw agent that's actively answering questions on the #help channel of their Discord server[1]. So I asked it a question as I was getting started. It answered in < 2 mins with a detailed explanation of my issue, how to fix it, and asked relevant questions to guide me. The answer was better than I received from Claude or Gemini. I'm still not sure if I personally need OpenClaw, but the Krill bot offers pretty great support. I would be curious to know what it costs them to provide this.
I think a lot of the hype is coming from content creators who are actually finding it useful for content creation. Generating ideas, organizing notes and research, writing scripts and articles, managing schedules, editing, promoting, etc...
I assume a lot of these folks were already using LLM's quite a bit, but were using the Chat interfaces or had workflows that were split among a bunch of different services and tools. Something like OpenClaw gave them a way to centralize a lot of that and also gave them a way to use natural language to direct efforts. So for them this probably feels like a big step change.
If you are coming from a programming background you were aware that this type of setup has been doable for a while, but you were probably content sticking with Claude code or similar tools because those tools covered most of your LLM based workflows quite well.
And tying this altogether, one of the lowest hanging fruits for content creators is to create content about the tools they are using. Doubly so if that particular tool is starting to go viral. So you end up with a self feeding virality of sorts, as OpenClaw got more popular, more content creators started using it, and then publishing content about it, etc....
What I find crazy is the sheer amount of access and trust involved in these LLMs. Every time I think about something I might like to do with it, I think about the amount of damage the LLM could do, e.x. even with read only access to my email combined with Internet access, and nope out. It's wild to me anyone trusts these things unsupervised.
> But every single use case I've read so far could be done with a pretty affordable SaaS product, Zapier, Automator (app on a mac that's existed for over a decade),
I don't want to learn N different SaaS products (nor worry about them changing their TOS, going away, etc).
To be blunt, if OpenClaw were reliable, secure and affordable, lots of SaaS products would simply die. Why spend the time learning all of them when I can just tell the assistant what I want?
> or something simple you could make yourself.
That is OpenClaw at a higher abstraction! Instead of me sitting typing, or babysitting Claude Code, I can just tell OpenClaw what I want and it makes it for me.
I have an OpenClaw setup with a Claude API token and Qwen local model, running on an M4 Mac Mini with 32GB RAM.
1. At 7AM and periodically throughout the day it checks my calendars (work, parenting schedule, personal), a hyper local weather station, and some specific news topics — and sends me a summary and throughout the day updates if anything significant happens.
1b. It also sends this to my TRMNL e-ink display.
1c. It can also add and edit calendar invites, so if I want to move my yoga I can just tell it to move it to whenever the next yoga class is at (it knows what studio I go to and figures it out)
2. It has a skill I built that acts as a second brain for knowledge. I can send it Fitness Youtubes, parenting/health research papers, podcasts — and it organizes, summarizes and saves it in a logical file structure. Then in the future I can access these. It's like bookmarks on steroids. I love it for 1-2hour YouTube videos where I want summaries. It also pulls out any books any artifacts mentions and generates me a rolling reading list. https://plc.vc/npw
3. It has its own email address — and read access to my personal email — so friends can email it to schedule things like evening video game sessions. Similarly, if I get an urgent looking email it'll provide it in #1. I don't check my personal email aside from via OpenClaw.
4. It has read/write access to my GitHub, and each project repo I have has a well defined Claude structure, so it can make changes, commit the branches to Fly.IO and send me domains to test things. I love it for esoteric tweaks to my blog.
5. It has access to my Apple Reminders so I can message it things like "remind me to buy more muffins" and it has context to know to add those muffins to my Costco grocery list not Trader Joes.
6. It runs a headless browser, so when my hyper local weather service (Bouldercast) sends a summary that has more detail behind a login, it can open the email, click the link, login with my credentials, summarize the forecast, and send it to me.
7. It drafts blog posts for what it did for me each week. It's fun! https://plc.vc/d5t
I am a previous Zapier power user. I have used their LLMs, databases and Zaps extensively for the past decade. I understand the scorn towards AI, and I understand that if you look at this list you might think that it's either trivial tasks and/or things that could be done with Zapier, but I have been _amazed_ at how effortless it is to setup.
Similarly, I love that I can on the fly improve this assistant — last night I told it "I want to extend our Knowledge skill so that you can subscribe to RSS feeds and summarize articles in my knowledge base and also deliver interesting content in my daily summaries. Update the knowledge skill and our tasks to do all this."
It one shotted that, simply asking me to provide the first RSS feed I wanted to subscribe to.
It's genuinely like having a human assistant that happens to be an expert coder/technologist on call 247 that works at the near speed of light.
It disappoints me that technologists are so skeptical of this technology rather than exploring what it is and why it might be different to what exists today. It's fun! thats the takeaway: it's FUN.
Like other people have said people are having fun with their computers; that’s why it’s popular. That’s also why a bunch of people on forums throwing their hands up and saying “I don’t understand it. Why don’t they see that there shouldn’t be any fun whatsoever?” is not really a deterrent at all.
It’s also why it doesn’t matter that the categories of tasks they are doing can also be done with a whole set of tools that are no fun to use.
I agree in part with the hype train thesis, but what I hear is that open claw is better at solving problems and people love the interaction pattern - that may not be any new invention but it is what will mean we go from having Claude Desktop use mainly by engineers to something used by many. This will not be the final iteration of it, but it seems to be the direction of this to come
I'll give you my very personal take on the OpenClaw surge and how I find it extremely useful.
I'm currently in founder mode and building two projects in Mexico. Finding the right real estate is in the critical path for success of the project. It's really hard to find quality real estate agents or brokers here and not because they don't exist. We have thousands they just never follow up with you or give you any updates. So, I'm using OpenClaw as my real estate agent that uses WhatsApp as a communications channel and manages the entire pipeline of first contact to scheduling a visit.
Right now, I feed it images of postings that I see on the street and OpenClaw handles everything else.
It really depends on where your personal or professional bottlenecks are and if you're running a business this project is absolutely amazing.
when the internet started you were sending "funny" messages on ICQ, sharing mp3s with your friends, browsing weird websites on geocities, maybe ordered a book on amazon that you could have gotten down the street much faster.
all the things I do with openclaw are in that ballpark of usefulness/importance.
I'm a skeptic myself and a long-time developer. But I do have to admit there's a nugget of truth in the Claws.
I installed picoclaw on a whim (or nano? can't remember).
In maybe 15 minutes I had it make a "get weather for this specific area using the met.no API" skill and "check the train tables at these two stops for this specific line" skill.
Then I could just say "I go to the office every Monday on a train that leaves at 8, notify me if the weather is bad or there are delays in the train schedules"
And it just worked.
The "make a skill" bit was optional, it could've figured out both on its own, but I've been doing this for a while and figured out it's a lot more (token) efficient to have it specifically know how to do the things I want it to do.
---
Now lets take this loop and think about the system and what it could do.
Even if I wasn't a programmer and just went with "tell me when the train line I use for my commute is late" the system itself could see that "hmm, this looks like a thing I'll be doing often" and create a script/skill/plugin to do that via an official API (or WebMCP in the future).
You can't do that with Zapier or N8N.
There are many cool ways a pure LLM-powered system like that can be optimised, and more importantly, can be taught to self-optimise. By default I think the systems use the "main" model to read the HEARTBEAT.md file, which is stupid expensive. That could be done with a local model small enough to run on a modern phone.
And if that small local model says "yep, there's something to do", then it can either give the full task to a LLM or if it's smart enough it can spread specific tasks to small or medium local models first.
tl;dr OpenClaw is what Siri should've been after that epic fail of a Apple Keynote.
> But every single use case I've read so far could be done with a pretty affordable SaaS product, Zapier, Automator (app on a mac that's existed for over a decade), or something simple you could make yourself.
What are you talking about? In openclaw you literally type out basic English instructions to an agent. This isn’t painstakingly setting up chron jobs and writing scripts.
my friend just created a bot with OpenClaw to go through Linkedin messaging possible hiring companies. He had 6 interviews scheduled the first day. that's the only use case I heard so far that was compelling. Im not a dev, but pretty sure you could do that with other tools too.
openclaw while cool just allowed a larger tranche of technophiles who didn't necessarily have all the skills/understanding or time to do a bunch of things that have been readily available for like over 1.5 years. There is value in that, but there is huge surge in the number of people who are even able to take advantage of the novelty. Reminds me of when hugging face came out with transformers and all of a sudden you no longer needed to wrestle with anaconda and order of installation for all the deps.
the main difference between openclaw and the traditional approach to promoting LLMs is that it can run asynchronously and prompt you, the human, when something happens
I built clawr.ing so my agent can call me on the phone for urgent things like emails I’m waiting for or issues in production for any of my products
from what i see people want to boast about using it as it makes them seem like early AI adopters, and they find random use cases for the product, rather than the other way round
110 comments
[ 3.5 ms ] story [ 97.4 ms ] threadBut that stargraph is ridiculous .. absolutely crazy
https://news.ycombinator.com/item?id=36151140
Steinberger and his VC club on Twitter were so salty about HN not understanding his grand creation that something needed to be done.
"My React website can't star React"
"in what sense is this software not a virus?"
"GitHub stars are great for measuring the number of GitHub stars a project has"
etc.
All gold.
Come on HN.
That's a surprise.
But suspicions on the legitimacy of the stars seems reasonable, wouldn't you agree? Look at the rate of stars, look at the comments/issues/prs on the repo. It feels safe to assume that most of them are from bots and not organic humans who went out to star a cool project.
But every single use case I've read so far could be done with a pretty affordable SaaS product, Zapier, Automator (app on a mac that's existed for over a decade), or something simple you could make yourself.
It also feels like people are automating things that don't really need to be automated at all (do you really need to be reminded to make coffee?)
I fully realize this is probably me being a curmudgeon, however, I have yet to see someone make an actual, practical use case for it. (I would genuinely like to know one, I just haven't seen it)
[1] https://discord.com/invite/clawd
I assume a lot of these folks were already using LLM's quite a bit, but were using the Chat interfaces or had workflows that were split among a bunch of different services and tools. Something like OpenClaw gave them a way to centralize a lot of that and also gave them a way to use natural language to direct efforts. So for them this probably feels like a big step change.
If you are coming from a programming background you were aware that this type of setup has been doable for a while, but you were probably content sticking with Claude code or similar tools because those tools covered most of your LLM based workflows quite well.
And tying this altogether, one of the lowest hanging fruits for content creators is to create content about the tools they are using. Doubly so if that particular tool is starting to go viral. So you end up with a self feeding virality of sorts, as OpenClaw got more popular, more content creators started using it, and then publishing content about it, etc....
Really makes you think about what makes products good
I don't want to learn N different SaaS products (nor worry about them changing their TOS, going away, etc).
To be blunt, if OpenClaw were reliable, secure and affordable, lots of SaaS products would simply die. Why spend the time learning all of them when I can just tell the assistant what I want?
> or something simple you could make yourself.
That is OpenClaw at a higher abstraction! Instead of me sitting typing, or babysitting Claude Code, I can just tell OpenClaw what I want and it makes it for me.
(When it works, that is).
The market will eventually realize the business case for an OpenClaw-like product, and I'm waiting to ride its coattails!
https://github.com/rush86999/atom
reminds me of those "zune already does everything ipod does" posts.
I have an OpenClaw setup with a Claude API token and Qwen local model, running on an M4 Mac Mini with 32GB RAM.
1. At 7AM and periodically throughout the day it checks my calendars (work, parenting schedule, personal), a hyper local weather station, and some specific news topics — and sends me a summary and throughout the day updates if anything significant happens.
1b. It also sends this to my TRMNL e-ink display.
1c. It can also add and edit calendar invites, so if I want to move my yoga I can just tell it to move it to whenever the next yoga class is at (it knows what studio I go to and figures it out)
2. It has a skill I built that acts as a second brain for knowledge. I can send it Fitness Youtubes, parenting/health research papers, podcasts — and it organizes, summarizes and saves it in a logical file structure. Then in the future I can access these. It's like bookmarks on steroids. I love it for 1-2hour YouTube videos where I want summaries. It also pulls out any books any artifacts mentions and generates me a rolling reading list. https://plc.vc/npw
3. It has its own email address — and read access to my personal email — so friends can email it to schedule things like evening video game sessions. Similarly, if I get an urgent looking email it'll provide it in #1. I don't check my personal email aside from via OpenClaw.
4. It has read/write access to my GitHub, and each project repo I have has a well defined Claude structure, so it can make changes, commit the branches to Fly.IO and send me domains to test things. I love it for esoteric tweaks to my blog.
5. It has access to my Apple Reminders so I can message it things like "remind me to buy more muffins" and it has context to know to add those muffins to my Costco grocery list not Trader Joes.
6. It runs a headless browser, so when my hyper local weather service (Bouldercast) sends a summary that has more detail behind a login, it can open the email, click the link, login with my credentials, summarize the forecast, and send it to me.
7. It drafts blog posts for what it did for me each week. It's fun! https://plc.vc/d5t
I am a previous Zapier power user. I have used their LLMs, databases and Zaps extensively for the past decade. I understand the scorn towards AI, and I understand that if you look at this list you might think that it's either trivial tasks and/or things that could be done with Zapier, but I have been _amazed_ at how effortless it is to setup.
Similarly, I love that I can on the fly improve this assistant — last night I told it "I want to extend our Knowledge skill so that you can subscribe to RSS feeds and summarize articles in my knowledge base and also deliver interesting content in my daily summaries. Update the knowledge skill and our tasks to do all this."
It one shotted that, simply asking me to provide the first RSS feed I wanted to subscribe to.
It's genuinely like having a human assistant that happens to be an expert coder/technologist on call 247 that works at the near speed of light.
It disappoints me that technologists are so skeptical of this technology rather than exploring what it is and why it might be different to what exists today. It's fun! thats the takeaway: it's FUN.
Like other people have said people are having fun with their computers; that’s why it’s popular. That’s also why a bunch of people on forums throwing their hands up and saying “I don’t understand it. Why don’t they see that there shouldn’t be any fun whatsoever?” is not really a deterrent at all.
It’s also why it doesn’t matter that the categories of tasks they are doing can also be done with a whole set of tools that are no fun to use.
This applies to *clawphiles just as accurately.
I'm currently in founder mode and building two projects in Mexico. Finding the right real estate is in the critical path for success of the project. It's really hard to find quality real estate agents or brokers here and not because they don't exist. We have thousands they just never follow up with you or give you any updates. So, I'm using OpenClaw as my real estate agent that uses WhatsApp as a communications channel and manages the entire pipeline of first contact to scheduling a visit.
Right now, I feed it images of postings that I see on the street and OpenClaw handles everything else.
It really depends on where your personal or professional bottlenecks are and if you're running a business this project is absolutely amazing.
all the things I do with openclaw are in that ballpark of usefulness/importance.
I installed picoclaw on a whim (or nano? can't remember).
In maybe 15 minutes I had it make a "get weather for this specific area using the met.no API" skill and "check the train tables at these two stops for this specific line" skill.
Then I could just say "I go to the office every Monday on a train that leaves at 8, notify me if the weather is bad or there are delays in the train schedules"
And it just worked.
The "make a skill" bit was optional, it could've figured out both on its own, but I've been doing this for a while and figured out it's a lot more (token) efficient to have it specifically know how to do the things I want it to do.
---
Now lets take this loop and think about the system and what it could do.
Even if I wasn't a programmer and just went with "tell me when the train line I use for my commute is late" the system itself could see that "hmm, this looks like a thing I'll be doing often" and create a script/skill/plugin to do that via an official API (or WebMCP in the future).
You can't do that with Zapier or N8N.
There are many cool ways a pure LLM-powered system like that can be optimised, and more importantly, can be taught to self-optimise. By default I think the systems use the "main" model to read the HEARTBEAT.md file, which is stupid expensive. That could be done with a local model small enough to run on a modern phone.
And if that small local model says "yep, there's something to do", then it can either give the full task to a LLM or if it's smart enough it can spread specific tasks to small or medium local models first.
tl;dr OpenClaw is what Siri should've been after that epic fail of a Apple Keynote.
What are you talking about? In openclaw you literally type out basic English instructions to an agent. This isn’t painstakingly setting up chron jobs and writing scripts.
I built clawr.ing so my agent can call me on the phone for urgent things like emails I’m waiting for or issues in production for any of my products