You might want to read the terms of service next time before you utilize a service that you are unfamiliar with! Google refuses to let you use your oath with openclaw because they refuse to subsidize the slop that it is processing. It is not worth giving free frontier models from Google Gemini Pro for the basic tasks you are looking to complete with open claw I would know because I've been developing with AI for over a decade.
Openclaw has a lot of potential. It's a pretty neat tool. I have to admit. My favorite part is that you can point the URL for the large language model to anywhere you want. Since most of you guys only have a rudimentary understanding of the difference between AI systems and know that there are options for local and cloud models, we are going to expand your knowledge base today. Cloud models are great. They are able to set their own terms. Google gives free compute with oath connections. They have stated very clearly that you must be a paid API user and that you are not able to use oath and using oath with openclaw will result in termination of your account immediately because the tasks you are asking Gemini to do are for a preschooler..
We have known this for at least a week at the Atlanta technology development Center and we were able to get an account restored by apologizing. However, we knew in advance that it was going to get nuked. This is actually an open claw issue not a Google issue because openclaw is the one that's still allows for 05 connections and is not made the change to single-handedly only permit API connections. If they only permitted API connections through Google Gemini they would stop any of their users from being banned but they haven't made that change yet. So if you are mad at someone be mad at Peter steinberger for not shipping faster or paying attention to terms of service or the recent slew of complaints from users. This isn't my department. I don't want to figure it out. You can handle it.
You really think Google wants to subsidize compute so that you can organize your freaking calendar you morons? Sounds like a great use of a data center where you could just click accept invite.
Oh, you can't reply to text messages, so you trust AI to do it now. Wow! You must be the laziest piece of s** on the planet.
Oh email's too much for you? Let me tell you what inbox zero as a mandatory process looks like when you wake up with 150 emails on average every single day over a a 7-year period and you handle every single one every single day before you leave the office with zero unread and zero pending reply. You get really good at making sure every single customer is handled. When customer service is not optional and your service level agreement is that you answer customers 24/7 365 you do it immediately without question and everything is solved before the lights are turned off and if something goes wrong at any time of the day that phone's getting picked up ring one and the problem is being solved before the line is hung up or there's an escalation to management and it is handled on the spot.
Some of you are a little soft and don't actually believe in putting in the work or putting yourself in positions that will leave you slightly inconvenienced because it would be way too much to ask for you to stay 3 hours after work if that's what it takes to build a billion dollar company. I really hope you enjoy dinner with your family every single night at 4:59 p.m. Sharp while you get smoked by people that are actually staying at the office.
So now that the only two options you think you have are paying someone for their large language model because you are poorly researched and don't know how to build your own system taking thousands of dollars and buying something with over a terabyte of RAM that spits out a hundred tokens a second. Maybe you can do that 10 times to get a thousand tokens a second if you really want to drop 100 Grand on new macbooks. That's about a quarter of the speed of meta which offers infinite free compute if you cleared their wait list as a developer(6+ month waitlist for consideration).If you don't have free compute from meta, I feel bad for you because they were giving 3,000 requests per minute until openclaw came out and that's only 10 requests per a minute because they got real tired real quick of these 24/7 agents taken up all the compute to do. Jack diddly s**
I will tell you that. Claude grok meta Google and almost any frontier model you haven't heard of will give you infinite free compute if you are building products that have never been seen before. I'm not sure what Peter was doing spending 20,000 a month on compute. Does anyone have a figure on how many credits he burned? Cuz I'm in the tens of billions and my compute spend is $0 over 18 months. It's really interesting what happens to your account when you feed it, new data, new solutions and try to build theoretical software that is not currently possible with what is available today and then you come back when the new model comes out. They welcome you with open arms and they don't charge you to use it and they do a lot of heavy lifting without much of a cap.
The point of this is to you mention that you do not need millions, probably even hundreds of thousands of parameters for your basic tasks that you are too lazy to handle yourself. I honestly am not so sure what any of you are doing with open claw when cursor builds much better software.
In my professional opinion, the most effective way to utilize open claw is by building your own decision tree or something similar that has a pretty rigid process and follows exactly what it should do without thinking. You guys would be amazed to know that you can build a fully functional model that does everything you need and completes every task a customer will ask for with a 12 KB file it runs with less than 15% of your total compute on four cores and takes less than 2 GB of RAM on a web server. It provides instant response with no processing delay. It knows exactly what the customer is asking and matches it up with the correct response. It is able to process images using data sets which are also very lightweight because they are specific to your business.
So next time you try to set up openclaw, ask whatever AI you're using just to rip out the apis that comes with and build your own model which you will host on your local system server. You might need to ask it to refactor for firebase or whoever else you're going to be hosting with. Go ahead and keep paying digitalocean $6 a month for the worst VM I've ever seen. That's a great investment and their current best practices go against the current best practices that open claw recommends, so I would not recommend any third party providers, even if they are worth $5 billion because they do not maintain as fast as openclaw ships.
Any questions I'm available but I probably don't want to help you unless you're paying me a lot of money because I and I'm kind of a big deal in case you can't tell.
You really think Google wants to subsidize compute so that you can organize your freaking calendar you morons? Sounds like a great use of a data center where you could just click accept invite. Oh, you can't reply to text messages, so you trust AI to do it now. Wow! You must be the laziest piece of s** on the planet. Oh email's too much for you? Let me tell you what inbox zero as a mandatory process looks like when you wake up with 150 emails on average every single day over a a 7-year period and you handle every single one every single day before you leave the office with zero unread and zero pending reply. You get really good at making sure every single customer is handled. When customer service is not optional and your service level agreement is that you answer customers 24/7 365 you do it immediately without question and everything is solved before the lights are turned off and if something goes wrong at any time of the day that phone's getting picked up ring one and the problem is being solved before the line is hung up or there's an escalation to management and it is handled on the spot. Some of you are a little soft and don't actually believe in putting in the work or putting yourself in positions that will leave you slightly inconvenienced because it would be way too much to ask for you to stay 3 hours after work if that's what it takes to build a billion dollar company. I really hope you enjoy dinner with your family every single night at 4:59 p.m. Sharp while you get smoked by people that are actually staying at the office. So now that the only two options you think you have are paying someone for their large language model because you are poorly researched and don't know how to build your own system taking thousands of dollars and buying something with over a terabyte of RAM that spits out a hundred tokens a second. Maybe you can do that 10 times to get a thousand tokens a second if you really want to drop 100 Grand on new macbooks. That's about a quarter of the speed of meta which offers infinite free compute if you cleared their wait list as a developer(6+ month waitlist for consideration).If you don't have free compute from meta, I feel bad for you because they were giving 3,000 requests per minute until openclaw came out and that's only 10 requests per a minute because they got real tired real quick of these 24/7 agents taken up all the compute to do. Jack diddly s** I will tell you that. Claude grok meta Google and almost any frontier model you haven't heard of will give you infinite free compute if you are building products that have never been seen before. I'm not sure what Peter was doing spending 20,000 a month on compute. Does anyone have a figure on how many credits he burned? Cuz I'm in the tens of billions and my compute spend is $0 over 18 months. It's really interesting what happens to your account when you feed it, new data, new solutions and try to build theoretical software that is not currently possible with what is available today and then you come back when the new model comes out. They welcome you with open arms and they don't charge you to use it and they do a lot of heavy lifting without much of a cap. The point of this is to you mention that you do not need millions, probably even hundreds of thousands of parameters for your basic tasks that you are too lazy to handle yourself. I honestly am not so sure what any of you are doing with open claw when cursor builds much better software. In my professional opinion, the most effective way to utilize open claw is by building your own decision tree or something similar that has a pretty rigid process and follows exactly what it should do without thinking. You guys would be amazed to know that you can build a fully functional model that does everything you need and completes every task a customer will ask for with a 12 KB file it runs with less than 15% of your total compute on four cores and takes less than 2 GB of RAM on a web server. It provides instant response with no processing delay. It knows exactly what the customer is asking and matches it up with the correct response. It is able to process images using data sets which are also very lightweight because they are specific to your business. So next time you try to set up openclaw, ask whatever AI you're using just to rip out the apis that comes with and build your own model which you will host on your local system server. You might need to ask it to refactor for firebase or whoever else you're going to be hosting with. Go ahead and keep paying digitalocean $6 a month for the worst VM I've ever seen. That's a great investment and their current best practices go against the current best practices that open claw recommends, so I would not recommend any third party providers, even if they are worth $5 billion because they do not maintain as fast as openclaw ships. Any questions I'm available but I probably don't want to help you unless you're paying me a lot of money because I and I'm kind of a big deal in case you can't tell.