Hermes Desktop vs Claude Code vs Codex vs MiniMax Code vs OpenClaw vs Antigravity — Live Showdown

Okay. Yeah, I am live. So, hello and welcome to Toby on Fitness Tech. Uh so that I can see what I'm doing. Yep. So, hello and welcome to Toby on Fitness Tech. I wanted to do a little bit of a different uh video today than I normally do, and I wanted to discuss Let me minimize this for a second. So, I wanted to actually discuss a bunch of different harnesses. And I'll tell you why, because I think in the agentic world, not only does model matter, but but harnesses are going to matter also. So, what I wanted to do is go over the ones that I've been testing recently, and then uh go over where I think each of them are in terms of capabilities and capacity, and um my personal reviews of each of them. So, let's get started on why I'm actually looking at harnesses, right? So, uh a couple months ago the I was using OpenClaw for pretty much everything, and I was using uh Anthropic's models inside of there. While I was in Puerto Rico, they actually banned us, and everyone got kicked off of uh OpenClaw with Anthropic using their subscriptions. So, I'm a big fan of using subscriptions, and that's the other thing I wanted to show you guys today. I'm going to show you each of the different subscriptions that I use, their pricing, and also what you get for that price. Because it's wildly confusing out there. People didn't even realize that uh OpenAI's pro plans, their their pro max plans, their biggest plans, they do have subscription limits, and some people didn't even realize that. That they thought it was just completely open once you get up there. No, no, no, no. Every one of these has uh an amount of tokens essentially that you get for the cost of what you're doing. And and so we have the harnesses, we have the subscriptions, and the reason why I'm testing all of these harnesses is because I was just going to use OpenClaw for everything. But when Anthropic banned it, uh I had already before that been using Codex in particular. I had a small Codex plan, and so I was using Anthropic inside of OpenClaw, and then I was using the free Codex plan, and I bumped that up to the $20 a month plan because at one point in time the free actually got to be very, very nerfed in terms of what they were doing with it. So I started with free Codex and the $200 plan uh from Anthropic inside in the the Claude plan inside of OpenClaw. We got kicked off of that, and so now I have a completely different tier structure, but I'm paying about the same amount of money for all of these different subscriptions as what I was the $200 subscription from Anthropic. And I'll explain all of that in detail, but in in the in the image and and in the actual preview for this, we'll be talking about Hermes Desktop, Claude Code, Codex, MiniMax Code, um OpenClaw as a harness, and Anti-Gravity. So Anti-Gravity is the most probably confusing one in terms of not in terms of usage. Well, yeah, in terms of what usage you get for the money that you're paying because it uh it's surprising the different tiers. And luckily I've used three different tiers of Google's plans because uh theirs was the most complicated. Uh everything from their their most expensive plan to their cheapest plan that includes AI, and I can discuss all three of those plans and one is actually the sweet spot because there is an answer for at least Google if you're going to use anti-gravity of which one you want to pay for. Uh and it's clear and it's like a a no-brainer choice. And then I wanted to talk about the state of OpenClaw uh because it became so frustrating for me that I actually moved over to Hermes uh for most of my work. And I still have OpenClaw around and I found sometimes the best thing to do is to point Hermes at OpenClaw and say, "Hey, do this the way OpenClaw did it." Now, you notice here I created a thumbnail. And this was using this this harness right here is actually Codex. So, uh this is the Codex harness and in my opinion well, and it's not just my opinion. You can look at the other harnesses. Uh they've all been modeled after what was done in Codex. This is the right way to build a harness in terms of uh look and right? It it is I mean, the fact that uh Claw the Claude Code desktop app now looks like this. Um the MiniMax code app looks like this. Anti-gravity now looks like this. Like they all look like it now. And they did not before. Uh the in particular, Claude Code desktop was particularly terrible looking in my opinion up until recently. Uh this one has massive massive changes. So, let's take a look at each of these desktop apps. The only one that I don't have installed I use the Claude Code command line. So, you'll see the Claude Code command line back here and I just run, you know, dangerously skip permissions to be honest with you most of the time uh because the work that I'm doing is similar to the way OpenClaw works where I give it permissions to this particular machine and it goes out there and does the work. So, that's the way I run this typically. and why am I not using their desktop app even though they made it uh look far, far better than it used to? Well, the reason why I'm not using their desktop app is because it started hallucinating wildly on me. So, the Claude code desktop app seemed to really still have problems, and I'm not willing to investigate it because I tried the Claude code desktop app once, and it was horrible and unusable in my opinion. And then I tried it again, and I'm like, "Ah, they finally fixed it. It looks so much better. It looks like Kodaks." Um they finally fixed the way the way it looks and the way the user interface works, and then after 2 days it just started hallucinating like crazy, something that I've never seen from the command line. So, I do just stick to the command line for my Claude usage. Now, where am I at as far as Claude usage in terms of how much I'm paying? Well, like I said, I used to have the $200 a month plan. So, now I am down to the $20 a month plan billed uh annually. So, that's what I decided to pay for, and that brings it down in cost quite a bit. So, what I found with switching over to annual, uh I do hit the limits, by the way. This is the only one of these plans where I consistently hit the limits. So, I consistently hit the the 5-hour limit, and I bring it pretty close to the weekly limits usually. You can see it just reset yesterday, and I haven't touched it very much because I've been playing around with Mini Max, and I'll talk about that next. So, in terms of my ranking of Claude code and my ranking of should you have this, uh unfortunately, Opus 4.8, in my opinion, is still one of the best models, if not the best model, and not using it, and not having this $20 plan at least uh billed annually, I think is a mistake as a as a uh you know, if you're doing anything inside of AI. So, I do think if if if you're paying for just one plan, this is actually the plan even though I'm upset with Anthropic with what they did with Openclaw. I'm upset that it doesn't work inside that that harness anymore. I think that it's very disappointing that they've done that and you have to go into the the extra usage. You can see that uh I have the extra usage stuff turned off right now. You know what I mean? And I I have a $35 credit still. But, I'm leaving this turned off. But, uh everything if you try to use Openclaw, it hits this extra usage instead of hitting your Claude plan. So, this is where it'll actually hit. Let me go over and move this a little bit so that you can see the rest of it. But, uh it will hit this portion instead of hitting your your plan, which is very important to realize. There is hacks around this and I may or may not have used some of those hacks at one point in time. But, you just get such a low amount of usage with this $20 plan that it's not worth it uh to to do that. It's better to use an explicit set of things that you want Claude code to do. And I still find that it is very good at doing tasks. Uh is it the fastest in the world? No. We'll talk about which one I think is the fastest. And and in fact, it was pulled up first. So, spoiler alert, the thing that I think is the fastest is Codex with GPT-55 and I have it set to extra high right now. Um but, I I actually should dial it down to high. This is where I would prefer to keep it, but for I just had it up at extra high whenever I was building this image. This is for sure the fastest, most accurate, in my opinion, of these harnesses, and this is the one where I'm paying $100 a month. Now, you notice I have So, I have the Claude stuff in my browser. Right, so this is pulled up This is Claude AI new settings usage, right? That's where this this lives. Okay. That's where the usage meters live for Claude. So, however, for Codex, I could not find a web interface to find the usage meters. So, for the usage meters for Codex, it's settings, usage remaining, and down here it will show you and this is at um This is the $100 plan, and I will tell you what So, up until a couple up until a week ago, I could never hit their 100% usage even with trying because they were doing that double promo. And then I was really nervous cuz I had it down around 30% usage, and I still had uh days and days left on my on my thing, and it just reset. So, it just magically reset out of nowhere and gave me back 100%, and you'll find that and I don't look into it very deeply, but you'll find with Open AI in particular, it doesn't reset like you never get extra usage out of Anthropic. Like it just doesn't happen. In fact, they've only ever lowered the usage. Um they did bump it up a little bit all of their tiers whenever they made that deal with uh with uh Elon and to use his data centers, so those got bumped up to where it's actually usable again, but you don't get resets. You know what I mean? You don't get random resets of hey, you all of a sudden have 100% usage again, but you will see that with with this in particular with Codex. And you can't rely on it, obviously, but it it just happened to me where I got bumped back to 100% out of nowhere and I had days left where it was not going to bump back. So, that's why you see I'm almost at you know, I'm almost at 100% available. That was down at 40 30% whenever I woke up this morning and I was not expecting to see the reset today. So, that's pretty awesome from from Open AI. I found they've been the most generous with their plans in terms of that. Uh but the other one that I really wanted to talk about is MiniMax because this has been talked about a little bit online that they created an M3 model. They have a new multimodal model uh similar to what what similar to what we're we're seeing out of Google and some of the other, you know, players. Uh and and I believe ChatGPT 5.5 is a multimodal model. Uh I should probably fact-check that real quick. Let me pull that up. So, GPT 5.5 multi model. Model. Yep. Okay. It's designed natively for Omni Omni model intelligence. So, yes. Yes, sorry. I wanted to fact-check myself just to make sure I was was right on that, but uh that means that ChatGPT 5.5 MiniMax M3 and also Google's uh Google has a new Gemini Omni model that they they've been building, too. So, these are all these Omni models that can do everything, and that's why it didn't have to I wasn't sure because Codex will, behind the scenes, call different models and tools and everything else to actually build me the thumbnail that I wanted. But, yes, it is capable of that. And it's pretty incredible, to be honest with you, that these models have come so far, so quickly. So, talking about MiniMax in particular, they just launched M3, and they changed This page has looked massively different uh the last 3 days. So, this looked totally different before the launch of M3, and it showed you breakdowns of all of the different things that they had within MiniMax, and which ones you had access to, and how much usage you had. Now, it's just here's your token quota, and here's how much you've used, just like the other models. You know, just like what you would have with Codex. And I was very nervous that the usage that they were going to give you would be so neutered that it's unusable, but they also gave whenever they made this change, they give you 20,000 tokens, and I've been using it the most aggressively I've ever used MiniMax, and I am still at 6,000 tokens, and you can see right now I have a ton of my quota for the 5-hour limit, but realize I was I was taxing this thing and hitting it as hard as I could. The weekly limit is set to unlimited right now, which is just awesome. Uh and in terms of being impressed with a model and a harness, I went into MiniMax code thinking it would be absolutely terrible, and it is actually pretty good. You know, it is As far as a harness goes, it seems to be as capable as any other, if not more capable. So, to be honest with you, I was shocked at MiniMax Code as a harness. So, I've only been playing with this for like 3-4 days now. You can see I only have like three task histories, whereas if you go into Codex, I mean, this is, you know, all kinds of projects from all different periods of time, right? So, Codex is the thing, and I have a $100 Codex subscription. Right. Here with MiniMax, the subscription that I have on that is this this token plan is if I go to token plan, I have the $20 a month plan, uh but billed annually. So, realize with Claude, sorry, with this one, with Claude, so with Claude, I have the uh $20 plan billed annually, basically, and with MiniMax, I have their $20 plan billed annually, which brings down the cost to under, you know, way under, you know, way under $20 a month on both of them. And I would say if you're paying for plans, I have been so impressed with what MiniMax is capable late lately that it's blowing my mind. And you can see, what's funny is even more crazy, right, the um you know, it says full access to M3, M2.7, image, speech, and music. It's just uh it's just incredible, and everything shares the same quota, so they finally fixed this. Um Oh, sorry. Actually, I'm down here. My bad. I'm not even to there. I am actually down at the $100 plan, not even the $200 plus plan. Sorry, I'm clear down here. Uh and so I'm paying even even less than that right now. So, I'm at $100. I would say I'm at $100 yearly for this and in terms of pure usage, you get at $100 yearly, you get 100 times the usage you're going to get out of out of Anthropic. Now, in terms of model quality and re-prompting, I I found that with the reason why if you look at this uh if you look at my usage numbers, right? And you go over 30 days. This This is This graph was all new today, by the way. If you look at it, it's like uh almost, you know, almost no usage and then it ramps up and I I believe like this is when I first got it and then it ramped way back down uh because because I didn't find M2.7 to be an improvement over M2.5, to be honest with you. I found that I still had to re-prompt three times and it was just pretty disappointing until I put M2.7 inside of the Hermes uh the uh Hermes Hermes container and once M2.7 was inside, you can see this is this ramp up here and it barely touched it in terms of uh token usage, right? And then you can see this is now me pointing This is me pointing OpenClaw, both Hermes agents that I have at and and using uh Hermes Hermes desktop. So, let's see what it actually built at the end. Uh pa pa pa. So, here's what Look at the difference though. Like this it it did moderately okay. But, you can see the difference between what gets generated by M, you know, Minimax M3 versus what was generated with Codex. So, I one-shotted it with Codex. This is just Uh by the way, to be fair, it's one-shotted with Codex, but this is giving it this image uh the source image for this and this image that's behind it that actually was generated by the Hermes agent. And then uh the description that was already generated for this uh from what I'd voice to text. So, this is the and then said, "Hey, build me something." Right, so this is it one-shotting it though basically. And it was able to do it uh incredibly quickly like in under 5 minutes and and build something that I really like. Whereas Minimax, you know, inside of Hermes, right, took Oh, boy, 15 minutes to get here. I would want this image to be down. I would want this to have the proper spacing, right? I I would also want it to not collide with this as much as it does. So, I want the head to head to be down here and I would want the five AI agents to be up here, right? So, I would have to re-prompt it with all of the description of what I wanted moved around and where I wanted it moved it moved to. But, otherwise, it is not terribly bad for uh for something that is $100 a year versus, you know, something that I'm paying $100 a month. Realize the difference in price between these two is this MiniMax M3 subscription, which can power uh two Hermes agents at the same time, plus powering the uh MiniMax code, right? It can it it can power all of that at the same time for $100 a year, versus this Codex subscription is $100 a month is what I'm paying. Now, do I think uh and by the way, up until today, I had Hermes on the DGX Spark, uh one of my machines. That was pointed at uh ChatGPT 5.5, right? And then uh I had Hermes on this machine pointed at MiniMax 2.7, and I just upgraded that to uh you know, MiniMax M3 uh whenever I put whenever I installed the Hermes agent desktop. And so, what we're looking at here is Codex being, you know, massively massively more expensive. But, it is especially with ChatGPT 5.5, it is incredible what you can do with these with these tools. So, my personal opinion is uh MiniMax is the biggest bang for your buck, but realize that I used to describe MiniMax 2.5 and MiniMax 2.7 as the hold my beer model, which uh meant that it would it would you know, that's a saying that they say in the country, and what what's going to happen is it something is going to happen after that, but it's probably not going to be what you wanted. Right, so that's the description of it for me. Uh that's one of those things where I came from the country, and that would be a thing, you know, that you know, you'd hear hold my beer, and then uh Ah, going to get into a fight, or someone's going to try to pull something out of the ground that they shouldn't. Like all kinds of chaos would ensue, but something would happen, right? MiniMax was uh MiniMax 2.5 and 2.7 really had a capability to do things, just not correctly. And I would say with M3, they're getting much, much closer to being able to do agentic tasks properly. It's just it still requires multiple prompts, whereas with OpenAI, usually it's a single prompt quality results. Very rarely do I have to prompt it again. And that's specific to ChatGPT 5.5, by the way. Uh and with earlier models, it really did work well within the Codex harness, but I found within uh Open Claw specifically, it worked terribly up until recently. So, just realize that Open Claw for sure worked better with Anthropic's models, and now it's very much leaning I mean, they have the founder now, you know, from from Open Claw at uh OpenAI, so it's not shocking that this has come a long way. And the the reason why I moved most of my tasks over to Hermes Agent is because Open Claw updates are just awful. Today's update was the first update with Open Claw where I did not have to actually run and even then I still ran I still really I really expected it to be broken after the update, and for the first time in a month and a half, the update didn't randomly break something. And every other update has randomly broke something. So, pretty wild to think that we've got, you know, that kind of >> [snorts] >> bugginess, you know, from Open Claw, and it's just consistent, whereas Hermes Hermes Agent though, Uh everyone's saying that this thing's great and like there's no problems with So I've never I I had one instance where OpenAI stopped working within it and it turns out it was a bug on their end. They opened a ticket with them and they fixed it. I I expected it to be a bug on Hermes' side but no, it was actually a bug on OpenAI's side which kudos to them by the way for fixing it so quickly um and allowing, you know, and allowing the patch to go through whatever needed to go through for Hermes to be back up and running quickly. Uh but it definitely created a chaos situation where everyone's agents broke all at once that were on a OpenAI including my DGX Spark. So that's how I knew it was specific. It wasn't Hermes, it was Hermes plus OpenAI because I have one agent one agent that was using uh MiniMax on this machine itself. This is an M3 and the other one on the DGX Spark was using OpenAI. And that's because the the tasks that I offloaded to that machine I felt were higher importance to get right the first time. Whereas the MiniMax agent on this machine was just a tinkering thing. Now I have all of them switched over to MiniMax M3 for right now just as an experiment to see A, how much can that $100 sub- or $100 a year subscription give me and how how how much can I, you know, how much can it actually handle because I've been shocked by MiniMax M3 in terms of the quality of this for the price. So not to be a shill for MiniMax but it's it's night and day. The difference from before to now. Now in terms of Hermes agent itself, where would I rank it? I can tell you it seems far slower than uh it seems far slower with any, you know, OpenAI uh you know, MiniMax, any model than any other harness, including the desktop app versus like if you compare MiniMax code's response times to Hermes agent's response times, it's nice at night and day. It is like Hermes is so much slower. In terms of the quality of the output, I don't think it's any better than MiniMax agent would be, but again, this is preliminary because Hermes agent I've only been using inside the desktop app for today. Whereas but I've been using Hermes agent outside of the desktop app for quite a while and I can on Telegram and I can tell you it's it's slower. Like it's slower than any of the other It's slower than you know, the only one that rivals it in terms of speed is definitely anti-gravity, which I'll get to next. But my personal opinion on these things is Hermes agent is definitely worth experimenting with. I don't think it is as great as people's hype. I do think that it's better than open claw in that in that it doesn't break every time you update it, but I will say the initial onboarding I had the Hermes agent desktop app I had to have I spent like an hour or so getting everything so that I could just get the switcher so that I could switch between models. And it showed all my models. That took like an hour. You know what I mean? And what's funny is I actually had offloaded the task to Claude code because I was frustrated with how slow Hermes desktop was to respond to me. And Claude code's response came back as Hermes agent was actually fixing it. With with MiniMax M3, Hermes agent was actually able to fix it at the almost the same pace as Claude code, uh which is actually pretty impressive. Like if you think about it, uh because of the difference in the power of these models. I mean, this is Opus 4.8, the most expensive, the most premier model out there, and this is MiniMax M3, which is far cheaper. So, if if I were going to rate this thing, it's I think Hermes Agent Desktop and Hermes uh in general are worth playing around with, but just realize that the setup on these things is not, you know, people are like, "Oh, yeah, you can just point it at your open claw, and it'll magically start working and have all of your memories." I don't believe it's that simple. And I do believe you have to be prepared for the migration. And I can tell you definitively when Hermes was first launched on this machine, it had none of the capabilities of of uh of Open Claw. And I had to actually pointed it at Open Claw and say, "Hey, figure out how Open Claw does this, and then you can do it, too." Uh one of the things that Open Claw can do is it can launch a real Chrome browser with your logged-in profile, and then navigate it. And I know that doesn't sound major, but that gets around all of these um all of these CAPTCHAs and stuff like that that are on the internet. And it can use your logged-in profiles, which I know is somewhat risky, somewhat dangerous, unless you've set guardrails around it, but realize that gives it the capability to break all these CAPTCHAs, to go on websites and do things as you, which is pretty powerful. I mean, it's not pretty powerful, it's extremely powerful. And uh Open Claw has that natively out of the box. I didn't have to tell it to do that. It just knows how to do that at this point in time. And to get Hermes to do that, it was within This was MiniMax 2.7 at the time, not Minimax M3. It was like seven prompts or something like that to get it to do that, but it did eventually figure it out at the end of the day, and it was able to do all of the things that my uh my Open Claw agent was able to do. And you're like, why did you use Hermes agent instead of Open Claw? Well, at the time my Open Claw was broken because I had run an update and I got tired of trying to fix it. So, that's what happened. I was just sick and tired of trying to fix Open Claw's updates. Uh with, you know, usually what I do is point Claw code or point Code X at it and say, "Hey, fix this thing." What happened at the time is Discord was broken, uh which is one of the ways that I I talk to and segregate out work for Open Claw. So, instead of having this concept of sessions, uh which you have here, right? And um and and inside of Code X you have, you know, projects, right? I have all these projects set up with individual chats inside of those projects. So, I have a pretty elaborate workspace here. Uh within Open Claw I was using Discord to to kind of build this structure. And and what's cool about that is is that accessible everywhere. Uh now I wanted to talk about the last of these tools because this one was the most frustrating not in terms of setup, but in terms of understanding the usage that you get for the dollar you pay. Because my gosh, is it confusing, and I finally figured out what I believe to be the optimal subscription to get for everyone, and what like what I think provides real value if you're using Google's models. And so, here's anti-gravity. And if I go in here to find out your usage, by the way, anti-gravity is another one where I cannot find it on their website, but in order to find the usage that you have available, you go into settings and then you go into models and you can see right now it looks like this. This is not how it is not how it looked earlier today. These were all like I used all of it the quota. But um you know, I had the end of my work day and then and then here. So um so like last night this was all used up and into this morning it was all used up. And so now because I would have had reoccurring tasks that were running as I was working through the day. So I had stuff that was still in process, but now everything's freed up and uh one big thing is switching models before. So these have different you can't see it here. I I should probably send it some messages so that you can see, but I'll just describe it for you real quick. So all of these Google models have one quota and then the Claude models have a separate quota. They the bars go down separately. So when you pick a Gemini model, you only have one like you don't actually have individual quotas for all of these. They all go down together. But the Claude models go down separately and if you had to switch context before between Gemini and Claude, it couldn't do that. Um and I don't know when that changed or if it changed because my subscription changed. So I don't know which one caused it to change, but I can tell you now I was able to today switch between Gemini and Claude. It had all of my contexts and switched providers and then just finished the task because I ran out of Gemini credits. So that's pretty cool to see. You can just switch the model and say continue on this and boom. It continued on it and finished it. Which is pretty awesome that you get these models included with these models. So the Gemini models the the Claude the Claude models. The Claude models obviously have a quota together and then the Gemini models have a quarter together. But, in terms of subscriptions, so let's talk about this set of subscriptions because I have never been more disappointed in a subscription service than this outside of what I have today. So, their subscriptions, because I've played around with all of them now, thankfully, if I go back into Safari, so their subscriptions look like this, right? They have AI Plus, AI Pro, and AI Ultra. What a mess. So, I can tell you inside of anti-gravity, I don't I didn't see any difference between AI Pro and AI Ultra. That's right, this $99 a month plan did not have >> Rachel Peters >> It did not have uh you know, it says up to 20 20 20x more limits in Gemini than the Pro plan. Well, I'm not sure where they're getting at that because my limits for sure, when I was in anti-gravity on that Ultra plan, looked like this and depleted extremely quickly. It was not good. It was not useful. Uh I I got a cheaper deal than this. I did not pay $100 for the month. I can't remember what I play paid, but basically I had a month or two at a discounted rate, and so I took that. And now, right now, I'm actually at the So, then what happened is I I had this plan, and I went down to this plan cuz I needed the storage. So, I needed the Gmail storage. So, that's the the impetus to move down. And then I wanted to play with anti-gravity again. And I can tell you that anti-gravity inside of this plan is absolutely useless. You might as well not even have it. Like it it it gives you a weekly or at least for me when I was using it, this model quota was weekly instead of like a 5-hour window. Which at a weekly level it is utterly useless. Like it's actually more frustrating to even bother uh even bother attempting it than to you know, than to to to just not use it. So, and and just in general trying to figure out what you get with it was absolutely crazy. But I again, I needed I needed the storage. So, I just decided to switch over to this. Then I got offered this for $10 a month for 6 months. And I was really worried what would happen to my Plus plan because by the way, as you're going through this and switching from Plus to Pro, it doesn't tell you what it's going to do with the Plus plan. They reimbursed the full price of the year for the Plus plan when I switched to Pro. Like they didn't prorate it. I couldn't believe it. Like they just basically gave me the money back cuz it'd only been like a couple weeks on Plus. And they just literally gave me all the money back, reimbursed my card and switched me over to Pro. Which is absolutely awesome. So, first of all, kudos to Google on that. I was not expecting that. I knew I was hoping it would be something prorated, right? Or you know what I mean? And that's what it said whenever I you know, I asked Gemini what would happen and it said, "Oh yeah, it'll be prorated." Uh but what they sent me back was the full price. And I can tell you the thing that I really really missed was the the YouTube ad-free that I had up in Ultra. And I do use YouTube quite a bit. And I like you know, I I like to have it in the background. And I like And I know there's hacks to do all this, but I I just want it to be simple. At my age and with my amount of time I actually have, uh I liked having ad-free and I liked having the ability to background it on iOS. And so, I desperately miss that and you get that with Pro. Plus, you get the 5-hour rolling windows for anti-gravity and this thing is actually useful. And it's good. That's what I couldn't believe. Gemini 3.5 flash is good. So, I've been really impressed with it so far. I switched over a couple days ago uh to this and if you can get that $10 a month for 6 months, uh plan, you know what I mean? You can get that cuz I don't think like right now I'm not logged in and I'm pretty sure I will not see um yeah, I'm not logged in with my main profile here and you can see I don't see the you know, I don't see the the promo that I got, right? So, I got $10 a month for 6 months on the plan. And uh that is I mean, that is huge, right? I don't see the that promo that I got. So, on this account. So, if you can if you get that promo, I strongly recommend everyone get that. I mean, that that is uh in my opinion of these plans, it's pretty incredible, especially if you use YouTube and especially if you want ad-free YouTube or background. Like, it makes it makes no sense to ever buy this plan when this plan is available and and you can actually use Gemini. Like, it it is you know, it says 4x higher usage, but that's not my experience. I mean, what I see in anti-gravity is this plan, AI Pro, matches AI Ultra, which it should not, by the way. But, that's what I'm seeing in anti-gravity from when I had the AI Ultra plan to now having the AI Pro plan. I don't know if I don't know why, you know, but I can tell you that is definitely the experience that I've seen with it. Anyways, that's what I wanted to talk about today. I wanted to to get I wanted specifically people to be able to see all of the different screens that you have for each of these experiences in terms of the usage. So, each of these each of these looks totally different. It's in a totally different place. Two of them are in app, right? And two of them are on the web, right? So, it's just a minefield of trying to find how you can track your usage inside of these plans. I strongly recommend plans over individual tokens because I don't want to have to worry about or think about the prompts that I'm giving, clearing sessions. I don't want to have to think about that is I mean, you do have to think about clearing sessions a little bit and freeing up memory context, but it's not as crazy as if you're paying per token. I feel like I would overanalyze what I'm sending to the AI if I was paying per token, and I strongly recommend of all of these plans, I would say that in terms of ranking them, if you can deal with multiple prompting, the MiniMax plan is the best. Like that $100 plan will actually get you uh get you a ton of usage. Now, if you want frustration-free, best for any tool that you could possibly want to use, whether that be Hermes, whether that be OpenClaw, OpenAI is your choice today. If you're locked into the Claude ecosystem or you want very specifically Opus, the latest Opus, uh the latest Opus's, then you're going to have to go with that. Like you don't have a choice if you really like Opus and Sonnet, other than to pay for a Claude subscription. I really debated about never paying Anthropic another dollar after the open Claude thing. But, I really really do like Opus 4.8. Uh I There is definitely model degradation for a while there, but it feels like it's back. Feels like it's just as smart as GPT-55, and each of them have different quirks of different things that they make mistakes on, and it's so good to have both of them to AB compare them. So, in my opinion, I have both, but what I've chose to pay the big bucks on right now is Codex. And I like to keep all of my subscriptions under that original $200. So, to give you an idea of the way my brain works, I want to keep all of my subscriptions in totality under the $200 that I was originally paying to Anthropic. So, that's the goal, and that's why I have all of these subscriptions to test them all. And even when I had that Anthropic subscription, I had the $100 MiniMax subscription to use that to actually back up in open Claude so that it could pass a whole bunch of tasks through MiniMax, and I would have a reliable uh basically an AB thing where I would have have my main orchestrator, which was which was Anthropic's models, call uh MiniMax 2.5, do a bunch of the tasks there, and then and then send it back up and have uh the the the main model, the orchestrator, that was running the Anthropic models, test that out and see the quality of it. So, that was the the original workflow that I had was you know, a $200 uh Anthropic subscription plus that $100 a year uh subscription for MiniMax. So, it was Those were the the the originators. That's why I have that MiniMax one is actually because I hit the limits almost every every time on Claude $200 plan. I can tell you I barely ever hit the limits on Codex. This week was the first time I thought I was going to hit it and then it man magically just reset to zero today. So, I would strongly recommend people and the $20 Codex plan was just not enough. I was hitting it all the time within within open claw. And so, I had to bump up to the $100 plan. I originally had the $20 Codex plan. I well, I originally had no Codex plan and I was using the free tier and they used to give you a ton of usage on the free tier, which is what got me hooked on the Codex app. So, in terms of being hooked on one of the apps, the Codex app is the one that I'm the most hooked on in terms of usage. I mean, you can see it has the most and this is after deleting a bunch of things from the projects because it was just too much. So, my personal opinions on this again to kind of summarize, anti-gravity with that with that pro plan is incredible. Mini max is just incredible across the board and M3 is really impressive as long as you can deal with multiple prompting because you are going to have to do that. Codex as an app in my opinion is the best, but the $100 plan is what you're going to have to go with if you really want to be able to use it, especially powering agents and and within the Codex app itself. The other ones just don't give you enough usage. Claude code, it's going to vary. I I find their desktop app to be garbage and maybe they fixed it, but I'm not giving it another shot after giving it two shots and it being horrible the first time with the UX being horrible. They fixed the UX and then it just started hallucinating on day two, something I've never seen from their command line. So, uh code if if but if you use that command line or you use that app, and you use a subscription, your subscription's going to be based on what you can tolerate. Right? What you have the money for and what you're willing to spend. Uh and outside of there, I don't think the extra usage is a good idea. I don't think powering an open claw with it is a good idea. You know, unless you've got, you know, unless you have some sort of sort of backing or business that you're running all this through, right? I'm doing this all self-funded, you know, no no business really, no plan for a business either. Just trying to learn and and experiment uh while these while these harnesses are are coming to pass. And Hermes desktop and Hermes agent, I'm impressed and disappointed in in the same in the same breath. And the reason why I'm saying that is because I am impressed at at the ability for updates to not break the system. So, I've never had a an update from Hermes break the thing, but also disappointed because of the amount of hype online about how great these things are. And I just don't find them to be that much better, you know, especially the setup, the ramp up and setup is actually more difficult than they make it seem online, like other influencers are trying to say that this thing uh you just install it in one shot it and then it works. And that wasn't my experience at all. E- each time I had to have another agent or another agent system uh make this thing work properly and make it use especially the the MiniMax models. It just wouldn't use them. Like, it just wouldn't use my OAuth for MiniMax until I pointed Claude code at it. And this is the benefit of having the Claude code command line. I just point Claude code at it and say, "Hey, here's a half-working Hermes install. Um you already know about my MiniMax subscription." Believe it or not, the Claude code command line knows about my MiniMax subscription and can call it. So, for people thinking that you just have to call Anthropic's models inside there, that's not true. I mean, it's capable of calling other models. And actually, I have a wrapper for it to call all of these free tier models that I have. So, inside my Hermes agent now, you can see that I have I have my subscription ones, OpenAI, MiniMax, and then the Google stuff, right? That I'm that I'm paying for. But, you can see NVIDIA NIM is in here. And if we go down to Hugging Face, which has a whole bunch of free models, and all of these are there. So, that is the that's kind of where I'm at with these. It's pretty incredible. I'd strongly recommend playing around with it for yourself. I'm sure my opinions are not going to be your opinions. There will be, I'm sure, a lot of people that have very different experiences and opinions of this. So, that's kind of the conclusion here. And yeah, I appreciate your time. Very different video than the fitness-related stuff, but figured it was time to go back on and do a stream and talk about these these harnesses and agents just because I've spent a ton of time with them and no one's actually brought them all together. And I can tell you it's it's a crap shoot right now. And I think the harnesses in general are where we're going to get AGI. So, everyone's talking about artificial general intelligence and thinking the LLMs are going to provide that. I was listening to an interview of the original one of the original founders of the the large language models of these LLMs. And I I think he's fundamentally right. And LLMs are not capable of two of the key things that are necessary to uh to truly create AGI. And um one, they don't have goals. And two, LLMs do not have the ability to be retrained once they're brought into the wild. Um at least the at least the frontier models in the United States. There's a caveat, the open weight and open-source models do have that capability. And we could see them leapfrog the frontier models because of that capability. However, the frontier models do not have that, and right now they are the leaders. But what those frontier model providers are learning, and it's pretty eye-opening, and they're learning it quickly, is I believe, and I think they believe also, these harnesses are going to be the true way to get to AGI. And not the model themselves. We're close, if not already there, in terms of model capability uh to produce AGI, artificial general intelligence, but I think the harnesses are where the final fight's going to happen because goals and retrainability have to be a thing, and I think these harnesses are where it's happening. You know, Hermes agent is literally trying to learn your preferences, have persistent memory, fix its issues. Like, that's the whole gimmick behind Sorry to use a like a pro wrestling term, but the whole thought process behind uh this Hermes agent. And I personally think that it's going to be a race, and we're seeing that now, where Google's got a concept that isn't anti-gravity, that is actually their new Spark system. Uh we're seeing which is an agent in the cloud, basically. Uh Hermes agent, obviously locally. Open Claw, which is obviously local to you. And then these other end Codex is actually getting discontinued next week, or soon, where they're going to be moving into a one one OpenAI one OpenAI system, which I imagine is going to look a lot like Open Claw or some sort of a merged system between all of these because honestly uh chat GPT Atlas should not exist. Codex should not exist. It should really be one package system. Uh and that that's what it looks like they're going to build soon. So, I I'm pretty hopeful that they're actually understanding the concept of the harness needs to be able to have goals and have and that's what everyone's showing off is their harnesses having goals, having retrainability. These are the things that are going to make or break uh AGI. So, thank you for your time. Uh little bit of a different video, but I wanted to get back on stream and get back to doing that and talk about something that I've been putting some time into.