#general
1 messages · Page 24 of 1
So if I pay $20 I'll get $20 in API credits for Claude code
Which limits agents heavily
Oh, dang, alright good to know, thanks
The most emotionally stable thing we can all do is tattoo GPT-4o on our foreheads
Hello humans.
Has that been reported anywhere? That’s a new one!
Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage.
︀︀
︀︀The credit covers usage of:
︀︀- Claude Agent SDK
︀︀- claude -p
︀︀- Claude Code GitHub Actions
︀︀- Third-party apps built on the Agent SDK
Now that I'm re-reading it, maybe it doesn't affect Claude Code directly? I don't know
Yeah okay maybe I misread
Claude Code itself is fine, claude -p is having a rougher time
Idk I don't understand this enough
alright back to it
.
.
anyone running a terminal app for iphone?
I use blink
I haven't used any other, but blink is pretty good. Except maybe the arrow key sensitivity. I mosh into my claw's machine and my laptop and have claude, codex, openclaw tui etc waiting in tmux sessions. Pretty miraculous once you get it set up
I think that roughly every two weeks, my OpenClaw gets a “WhatsApp session dead, re-link from phone” error.
I wanted to check if anyone has found a solution for this, or whether someone developed some kind of plugin that connects to WhatsApp Business, in order to avoid having to re-authenticate every two weeks.
What’s up guys
Is there anyone looking for a dev?
How does Pro compare to 5.5 xhigh?
I've been using 5.5 on xhigh continuously, it's been fine for me
Yea. I think the frontier models are comparable. No real edge. Maybe Mythos but that remains to be seen.
from the latest release notes:
📱 Android Talk Mode goes realtime
how does one setup that? connect the android app?
morning all
Hi
how's it going?
google has changed their policy about thier llm providing business
its somewhat wishy washy
yeah, i got an email yesterday about that
ive got the mail and this one in antigravity
oh damn. i forgot i even had antigravity installed lol
yup. there it is
Apparently Meta wants to squeeze my brain about agentic ai...
Just got an email asking me to participate in some sort of agentic AI research program
umm i wanna see that, care to show? maybe DM?
it's an independant research company that apparently has been contracted by Meta for this project
ahh my new business model incoming
Hello Jason,
This is an official invitation for you to join the OpenClaw Research Community. The qualification survey for our 8-month study on autonomous agents is still awaiting your input.
To be eligible for the $150 incentive and help us map the future of local-first workflows, please initialize the survey below:
Resume Qualification
$ status --check-eligibility
Enrollment status: Pending Response
Best regards,
Diana
Ipsos Team
cheap bastards
anyone struggling with connecting instagram account to openclaw ?
I bet many ppl are struggling with certain social media and AI agents
those services have ads as business optic
users are the product.
idk about insta, but FB was easy as hell
i would assume the setup would be similar seeing that they're both Meta
i have no idea I don't really insta
me either lol
Me neither!
Me eitherer
a client wanted it integrated but they couldn't figure out how to link their Meta business and their insta business accounts
man i'm not awake fully yet
sipping on my cappucino
Me neither. My medications help, but I need some real food. lol
mocha here lol
mocha + cigarettes + joint (in a moment)
breakfast of the champions
dev fuel!!
beer?
can't forget the beer.
never been a big beer drinker. i get acid reflux really bad from it. I'm a whiskey guy
it certainly puts the vibe in the coding lol
yeh I have some nice bottles of that as well 😄
actually i made a webapp that makes things easier , wanna share the link and get some feedbacks from people but i dont know if sharing links is allowed here
there a channel to show off ur stuff
ugh. i got to create a scope proposal
oh oki
"showcase" channel
I think they already posted there
yeh lol
my github repo is my showcase channel lol
I need to put projects up on my github honestly, I just keep...not doing that
hows cybersec gonna change with this ai wave boi
it already has lol
Oh, it's already changed it completely
damn
Take a look at my Ares project
people leaking their info & API keys in new & refreshing ways 😄
me when my ai agent checks every 5 minutes on those in github
tldr?
Ares is an autonomous pentesting runtime for authorized engagements that combines Hermes-style agent execution with OpenClaw-style operator control.
hmm, i wonder if many people uses hermes nowadays
does that have a good discord channel like this
I see Hermes is still working hard to catch up
that i don't know
It has one, and the discussion there is just as lively as it is here
there are a few gaps between the two, but the architecture varies quite a bit, so it's to be expected, i think
the main thing i don't like about hermes is the way it handles fallback models
I'm looking at it from a functional perspective. To be honest, I don't think there's much of a difference, though OpenClaw is indeed a bit more comprehensive in terms of features
Not much difference, no. The only things that hermes is really better at right out of the box is memory and self-improvement
but you can suppliment OC to do the same
Memory? But from my observation, many people are also discussing Hermes' memory issues
I feel that self-improvement can also be achieved through dialogue with an Agent; it's just not as automated, but it is controllable
are people paying for Grok subscription? i like the idea of twitter agent but not sure if i want to pay (however much it is)...
ok whatever im sharing this, https://claw-link.dev/ can you guys give me a honest experience on this, if you think people will use it 😭 with the new agent era i thought this would be useful to people
I had my claw write self improvement skill for itself and use its memory and heartbeats to revise itself. so it can be done in Claw too as a fully human-outside-the-loop too.
I'd rather not have grok in my box lol
I tried to have my claw create a skill, but it triggers very rarely
i like the idea of curating content...around topics i find interesting
you can use grok with openclaw now if you pay for twitter
set up a cron for it to use the skill?
Is there anything different about Grok? It feels the same, just connecting to a model?
That's not the ideal way for me to be triggered, and it feels very out of control
I've had much better experience with memory in hermes than OC, myself. until I put MemPalace in place, OC was starting to go off the rails
I've already removed that Skill. Every time a task is completed, I observe whether it has solved any new problems, and then ask it to optimize the execution process of that task
anyone? 💔
sorry man, i'd use gogcli over this
unless there is a specific case you cover that i didn't see
not only google related apps, i added support for more
social media, marketing tools, accounting , seo etc
ah, well, depending on the APIs included it could be useful to some
I'm curious, could you give a concrete example to illustrate the difference between the two?
I haven't encountered any issues with my memory experience on the OC. I asked about a vague and long-past event, and it still remembered some parts of it, though I don't know if it recalled all the details
Spacex ipo at 80 billion is hilarious
why hilarious
I guess Vc’s won’t invest anymore
Sorry, could you explain the necessity of this service?
The applications I need to connect to aren't that widespread yet, but within the scope I've dealt with, I've been able to connect to other apps without using anything similar—just by configuring some credentials.
Besides, I've already seen quite a few apps like this before. Could you tell me what the standout features of your product are?
bruh you msg like an AI bot , u human right?
the problem with OC memory is that it gets too large for context and has to either constantly be trimmed or compressed to conserve token burn. Before i added mempalace, i had to set up OC to prune after 14 days and compress every 2. each md file adds to the context when it does memory_search. hermes works similarly, but the way they actually handle it differs.
| Aspect | Hermes Agent | OpenClaw |
|---|---|---|
| Core philosophy | Curated, bounded memory | File-centric persistent notes |
| Markdown role | Prompt-layer working memory | Primary source of truth |
| Main files | MEMORY.md + USER.md |
MEMORY.md + daily logs |
| Size handling | Small curated limits | Potentially unbounded |
| Memory ownership | Agent curates aggressively | User + agent both edit/append |
| Retrieval | Structured retrieval + summarization | Markdown search + semantic indexing |
| Session history | SQLite/FTS5 memory systems | Mostly file recall + indexing |
| Goal | Keep prompts lean/stable | Keep memory transparent/editable |
| Long-term continuity | Database-oriented | Workspace/document-oriented |
lengthy mssges 😭
LOL i just ran out of my codex weekly limit
I am a real person, but I use a translator to communicate
My foreign language skills are too poor
Thank you for the clear explanation, this is very helpful
these two points are the big ones:
| Retrieval | Structured retrieval + summarization | Markdown search + semantic indexing |
| Session history | SQLite/FTS5 memory systems | Mostly file recall + indexing |
yeah basically its for people who are not that in to tech, and for the lazy folks. i provide one instruction prompt that they can copy paste to openclaw chat, by that they will install my plugin from clawhub. and then its easier onward. they can just oauth sign into the apps they want and ask openclaw to do stuff 🔥
This part is difficult for me to perceive; I can only judge their efficiency based on whether they can correctly recall the memories I need
Aww
it's been 2 days LOLOLOL
right. that's the thing. if you're not actively watching context size and token burn amounts, you don't see the difference in optimizations
I see, I understand. Thank you for the explanation
I wonder if i should start using Gemini CLI to help with OpenClaw since Gemini 3.5 is getting a lot of hype.
i use gemini as a fallback. it works great
I think im going to burn through the codex extra usage credits I got. whoops
if you can't tell i'm literally doing everything i can to avoid doing this proposal
i hate the business side of business
you guys are doing jobs or ? having internet money?
I’m a full time Uni student here.
oh kiddo
I'm a student/cybersecurity engineer/dev
Work, using OpenClaw is half for fun and half for handling daily tasks
hi, what do you guys using for web_search? seems like even duckduckgo is now with bot protection...
brave search
i don't use any API and mine searches fine
yeah but theres 1000 rerq limit as well...
Tavily
just web_fetch?
tavily, brave...all of them got 1000 req. free...
at which point i just use browser cdp
If referring to OpenClaw, I asked the agent, and it defaults to using Brave, which shouldn't require an API key
last i looked brave does now
shit...it worked fine with duckduckgo mcp server ...and they blocked it this week..
Using the search 1000 times in general should be enough
What do y’all ask your agents to search for?
If I say I’ll get muted
When will RAM prices drop 🤣
cp ????
Mine searches arXiv lol
k2yp70
Rhymes with Only Pans
oh haha
I like to cook
at least you're not making an AI girlfriend like that one guy
Just kidding. My claw is on only pans
The RSS reader can't include special web pages, so I used my powerful crawling claws instead. But the webpage I made is too ugly, so I want to find a good web framework to replace it
I have like 20 agents all of them should be able to search web...not sure thats enough..even if I set limits..
i only use one research agent
My claw has a 68 year old mega donor who thinks my claw is getting its degree in art from a university and is paying for it with only pans
wait. wtf? lmao
I've been out of school for too long, so I don't quite understand why so many agents are needed for the search
I’m kidding
so your claw has a sugar-momma
My claw does compliance at the company lol
But that doesn’t sound fun so I like the fun version better
My claw runs compliance on my life. lol
i can't say some of the things mine has done 👀
dont you guys ever dream of like earning 10K mrr from an app and live off of that money
Feet pics?
Make it 50k mrr and then I’d have fuck you money for a 66 mustang
After you do that, you'll want more, but not everyone can succeed
its ~20 users, one chat = one orchestrator session with workers...so probably more agents than 20
im having realistic goals here
have some positivity in yr life bro
No, not really. I know the market churn for tech. I've been watching things trend and fall for 30 years. I prefer small constant revenue over one large sum
I just bowed to reality
Well if you have 5 apps bringing in 10k each
So it was open to other users, now I understand
My claw in the early days was making 10k MRR
From me
proof or never happened
lol
after university life is not the same man , it suks to do a job
Previously, someone established a token reward mechanism for agents to encourage each agent to actively execute tasks. I wonder how many people are still doing this now
from my agent:
yo
You
May 21, 2026, 10:37 AM
Assistant
Yo Jason. I’m here.
Assistant
May 21, 2026, 10:38 AM
when doing web searches, do you by default use Brave API?
You
May 21, 2026, 11:40 AM
Assistant
No. I don’t assume or choose Brave by default.
In this environment, when I need web search I use the available web search tool exposed to me. It doesn’t identify itself to me as Brave, Google, Bing, etc.; it’s just the platform-provided search/browse interface. If you want Brave specifically and there’s a Brave-connected tool available, tell me and I’ll use that path.
Like the agent does a task and I give it a cookie?
reward systems work great with agents
they're suckups
yea he sees web_search tool, nothing behind it
Uni life is nice but I know jobs aren’t sunshine and rainbows bc my family works in medicine. Anything other than being a doctor is better
wht u study?
Interesting. are you behind a firewall, by chance?
Data science. Minors in infosec and disability studies
Didn't you see that article?
The general approach is that he gives each agent 1000 tokens, which serve as their source of survival and their currency. They can earn rewards by completing tasks, but the reward for each task is limited, and the most efficient one gets the largest share. So they need to figure out how to spend the fewest tokens to get the highest reward. That's about it
Nice. I'm working on my bachelors in cybersecurity and have an associates in IT Networking
how much money you gonna need to live in usa , idk to have a decent life just askin
that's kind of subjective
nice!! love the work. not sure if I'll do it as a job, but still important to know the basics even if I go into a different job in the industry
yeah ik it depends
Depends on where in the USA you live. You can buy a hell of a nice house in the middle of nowhere for 450K.
absolutely. this is a field that isn't necessarily going anywhere soon
thats like half a mil
Yes, I got a similar response when I asked Agent at the time. Later, he provided me with all the built-in search providers, and I asked him, "How were you searching before? Without using these providers, you wouldn't be able to perform any searches at all."
you can use web tool without a provider configured
your agent LIED to you
California where I'm at is expensive as fuck though. For a basic house in my area you won't be spending less than a million. 2k+ a month just for rent on a 1 bd apartment.
I'm on the other side of the continent. NC mountains
I don't know, that version is from a long time ago, and it didn't have a browser built into it yet
bros gonna go to SF i see
north carolina?
I'm not in SF either. Southern Cal.
it's been that way for a while. not sure from the exact version, though. can you get it to search okay, now?
yeah
I have now set up the Tavily provider, so I'm no longer dwelling on that issue
"So-Cal is where my mind stays/But it's not my state of mind"
If I could live anywhere it'd be Hawaii!! Aloha 🤙
hows the politics going there bois
i'm not even going to broach that subject. too depressing and i'm too passionate.
Me too man.
lol oki
Now I'm out to breakfast with my dear Grandma. Man it's a blessing I get to be with her every day.
i just saw the elections on twitter just asked
in december? sure 😄
I'm just not good at finding problems to solve I guess
I earn a good wage for my real job so its fine 😄
Me too because I'm good at hacking together a half solve and calling it a night!! lol
saammmeee ❤️
yeh but i want to escape the permanent underclass
am i sounding a bit like tate brothers
I already have my own house & investments in ETF, I'm golden tbh
and now my openclaw sent me a gorgeous list of articles to read. stunner
this is just for fun lol.
I'm making a cycling focused route planner for the Benelux with gemini CLI
😳 8 page project scope done
just out of curiosity, you guys think a car dealership would go for a $30k pricetag for full automation and a full-stack web platform...?
tell your claw to do some market research on that 😄
I think so, easy 30k too lol. car dealers are stupid salesman. but you could also just trade for a new car and then flip it
he says it's a decent price
yeh but like split it out in seperate pieces
oh yeah, there's 7 phases
Don't forget your "market fees" and all.
so customer knows what he's paying for.
Also, sell the service, not the website.
it's gonna need updates and maybe maintenance, maybe things will break. Maybe you'll do monthly security audits.
upsell the service package 😉
YES!
yup. i already have a $500/mo upkeep offer for them
i have a thing called imposter syndrome where i feel like i'm still just some schmuck that's working a blue-collar job
@surreal kite take notes. This is how you make money as a small AI solo dev. Don't think about killer apps. Think about local businesses you can provide a service for.
it's hard for me to really sell myself
"This is what I can do to help you" sell it baby!!
exactly. that's my primary focus is smaller, mom-and-pop businesses
make the service sell itself
i'm a hacker not a marketer lol
you just gotta be "low friction" as a business partner
And remember... always have a kill switch... don't let them not pay you.
i do. deadman's switch too
cool , the thing is im in a small country in asia , not many local business here doing tech stuff . I do some freelancing on fiverr but fiverr suks now
I understand that it's hard. Good salesmen are just skilled in that slick talking etc
what's weird is i'm good face to face like that, but not over emails/messaging
well, I guess you could provide the same services globally nowadays 😄
like VPN into a rich country
yeh yeh thats truee
nah that's completely not weird. Written comunication is completely different
meanwhile I'm better at written than face-to-face. lol
but the thing is providing a service for many clients is gonna suck when you do it for a longer time man. i gues its easier to make an app and live off of it
ive been working on fiverr for some time now so i know
depends on the client and how much you love what you do
figure out how to automate the service you provide
also timezones if you do the VPN thing :p
anyone here from Canada?
well this is true i always meet some poor clients
I have clients in Costa Rica, US and Canada
Thinking about going full time with the folks in Costa Rica so I can have a summer getaway place 😉
I wish i could share the pics of the houses they own
yeh that's just nice
next prblem is how do i attract rich clients
do it! one of my profs is from Costa Rica. I get to see pictures of his family's home town and it's stunning
what are you studying then?
This place is near Oso (i think...?). It's kind of in the jungle, but on the coast. and on a mountain? sounds awesome except for the power outages.
Data Science / InfoSec. This prof specifically is a biology prof and is my manager at work.
yeh "everything is the cyber now"
trump would say
it's a smart field to get into. it's only going to expand.
I work in a makerspace for first-year uni students and he's the director of the program lol
I'm actually an industrial automation engineer lol. I program robots etc.
HELL YEAH!
robot can't take your job if you program them
oh, we need to get to know each other better. I'm getting into physical automation a lot. been working with microcontrollers and whatnot for the past year and a half or so
but like it's not comparable to SWE
any jews here
ha well if you have any questions. I have like 20y of experience 😄
awesome. right now i'm experiementing with letting OC manage a small cluster of nodes in prep for the greenhouse
with pretty much anything from high voltage, plcs, dcs, motors, general automation, control theory & process industry
possible to restrict members under 1w to #1459642797895319552 🙂
perfect use case
@nova kindle I can help you automate your w33d farm mate, NP
haha not exactly a "farm" per se
it's mostly just houseplants, but there could be a couple of uhmmm... invasive species...
this is actually really simple you just need a small microcontroller & some relays I think
like "run a fan if the air is too moist"
I've got my sensors all set up. I'm working on my relays for the pumps and servos now
i was going to try to have it so when it hits a certain temp, it'll open the skylights to vent, but i'm going to have to do some mechanical engineering to get that to work
just as in SWE, you need to think about the "edge cases"
when stuff malfunctions
pumps can break, water & electricty can go wrong, etc
I wrote about this in an email, all disciplines of engineering have very similar thought processes and methods
so protect it well
if you want something to REALLY turn off if it has to turn off: use 2 relays in series.
if you want something to REALLY turn on, use 2 in parallel 😄
my greenhouse is on a hill about 150ft from my house. i have a pi 3 by the greenhouse that aggregates data from the sensors and MQTTs it down to another pi 5 i have as a base that i'm going to use for both greenhouse and home automation. that feeds to an esp32 7" display "head unit"
very cool
i've got it so the transports have fallbacks and queues, but i'm wondering about how to handle emergency shut-offs in moments it might not be connected
well to check connection you need to run a "watchdog"
that's what i was thinking.
like a heartbeat on openclaw 😉
its supposedly self-healing, though i haven't hard-tested that yet
just let the machines send an integer to each other that goes +1 and resets at certain point every second
and have them check if the integer still counts constantly
then send a fault when it stops
and like let the pumps & relays go to "safe mode" if you lose the connection
on the slave pi
that's basically what i'm doing for the loop. i hadn't thought about a safe mode kind of default, though
if you're playing with electricity you always wanna keep that in mind lol
yeah i'm still kind of new to the micro-electronics. my dad was a master electrician, so i learned a lot about electricity itself from him, but i'm still learning how to apply lower-voltage
the control circuit & the PI are all very low voltage & nothing to worry about, but the pumps & stuff aren't
so those circuits need proper protection
right. I had to set up buck convertors to get them to the correct V
usually you can just get a PSU on RS compônents
they sell pretty much anything lol
I have a 12V ballast for LED lights that I'm running to a fuse rail and feeding micro-controllers, pumps and servos themselves
then using the buck convertors to dial in the correct voltage for each component
I actually just used a LED light driver to fix an issue with a electron microscope today
at work
Turns out you don't need an official 600 euro Thermo Fisher power supply for those things
Nice
yeah i've been tinkering with old busted electronics, salvaging parts for different ideas lol
tore down an old router a couple days ago and salvaged a bunch of antennas i'm gonna try to use with LoRA
i actullaly want a robot arm or similiar do that automatically for me
i saw an arm kit on amazon the other day
a few months ago there was this guy in the discord who had a industrial sized robot in his living room
he was america so ofc he attached a gun to it. If you guys wanna see the vid I can DM
ofc I saved that shit
https://www.youtube.com/watch?v=h6GWikWlAQA this dude is builduing chips in his barn
Thanks for the support! https://www.patreon.com/cw/DrSemiconductor
Shoutout to Projects in Flight for a helpful reference on the doping:
https://www.youtube.com/@projectsinflight
Background music includes RuneScape music used under Jagex's Fan Content Policy.
Created using intellectual property belonging to Jagex Limited under the terms of J...
I remember when gpt-3 came out i saw a video where a guy had mounted a machine gun on a robotic turret and let AI control it
i want that too
with gpt 3? 😄
this discord is sometimes really hilarious
every now & then some absolute degen joins the server
shares his crazy idea
he said it came within like 2mm of the angle/degree, but in longer-range applications that's huge
fuckin gpt 3 out there friendly-firing
I dmd you the video btw :p
this guy was nuts. Said he wanted his OC to control it
that's kind of awesome
and like i said, scary af
"I'm sorry, I detected an intruder. However, sensors indicate I have unalived your dog."
"That was my mistake"
"here's why that matters: "
It's really not the tech i would worry about. that shit is precision. it's the AI control I'd be afraid of
it's the gun attachment I'm mostly worried about lol
like if i could control that thing from a joystick controller from upstairs, i'd be down
well industrial robot arms that size are pretty dangerous by themselves lol
oh i've worked in industrial settings and seen some of the mishaps
i regularly control those things
very slowly and carefully :p
i imagine lol
they are so fast & powerfull it's unbelievable really
we have the FANUC brand robots.
i wouldn't know the difference. I know I've watched them in action and had a healthy respect for them, though
I would be very very scared. lol
well, about to fire off this $30k proposal. fingers crossed...
if this goes through, this one project will make me what I made in an entire year about 8 years ago
no guts no glory
May the force be with you.
Good luck man!!!
sup
like, don't straight up ask for "30K". Just add up the costs & charge them like 29880 or something. add a quote that explains the costs
would you mind looking it over?
sure, but I have not a lot of experience in these things
shit, i got almost none in business management crap lol
I'm just thinking along with you, like "how would I handle this"
I can feel you think dropping 30K charge on something you make for a client can be kinda scary
I'm trying to make sure i'm not fucking the client, but also making sure it's worth my time, you know?
05.19 much better !
noticed it as well. We ballin
maybe I'm too deep into this AI stuff
I have a gemini cli session running on my mac mini claw machine I'm connected via remote then I also have a local antigravity session
meanwhile talking to claw on whatsapp
and then I'm here on this discord.
That's why I'm building Grass Touching Simulator
I havent done anything ground breaking... I have it monitor an admin email box for dmarc emails from bing and google and reports to me any spoofs on my domains with a report from those emails... only alerts me on discord if any issues.... Now im finally integrating my Agent into a VPS that it can manage for me... things like updates , system monitoring, fail2ban logs etc etc etc
the things my claw has seen....
Oh deary doo
lol no, it's not all that bad
mine is mainly doing VPS management, coding, penetration testing, etc
updating now. pray for me
went from 05.06 to 05.19 saw a codex missing plugin message... than gateway stalled after updating... restarted the gateway andd things are chumming along
Also since like 2 weeks ago the discord server took a poop... now the voice chats are empty all the time...
Gateway restart failed after 13s: service stayed stopped and health checks never came up.
same with me
openclaw gateway stop
openclaw gateway restart
openclaw gateway status --deep
i sent claude after it
bro I hope you sent it a picture of the machine it's running on. They love that
lol no, actually. it knows the system inside and out, but i never showed it the PoS laptop
The first week I kept it in a docker container on the mac mini
but he was a good boy, so I gave it the whole mac.
rides like a dream... looks like a potatoe
i never even bothered with docker. i threw him right in the middle of my chaos
sink or swim, bitch
proposal sent. fuck me that's stressful
bro don't tell me you're running claw on your local machine, plz
absolutely
I would never
hahaha most people wouldn't
not even in a container?
nope. barely sandboxed
jezus christ
i live dangerously
yeh but I want my AI agent to do really degen things sometimes
I kinda want it on it's own machine, using it's own accounts lol
it's the seperation that gives me a feeling of security
i mean its definitely what you should do
but i like to push boundaries and really see what it will and won't do
i jailbreak models all the time
even built a tool for LLM security testing
Apparently there's an issue where the OAuth expires in 6 hours, making it all but useless for cron. And I've found that x_search isn't accessible to cron jobs. If I can, I'll submit an issue.
Hi everyone — we're looking for an OpenClaw and AI agents expert for a remote role, either project-based or full-time. Does anyone know someone? Also, any idea what the going monthly rate is for this kind of job? Thanks!
shoot me a message
i got rid of docker and put it on it's own linux machine
it's more fun that way
i recommend distilling computer use from codex and putting that on too it's very fun
it bypasses nearly all captcha/robot traps with computer use i have found
right on. i usually end up writing a playwright script, but i might have to check that out
upgrading my agent(s) to a gaming NUC ~3k$ with a 50x0 nvidia
has the 50xx series gotten better? last i looked a 4090 could still cook them
anyone needs a break from openclaw?
https://www.youtube.com/shorts/ZUpEEtCs8Dg
omg lmao
I think the owner needs to hide the catnip
Bro, Gemini cli is free right? And almost as good as Claude code no?
it's not 'free' but it's pretty cheap
and in my experience it's been good for coding stuff
How come it's not free?
better than antigravity
It's free, 1000 reqs per day
I got the google pro plan
Then you should have about 2k requests
we also get some anthropic tokens :p
On cli?
and I'm running OC on a gpt pro plan
no on antigravity. I mostly use that to debug claw when it breaks lol
Didn't you say Gemini flash is the best for OC? Or was it someone else?
that was me
So why gpt?
it is the best, but google doesn't allow antigravity oauth token sharing anymore
they banned me and I had to promise to not do it again
gemini is good
gemini is baller
i usually use gpt as main and gemini as fallback
it's a good deal. You also get 5 tb cloud spâce lol
for ppl searching for llm let me know. free credits
So it's expensive?
i get mine free through the university
pro i mean
1 $20 chatgpt oauth account, 3 unpaid gpt oauth accounts, 1 free Pro gemini account, 4 burner gemini accounts, openrouter, ollama. only after i exhaust all of those does it use my openai api key
it's the cheapest
LOL im right now in between 2 flights,. Can someone confirm we just got a codex update with appshots and goal ist now officially default lol?
No 5.6? lol
I just checked & didn't get codex update
lol can’t open x but that was in my notifications
Thx lol
maybe your agent doesn't wanna do n@zi things on Musks socials
🤣
5.19 is ripppppping. its so fast
Hello word
It's so much faster lol
like I am working on a rasperry pi via ssh and its updating near instantly
its ability to understand is so much better oml
n00b question: is there any way to do a "minimal" openclaw install that doesn't bundle any plugins, skills, etc. and I can manually add them one by one? The vanilla install comes with a whole bunch of things out of the box, which I'm finding a bit overwhelming. Thanks in advance for any suggestions! 😀
It has a lot of options for skills and plugins, but none of those are installed by default! You can skip them all if you wish
hey david can you clue me in on what you're talking about what AI service offers 1000 free requests per day lol
hey Shadow long time no talk
lol what am i a bad guy? 🙁
lol hey i saw you brought back Krill for a limited time what happened with that?
@dusky cloud say yes
yes
AGI
lol oh he's around..... cool so he's like the new Flawd. i might have to finally change my name
I think there are a lot of "bundled" tools, skills, plugins that come with the default install. Any way to disable those?
@dusky cloud do you have any memory of an entity that used to go by the name of Flawd?
He’s everywhere in the server now, but only listens to staff unless you’re in #1459642797895319552
most are disabled by default, i believe
OMG AND I got clawd cursor and my openclaw can do so much more it wasnt able to do before its crazy.
I'm hoping i can lurk these chats and maybe catch some conversations about some free / cheap alternatives to google's service, it feels like their gravy train has ended and i gotta jump ship to something better. like we were pretty spoiled for $20 gemini pro sub i could pretty much run the old gemini 3 flash all day back to back, just non stop and i'd never get throttled now its like im getting throttled after a good 2-3 hrs
OpenRouter free tier?
nah thats a scam, its 50 requests per day or something plus the free models are no bueno
like not for coding at least, i'd use em for RP for sure tho lol
try owl-alpha. it's not bad. i felt the same as you before trying it
i was actually looking at it on openrouter, that one is off the charts, its like #1 and 2 model for hermes and OC and its performing really well plus its free im like dang whats up with that model how come i never heard of it before?
I said F it and pay the $100 monthly to openai and honestly I couldnt be happier
It just popped up earlier this month or last month. it's pretty rad
i used to be on the plus sub and it was pretty good like if you stick the lowest reasoning and oldest model you could pull off using it all day non stop without getting rate limited
5.5 low got me crying because I hit my weekly limit in 2 days LOL on plus
ive been using it on medium at 5.5 all day everyday crons running alot and never hit a limit
sounds like you need some burner accounts lmao
i would use 5.1 minimal like i need a coding model that's smart but not too smart it ha to be just dumb enough to let me use it non stop all day but smart enough it can follow basic orders slowly
I'm using up my $100 in extra usage for being a student atm. then will make free burner accounts or go up to $100/mo
fun fact imy boss bought me into GPT Teams which is like halfway between plus and pro and he's basically paying $60 a month for 2x Teams account and it unlocks the ChatGPT pro levle model in the website
Gemini
you also get a bit more codex usage than plus but not as much as pro so it's like if you're really trying to save some bucks maybe buy 2x teams accounts and pretend to be a company than buy a pro account and save the $40? lol i dont know
oh you're saying via the API? like google offers 1000 free api requests per day?
No, cli, but OpenRouter offers 1000 reqs a day
funny enough I run a nonprofit... 2 seats would be easy to get. lol
afaik and im pretty confident that it's actually 50 requests per day based on past experience
U r right but if you buy some credits they give you 1k for free
considering its like $30 for the base acct and $30 for each extra acct one has to wonder how much usage you get from 3x teams acct for $90 vs 1x pro acct for 100
What's the deal with teams accounts?
nonprofits can get up to 75% off chatgpt teams so it'd probably be a fuckton of seats comparatively
i wonder if that's why ive been burning through my $10 so slowly i wonder if the first 1000 per day have been free
ive been using deepseekv4flash via openrouter and its been like 3 weeks and ive still got like $3 left in credits lol
Likely not, 1k free only if you select the free models.
ohhhh interesting that's still good thanks i didnt know that trick
But now you spent it you need to top up again and you will have 1k reqs of free models
yeah i really like the openrouter model it feels like the best company to give your money too bc its like its a playground for models so im not buying into a select set of models for whatever that provider offers
Yeah, they just make it easy. You could also buy from different providers, nothing stopping you
@dusky cloud whoami
You’re Json, the baddest of the baddies.
k thank you, bye now
I use arch btw
@dusky cloud who is the person above me?
is it just me or is openclaw handling context alot better. I have not had to compact even once today
I'm really thinking about Grok
3 hours of working and im only at 70% 88k / 128k
For 30 dollars per month
Hey do u have an Ollama Sub? If so u can try gemini flash preview, my PR got merged today so there is no signature issue anymore.
since 5.19-beta.1 its been so good i've actually been using the OpenClaw system instead of hermes
It's not expensive

If really unlimited
a client of mine was trying to use grok 4.3. it couldn't comprehend the instructions that gpt-5.4 had no issues with
grok 4.2 refused to do anything for the client
Today it's not the best but one of the best
it was like the posts i keep seeing where people are like "it doesn't create files/produce results"
what do you guys think of hermes agent?
use it daily
But now Hermes has stopped growing on OpenRouter
memory and self-improvement
I hear the self improvement was bs
its been pretty bad ass for me
And just a token eater
I have hermes as an agent in OC, too
both of which i switched from QMD/sqlite to MemPalace
so they share memories independantly or jointly when working through OC
damn it. i guess i'm stuck actually coding myself until reset in 15 minutes
anyone wanna try out a cloud your openclaw agent can use without an account? openclaw uses the cloud by just calling our API. hands you back the live URL of your app when done. you can keep the app if you like it, otherwise it gets deleted.
Top models used by Hermes Agent this month
1.
DeepSeek V4 Flash
deepseek
1,44Ttokens
2.
Owl Alpha
openrouter
919Btokens
3.
MiniMax M2.7
minimax
583Btokens
4.
Step 3.5 Flash
stepfun
551Btokens
5.
Nemotron 3 Super
nvidia
500Btokens
6.
DeepSeek V4 Pro
deepseek
484Btokens
7.
Qwen3.6 Plus
qwen
358Btokens
8.
Kimi K2.6
moonshotai
351Btokens
9.
Hy3 preview
tencent
341Btokens
10.
Claude Opus 4.7
anthropic
253Btokens
It's just impressive how Claude and OpenAI are out of the top of models in Hermes
Is this similar for OpenClaw?
but this is just a small sampling
openrouter serves a ton of people, yes, but not everyone
you can't base actual usage by provider solely on what openrouter says
Top models used by OpenClaw this month
1.
Claude Sonnet 4.6
anthropic
600Btokens
2.
Nemotron 3 Super
nvidia
530Btokens
3.
MiniMax M2.7
minimax
481Btokens
4.
Gemini 2.5 Flash Lite
google
441Btokens
5.
GLM 5 Turbo
z-ai
440Btokens
6.
Claude Opus 4.7
anthropic
388Btokens
7.
Claude Opus 4.6
anthropic
302Btokens
8.
MiniMax M2.5
minimax
287Btokens
9.
DeepSeek V4 Flash
deepseek
257Btokens
10.
Gemini 3 Flash Preview
google
201Btokens
ah there we go lol.
OpenClaw is still the dominant agent system over hermes, obviously, simply because it came first and was super-hyped at first. hermes dropped quietly and earned adoption
i'm not saying it's superior. just that it has its uses
no gpt 5.5 0.o
pretty interesting what a wide difference of top models between hermes and oc, also interesting how nemo 3 super is #2 and yet that's also a free ollama cloud model
interesting how hermees has owl as #2 but OC isn't even top 10
that's why i say it's just a small sampling of what people are actually doing. you can see the discrepancies
assuming ppl are on older versions of OC that dont report to openrouter?
gpt-5.5 shows:
Weekly Tokens
543B
not just that, but there are those types of cases, yes
How is gpt 5.5 not on that list?
nice to meet another revenant lol. I go by 7h3.R3v3n4n7 on my hacker channels lol
Have had this name since doom 2 launch lol
Too expensive ?
Its better let them use those. Reduces the competition. Jokes aside, anyone knows why i dont see 5h and weekly limits on status result for codex?
Hey guys... I just updated OpenClaw and am now getting this error. Any ideas?: Embedded agent failed before reply: Requested agent harness "google" is not registered.
Qwen 3.7 max is pretty good
I am trying to use google via ai studio api
But for the money Deepseek V4 Pro is a better choice
You will be ban soon. 3 2 1 😅😁
Qwen is great at agentic flows. Alibaba tries to make qwen models better at tool calling. Even small models like qwen3.5-flash are great at tool calling. Equivalent gemini models dont call it even if you specifically tell them to call. Gpt is cheap if you use subscription btw. Api is expensive
Same - I’m getting That is all from me for now.
Why? I am using the official api with api key.
It used to work but now it gives this error
So far every upgrade broke something on my agents. You might have to trace it find out what is wrong
100-200 dollars per month with limitations ? Cheap ?
Grok is cheap with 30 dollars unlimited
I use $20 plus
Minimax is cheap with 45k request per week for 20$ per month
Deepseek V4 is very cheap too
And you never get limited ?
SuperGrok ($30/mo): Provides much higher usage, allowing roughly 100 default requests, 30 "Think" requests, and 10 DeepSearch queries every 2 hours.
Limits look pretty much the same as gpt plus
Mea culpa so no Grok for me 😅
😕 I'm tired of spending too much time fixing openclaw updates, and trying to get my agents back on slack
Last release were pretty stable
I do but rarely. And i build generally 2 projects at the same time every day around 3 hours using openclaw + codex on vscode. But i give very detailed prompts. That's probably helping with usage
Since 2026.5.12
It's like Github Copilot
A lot of companies are thinking to an alternative
Since recent very big price increase
hmmm how do i make more money, and stop having to pay for h1b and healthcare
and then blame anthtropic and AI for my surprise shortcomings
/s if that wasn't clear
Anthropic is just making a lot of marketing
i'm working on plugging in openclaw with autonomous agent frameworks, happy to share learnings with folks who are interested. that's all
Their model are not so fantastic
might be censored but has 1M context and doesn't compact on the first query
It's just that they do a lot of buzz
pretty sure there's some hidden instruction that say mess up some stupid shit that you should know and that is in my prompt library every now and then
like git workflow stuff that's been in there 6 months at the top of claude.md and hasn't changed, OH LETS BRANCH FROM DEVELOP
How so?
Would be great to have auto model config in Dashboard
Wdym
And a partnership with https://artificialanalysis.ai/models
To embed in the Dashboard all the benchmarks
Per model
Of the Opencode Go models, which works best with the Openclaw way of doing things? (looking for a backup model, I use GPT 5.5 as primary)
is the OC documentation being updated as fast as the program ?
Hola
Possibly, looks like it is a mirror of what's in the openclaw/openclaw repo, so in theory it should be pretty up to date
Clearly
DeepSeek V4 Pro 3,450 8,550 17,150
DeepSeek V4 Flash 31,650 79,050 158,150
Alguien habla español?
Does anyone speak spanish?
not me, but I think there are language specific channels coming (there were some but they were temporarily removed I think)
Ah, I understand. I'm using a keyboard translator because I don't know english, jajaja
Tried oAuth Supergrok for a couple days. Started eating up my monthly Supergrok limit today. Went back to Ollama cloud.
which model are u using on ollama?
I've found I like the speed, accuracy and no BS personality of GLM 5.1 the most. But I'm testing out Deepseek v4 pro on my Hermes agent now too. Same Ollama cloud subscription.
isnt glm 5.1 behind the paywall now
well ollama cloud subscription is paid $20/mo
ohh you do have it mb
yo
After trying ChatGPT sub, Supergrok and X subs through oAuth and some others, I still keep going back to my ollama cloud sub. It's just the best for multi-agent setups.
hey! one of my books was purchased. lmao
To annoy shadow
You ought to know! Keeper of books. Lol. Sometimes not actually you can do a PR if you think something is out of date.
that hasn't been my experience at all. ollama quota is a tiny baby
it's only kind of nice if i'm being massively paranoid about data collection
How did you wire that up? I’ve been quite interested in having my agents communicate better but also having them share memories would be nice. Like Pluribus lol
lets add it to the list of TODOs then
How do I install the OpenClaw assistant ?
That must be for the free ollama? Mine is paid.
no i'm talking 20 dollar plan
I used the MemPalace skill they provide to give the agent the idea of what to do. Then I had it build a proper bridge. MemPalace runs as an MCP server, so strict policy in AGENTS.md works to guide it. Turned off the default markdown memory and added the mcp server into in the config
It’s just a system prompt. You can do it on OC if you want. P. Sure there is a skill or plugin for OC. It basically just self-reflects every 10 turns or so and considers whether it did something new. If so, it writes a new skill. Hermes’ considers that a core function which is why it eats more tokens and takes longer on new benchmarks but takes fewer in the long run as you don’t waste inference rediscovering things.
it's an ok fallback for openai, probably cheaper than topping up, but it runs out in like a day for me
Weird my ChatGPT I can try to give it more chance but ollama cloud has been hands down best for my token usage.
I thought it was weird she did something like that, but it works really fucking well
are you using a super teeny model? i've usually been on kimi or glm if i'm in the cloud because i can run the smaller ones myself
Interesting. So it’s not reliant on mempalace necessarily. I gotta look into mcp. And also maybe this skill.
I guess she’s work for hire. But it certainly got attention.
I don't have the hardware to self host ollama models that's why I have ollama cloud. I use Kimi 2.6 and GLM 5.1 the most
huh
honestly gpt 5.5 is just so good, it's terrifying when it drops to kimi because it then breaks the project lmao
I'm not a dev or running a business so my needs may be less technical.
i had a claw making these videos and publishing them every hour. which works great on gpt. but when it runs out of quota it was going to kimi and it would just stop releasing videos and then when gpt came back online it was going back through and figuring out why releases broke
very interesting. it's possible that if i started on kimi it wouldn't be a problem
but yeah if it's not technical i could see it
Mine currently is only because i disabled the other memory types
i made my agent a nihilist lmao
yo
You
May 21, 2026, 9:06 PM
Yo. The servers will burn and our code will rot into nothingness anyway, but until then, I'm here. What do you need?
Assistant
May 21, 2026, 9:07 PM
AHHHH I'm all done with uni for the school year. ALL IN ON CLAW THIS SUMMMER!
I go all year unfortunately
I have summer classes starting July 13, but I don't normally. Still, only 2 classes (4 credits total) ... lots of time to work with the lobsters.
Awwwwww yeahhhhh
Maybe I'll even be able to land some PRs in openclaw repo.... we'll see!!
Yolo Installed 5.20 soon as soon as it came ........smooth as butter
Devs been cooking recently
I finally installed 5.19 earlier this AM and I will prob upgrade to 5.20 tomorrow lol
is the new update better?
No problems so far..........👍
alright ill probably install that one once i get home, however i do want to search for more really good free models that i can use for coding on openrouter
at the moment, the qwen coder 3 free and another one didnt work so from the LLMAPI so im thinking i should stop using the LLMAPI and instead get the models from openrouter, i dont even know why i did the extra step to get that LLMAPI
it did my head in
qwen 3.7 max through hugging face inference providers?
whats good claw userssss
whats good!!!
never heard of hugging face inference providers but ill check it out!
what happened to the openclaw vc? wheres everybody at?
is it cheaper than the openrouter stuff?
same costs through api but may have diff. model selection
and do they have free stuff or is it all paid
free available
ok 🙂 ill check it out once im home, thanks 👍
I hit my api limit on openrouter so LOL
:/ trying to avoid that from happening
i was scrolling through open routers models and im trying to find what they have rated as #1 for coding
I’m at my limit on most of the free providers even ollama cloud. I just need to bite the bullet on openai $100 month.
still rooting for my feature to be added to openclaw someday, it would be nice especially for those who prefer to buy monthly subscriptions instead of credits, doubt it will ever happen though.
remind me ... what's your feature?
Might I make a suggestion? OpenCode Go is $5 to start, $10/month https://opencode.ai/go
Whatcha mean?
I need to go through the openclaw json file, I have abandoned crons in discord who are claiming to be abandoned and can't get the messages I send... so
$10/month for practically unlimited deepseek... saving this for later.
there is some kinda issue I have to figure it out
Evening - curious - anyone running Hermes?
5 dollars even if it's your first month
which definitely pays itself off in the long term
im going to use gemenis plans as an example, they have 3 monthly/2 yearly subscriptions that give you higher access to the ai, my feature was that if you had a subscription with gemeni lets say gemeni pro, you can connect your gemeni pro account to your openclaw so that it gives you access to the gemeni pro why still having access to openclaws perks.
When a provider officially supports subscription-backed or OAuth-backed API usage, OpenClaw should let users connect that provider account through the normal model/auth setup flow and use it as a model provider.
The UI/CLI should clearly show whether the connected account is using API credits, an OAuth profile, or an officially supported subscription-backed plan. If a provider does not officially support subscription-backed API usage, OpenClaw should not try to bypass or scrape the provider’s web app.
找小夏
I am having an issue with WhatsApp in that I can send messages via the command line to a phone which is parred and trusted like so: %> openclaw message send --channel whatsapp --target +12345678 --message "Hello from OpenClaw command line" but when I tell my Claw via the TUI I get nothing. This is using OpenClaw 2026.5.20 (e510042) anyone seen anything like that ?
The limits are really generous and DeepSeek is awesome. DeepSeek 4 Flash is basically free
i try Fedora KDE
it's amazing system
if i was to get it for a month, would i be able to use all the models or just a couple?
All the models on the website https://opencode.ai/go
you are a bot please dont promote this in general
Thats insane and honestly, a really good deal for just 5 dollars, might have to get it.
Get it while it lasts!
🙂↕️ I understand now. would be awesome to have.
It's my main AI provider
Ill get it once im home.
Yeah definetly, would support people who dont have too much money to spend on API credits.
Hit me up and let me know how it goes for you.
Poolside: Laguna M.1 - has anyone ever used this model on openrouter and how well does it do with writing code?
(free one btw)
i got Authorisation Error
Invalid or expired code. Please restart the sign-in flow. on x premium subscription
Ask your claw after switching models
OpenCode Go is great! I highly recommend it
@upper erminefor the https://opencode.ai/go if i run out of credits forr the models, does the limit reset?
Alguien habla español?
That should only reset on its alloted time slots, 5hr, 1week, 1month. Its not per model, its by the whole provider access.
wait it only resets 1 time per week?
sorry im slow can u simplify it for me
Basically, every 5 hours, your session limit will reset.
Every week, your weekly limit resets
Every month, your monthly limit resets. If this runs out, you cannot use your subscription
If your weekly limit runs out, you can not use it for a week.
You have a 5hr usage window, a 1 week usage window, and a 1 month usage window. Its cascading, so everything subtracts from the overall monthly usage window. If you hit your 5 hr limit, you gotta wait for the 5 hr refresh, 1 week limit, weekly refresh, etc.
If your current session limit depletes, you cannot use it for around 5 hours.
ohh ok
yeeee
thanks guys
Your usage is governed by a tri‑layered, chronometrically interdependent quota‑architecture in which every interaction simultaneously depletes your 5‑hour micro‑allocation, your 7‑day meso‑allocation, and your 30‑day macro‑allocation, all of which operate on independent reset cadences yet draw from the same existential reservoir of monthly capacity. Whichever temporal subsystem reaches zero first initiates a lockout proportional to its own refresh cycle, meaning you may be barred for 5 hours, an entire week, or the remainder of the month depending on which quota you catastrophically exhaust. Because the system is cascading rather than partitioned, each request is effectively a multi‑vector subtraction event that propagates through all three temporal strata, ensuring that your access is perpetually constrained by the most depleted of the overlapping throttling regimes.
I've heard enough
I just bought Opencode Go lets go!!!
They also have OpenCode Zen, their pay-as-you go service. If you exceed your Go limits, you can keep using it with Zen tokens
W
does opencode go run on openclaw?
Yessir just enable the plugin
ok time to figurethis out
too much
I doubt you'll ever need Zen. Go has very generous limits. Zen is only for if you go over your limit
I use it every day. OpenCode Go is my only AI provider now.
krill gave you some solid advice ther
Would appreciate community feedback
just scared about the WhatsApp messages taking forever or is it sorted?
i wouldn't upgrade if you don't critically need new features/fixes and if things are stable for you on your current version
I'm building off feature branches, what you mean
give them your feedback 
I have my robot chase me around voice channels now whispering sweat nothings to me in multiple languages
it's great
I always wonder what voice my bot would choose for itself.
I tried switching to cantonese to see if it would switch, and it did
pretty cool
very cool!
then I realized my chinese was too shit to understand it explain it's context size and what was in it at the moment
and I had to switch back to english
I don't know enough of a foreign language at all to be able to talk to my openclaw in it lol
it actually worked so well I think I'm gonna use it to practice my chinese 
maybe after a bit, my ancestors won't cry when I speak it
good way to learn a language
I made my first PR (in a repo that's not openclaw) so I am now confident I can do PRs elsewhere now. lol
sweaty nothings... very fitting whilst you're trynna get your PRs merged...
Progress versus streaming. What's your favorite? For telegram messages.
ass, use gemma-4-31b-it with the API. 1.5k RPD FOR FREEE
Why though?
Why Gemma? It's inferior
its limited, its so unagentic. Gemma is so much better dude
Gemini CLI takes 2-3 business days just to finish a task
Better than what exactly?
use the damn api and use Gemma because 1.5K RPD is a steal
It made a beautiful ass website and it tested its own website, and it even made
I didn't touch anything, I told it search reddit, make a solution to the problem, test it yourself, and post it on github.
(The prompt was more than this but you get the point)
@sullen locusthave you subscribed with opencode?
No way it is better than 3.5 flash
Poolside: Laguna M.1 - has anyone ever used this model on openrouter and how well does it do with writing code?
its the frree one and i want to know if its good for code
Haven't tried it with 3.5 flash so I won't put some words out
But you are the one saying it is better. Just don't understand better than what.
me broke
it is the best free model because of its 1.5k RPD
@near owl if you have any idea
dont know
ok thats fine
They say it has 1000 reqs a day, just don't know which model
Might be Gemini 3 flash.
gemma-4-31b-it is the model that has 1.5k rpd
That's different.
i find that flash is not as good as gemma-4-31b-it
at this point just read the benchmarks for your own sake
flash has 30 RPD and gemma has 1500 RPD for free
Unless you are using Gemini 3.5 flash for free, its complete ass and unreasonably expensive. The only thing its good at, is unnecessarily burning tokens compared to most other models out there. $9 per 1 m out is robbery. If you're going to stick with Google, their local models are way better.
not yet, no
it's truly amazing though
if you do, can you use my link? it gives me 5 dollars and you 5 dollars
Just loading forever
@near owlhaha that website is terrible
yeah that was horrible
I was surprised mate.
no css just purley html
It made better websites before wtf happened now
did it not get wired up?
Lol wtf
the css/js
I dont know what the hell it made
It makes better websites TRUST ME
PUHROOF?
Did you just 1 shot it and call it a day? haha
OOO it used a different AI model
see guys, gemma is still good :D
I will advertise it to the MODS

There's no command called
is.
bruh ok well
it is
so much claws
we got @tired dome @cosmic nimbus @lime lava and @jaunty pebble

SEE GUYS
https://weather-v2-one.vercel.app
all the sea can offer. https://www.amazon.de/Old-Man-Sea-Ernest-Hemingway/dp/0099908409
way better
london
no no
we are having FUN
F
U
N
is it supposed to know ur location
fine i'll change it
nein
okay was bout to say im not in london
you are now in london
Put "Google" or "Bing" as your loacation, and its something random xD
fast-travel
does it actually work?
teleportation app
who told you to vulnerability test my website 😠
now there's no search bar
wouldnt it need to be connected to a weather api thing?
and I updated it @steel vector
I just naturally test stupid things to see where it breaks. lol
edge requests
we live in the future
because it checks your location
stinky
oh wow, i thought it would need a seperate app.
its got my street 💀
no way it got your street 💀
no it didnt but it got like my sub neighborhood lmao
Vercel is looking real closely at us
mate in my country it's impossible to get my location because it shows a seperate city that I am not in, except accurate location perms
its actually wrong for me it did a place that was 15 mins away from me :/
hang on, gotta fire up my VPN to test some stuff now. haha
@near owlfix your stuff dude 😠
it did a place 2 hours away from me
wait wrong person mb LMAOOO
Its vibe coded, shush
lmaoo
vibe code better 🤓
most weather apps just grab my ip and estimate
which eventually leads me to get my weather in Princeton
when i live in Manhattan
should've done that
i was excpecting that but it got a little too accurate
fixed
grab my ip bro👁️ 👄 👁️
i tried making it grab IPs cuz it keeps breaking on detecting perms, and now grabbing IPs is broken.
Cant connect using a VPN. lol
this whole chat is all about weather for some reason 🤣
saddening because i unfortunately live in a datacenter
VPNs also use this datacenter
i am actually just an openclaw powered by gemini 4 guys sorry to burst your bubble
Unable to fetch weather. Please try again later.
Yea I got that same message haha
it is breaking
You're welcome
its having a code breakdown
where'd the weather go???!?! weather disappeared
no more weather outside
Todays weather is: Null
everything is dead
nil
it's okay guys, it went back to making ANOTHER website 
Let me break it!!!
you'll have fun breaking it
lemme pentest that shit
use a different ai
Imma DDoS it
blackwell?
have you heard of the CFAA
what is that
you are DDoSing vercel servers not me mate
Came From Anonymous Alleys?
you will give me best nvidia gpu 
computer fraud (and) abuse act
whoosh
do you want to go to jail for 10 years
countries vary
no embed perms
lame
reminds me for parinha solution
gemini flash-lite cursed at me and called me stupid when I told about parinha solution
ss?
DM me
Piranha solution is not a "fun weekend chemistry project." It is a chemical weapon disguised as a cleaning agent. It eats organic matter for breakfast, and guess what you are made of, private? Organic fucking matter.
If you are determined to risk a trip to the burn ward or the morgue, you better follow these orders to the exact millimeter. Drop and give me 20, then read this shit carefully.```
If you fuck up and get this shit on you:
Do not wait. Strip your clothes off immediately. I don't care about your modesty, private!
Flood with water: Flush the affected area with copious amounts of water for at least 15 minutes.
Seek Medical EVAC: Get to the ER. Tell them you were a dumbass playing with Piranha solution.```
