Grok Bot, Hermes Agent, Claude Cowork, ChatGPT Work, and OpenClaw are all built from the same six parts. The video ranks them twice, then argues the strongest long-term move is skipping all five.
Posted
2 days ago
Duration
Format
Review
educational
Views
3.7K
126 likes
57 · 43
Big Idea
The argument in one line.
Every popular AI agent, from Grok Bot to Claude Cowork, is built from the same six core pieces, so the real choice isn't which one is best, but how much control you're willing to trade for setup speed.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You're already juggling or evaluating multiple AI agent products (Grok Bot, Claude Cowork, ChatGPT Work, Hermes Agent, OpenClaw) and can't tell what actually differs between them.
You've felt pressure to abandon an agent setup you just finished building because a new one went viral, and you want a framework to resist that churn.
You're technical enough to consider hosting your own runtime (VPS, Docker) and want to know if DIY is worth the tradeoff against a managed vendor product.
You want a vendor-neutral way to compare agent products on the same axes instead of trusting whichever one is trending on social media this week.
SKIP IF…
You've never used any AI agent or assistant product and don't yet have a workflow these tools would plug into.
You only use a single chat interface for one-off questions and have no interest in automation, scheduled routines, or multi-step agent work.
TL;DR
The full version, fast.
Every AI personal assistant, from Grok Bot to Hermes Agent to Claude Cowork, does the same six jobs: pick a model, run it somewhere, and manage files, tools, memory, and routines. The differences are about ownership. Grok Bot and Claude Cowork buy speed by handing that ownership to a vendor's managed cloud; Hermes Agent and OpenClaw hand it to you as an open-source runtime you host and secure yourself. Ranked for turnkey ease, Grok Bot and Claude Cowork top the list. Ranked for independence, building your own system wins, with Hermes Agent as the strongest open-source runner-up. The pitch: keep your files, tools, memory, and routines in a layer you own, and swap the underlying model without rebuilding anything.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Cold open naming Grok Bot, Hermes Agent, Claude Cowork, ChatGPT Work, and OpenClaw, and previewing the double ranking plus a build-your-own recommendation.
00:45 – 01:14
02 · The core pieces they all share
Every agent needs a model, a runtime (local or cloud), and some handling of files, tools, plugins, and memory. Vendors just use different names for the same layers (bots, projects, plugins, skills, gateways).
01:14 – 04:29
03 · Grok Bot
A turnkey, non-technical experience where every agent gets its own persistent cloud sandbox VM. Covers the $300/month X Premium + Grok Bot + Cursor bundle, and speculation that its viral traction is partly an X-platform push.
04:29 – 06:58
04 · Hermes Agent
An open-source agent that exposes the machinery Grok Bot hides: any model can be swapped in, 10 bots can run 10 different models at once, and the harness manages tools, prompts, context, and self-improving skills. Full DIY responsibility (servers, security, secrets) comes with it.
06:58 – 09:29
05 · Claude Cowork
A managed Claude workspace that sits between Claude Chat and Claude Code, was itself vibe-coded with Claude Code, and shipped near-daily updates after launch (a cadence that has since slowed).
09:29 – 10:48
06 · ChatGPT Work
OpenAI's answer to Claude Cowork, sitting between chatting and full Codex work. Newer and less battle-tested, but closing the gap on asset creation (docs, slides, spreadsheets).
10:48 – 12:04
07 · OpenClaw
The system that started the agent wave. Barely used today after Hermes Agent capitalized on its weaknesses, but its real contribution was being the first to stitch together many separate services, MCPs, and CLIs into one assistant.
12:04 – 15:13
08 · Why building your own may make sense
Makes the case for a DIY system with the same six core pieces (files, tools, memory, routines, permissions, interfaces) that swaps in whatever model or feature goes viral next, without ever needing to migrate everything to a new vendor.
15:13 – 15:52
09 · Ranking them for turnkey setup
S-tier: Grok Bot and Claude Cowork. A-tier: ChatGPT Work. B-tier: Hermes Agent. C-tier: OpenClaw.
15:52 – 16:37
10 · Ranking them for power and independence
S-tier: building your own system. A-tier: Hermes Agent. B-tier: Claude Cowork and Grok Bot. C-tier: ChatGPT Work. F-tier: OpenClaw.
16:37 – 17:20
11 · The complete capability comparison
A 12-row grid (always-on execution, model routing, browser/web, app connectors, files and memory, scheduled routines, agent orchestration, messaging channels, local device reach, reusable skills, finished deliverables, operator ownership) scored across all five products.
17:20 – 17:38
12 · How to choose the right assistant
Closing framing: minimize time spent chasing the next hyped product and instead invest in whichever setup you actually commit to.
17:38 – 18:18
13 · The free AI Assistant Decision Kit
Points to a free downloadable kit (interactive comparison map, research dossiers, a decision prompt, a one-use advisor skill) and the Early AI-dopters community.
Atomic Insights
Lines worth screenshotting.
Every AI personal assistant, regardless of branding, is built from the same six parts: a model, a runtime, files, tools, memory, and routines.
Grok Bot's full bundle (X Premium plus Grok Bot plus Cursor) costs $300 a month, only $100 more than a Claude Code or Codex Max plan.
Hermes Agent lets you run 10 different bots on 10 different underlying models at once; Grok Bot locks you into Grok models with no way to swap.
Claude Cowork was itself vibe-coded using Claude Code, before Anthropic made the Claude Code desktop tab usable for non-technical users.
OpenClaw's real innovation wasn't its harness. It was the first system to stitch dozens of separate services, MCPs, and CLIs into one working assistant, the equivalent of the four-minute mile for agent tooling.
For turnkey ease, Grok Bot and Claude Cowork rank S-tier; for power and independence, building your own system beats every managed product.
Anthropic shipped near-daily updates to Claude Cowork after launch, then slowed to weekly, then monthly. The same slowdown pattern is worth watching for in any hyped product.
The creator speculates that Grok Bot's viral spike on X may be partly incentivized by X and Grok now being merged into one ecosystem, similar to how Google's Antigravity got outsized YouTube reach right after launch.
DIY agent systems carry full responsibility for server setup, execution profiles, security, and environment secrets, work a managed vendor product absorbs for you.
The six core pieces the video maps onto every DIY build are: a file structure, tools, memory, routines, permissions, and interfaces.
Takeaway
Every AI agent is the same six parts, wearing different branding.
WHAT TO LEARN
Before switching to whatever agent product is trending this week, map it against the same six layers, model, runtime, files, tools, memory, routines, and ask how much of that control you're actually willing to hand over.
02The core pieces they all share
Every agent, regardless of vendor, needs a model to think and a runtime (local or cloud) to actually do the work.
Vendors rename the same layers, calling memory/tools/plugins 'bots,' 'projects,' 'skills,' or 'gateways,' which is why products that look different on the surface function almost identically underneath.
03Grok Bot
Grok Bot's core promise is a turnkey, non-technical setup where every agent gets its own persistent cloud sandbox, so work keeps running after you close your laptop.
The full bundle (X Premium + Grok Bot + Cursor) runs $300/month, only $100 more than a Claude Code or Codex Max plan, positioned as one ecosystem at one price.
Locking you into Grok models with no model-swap option is the direct tradeoff for that turnkey convenience.
04Hermes Agent
Hermes Agent's differentiator is model freedom: you can run 10 different bots on 10 different underlying models simultaneously, something Grok Bot can't do.
Its harness (the layer that manages writing, editing, searching, skills, and plugins) is designed to work well regardless of which model you plug in.
Full access to the underlying code means you can have another model inspect and rewrite how Hermes Agent behaves, but you also inherit full responsibility for servers, security, and secrets.
05Claude Cowork
Claude Cowork was itself vibe-coded using Claude Code, before Anthropic made Claude Code's desktop tab usable enough for non-technical users to skip the terminal entirely.
It offers a one-connector-button experience for organizational tools, plus a cloud environment so agents and sessions can run without keeping a screen open.
Anthropic's update cadence for Cowork went from near-daily after launch to weekly, then monthly. That kind of slowdown pattern is worth watching in any hyped new product.
06ChatGPT Work
ChatGPT Work is OpenAI's answer to Claude Cowork, positioned between casual chatting and full Codex-level work.
It's newer and less battle-tested than Cowork, but has closed much of the gap on producing finished assets like docs, slides, and spreadsheets.
Choosing between the two comes down mostly to model preference once you're not locked into a single vendor's ecosystem.
07OpenClaw
OpenClaw was the system that planted the seed for the whole agent wave, but it's barely used today after Hermes Agent capitalized on its weaknesses.
Its real contribution wasn't the harness itself. It was being the first to successfully stitch together many separate services, MCPs, and CLIs into one working assistant.
Studying its architecture is still worthwhile even if you wouldn't actually run it today.
08Why building your own may make sense
A DIY build reuses the same six core pieces (files, tools, memory, routines, permissions, interfaces) that every managed product also has, just under your own control.
The pitch for owning your layer: when a new product ships an impressive feature, you extract and rebuild just that feature into your own system instead of migrating everything.
This avoids the repeated cost of 'moving houses' from one vendor's ecosystem to the next every time a new agent goes viral.
09Ranking them for turnkey setup
Ranked purely for getting started fast: Grok Bot and Claude Cowork are S-tier, ChatGPT Work is A-tier, Hermes Agent is B-tier, and OpenClaw is C-tier.
10Ranking them for power and independence
Ranked for independence and long-term control, the order flips: building your own system is S-tier, Hermes Agent is A-tier, Claude Cowork and Grok Bot are B-tier, ChatGPT Work is C-tier, and OpenClaw drops to F-tier.
The two rankings deliberately point in different directions to make the point that 'best' depends entirely on what you're optimizing for.
11The complete capability comparison
A 12-category grid (always-on execution, model routing, browser/web, app connectors, files/memory, scheduled routines, agent orchestration, messaging channels, local device reach, reusable skills, finished deliverables, operator ownership) shows most products share the majority of these capabilities.
Where products actually diverge is a small set: model routing, messaging channels, and operator ownership, which is the real decision surface once table-stakes features are equal.
12How to choose the right assistant
The closing framing is to minimize time spent learning and re-learning a new tool, and instead commit to whichever setup you'll actually keep using and get leverage from.
Glossary
Terms worth knowing.
Harness
The layer of an AI agent that manages how tools like writing, editing, or searching get invoked, largely independent of which underlying model is running.
Sandbox
A persistent cloud virtual machine an agent product spins up so a bot can keep working even after the user closes their laptop.
Skill
A reusable, storable capability an agent learns and can call again later, instead of re-explaining the same task from scratch each session.
Model routing
The ability to swap which underlying AI model (Claude, Codex, Grok, etc.) powers an agent without rebuilding its tools, memory, or workflow.
MCP
Model Context Protocol, a standard way for an AI agent to connect to outside tools, services, and data sources through a shared interface.
“So I'll let you in on a secret. They're all pretty much the same.”
The thesis of the whole video in one line, delivered as a confession.→ TikTok hook↗ Tweet quote
02:28
“The real difference is Grok Bot doesn't make you think. It gives you that exact same experience without you having to work for it.”
Sharp, quotable contrast between managed convenience and DIY effort.→ IG reel cold open↗ Tweet quote
11:40
“It was equivalent to the four minute mile. All of these different services had never been combined together in this way, but once we saw it, we couldn't unsee it.”
A strong historical analogy that reframes OpenClaw's real contribution.→ newsletter pull-quote↗ Tweet quote
19:15
“If someone is telling you that there's a winner for you, they're lying or they're shilling a brand new product.”
A blunt, contrarian closing line that works standalone.→ TikTok hook↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphor
I've tested every personal assistant that's launched this year, and I've built real systems with all of them. Whether it's OpenClaw, Hermes agent, Claude co work, chat g b t work, and yes, even Grockbot.
I've even spent months building my own frameworks as well. But every few weeks, social media tells us to take the very system that we're finishing setting up and throw it away and replace it with a brand new one. So I'll let you in on a secret.
They're all pretty much the same. By the end this this video, you'll know exactly which one fits for you, when it's worth switching, and when it might make sense for you to build your own. And I'll even go over a couple theories as to why I think certain agents seem to be going viral overnight.
Let's get into it. So when I say that all of these different agents are basically the same, I don't mean that they're identical. I just mean that all of them have very similar core pieces that make up how they work.
So each one of them needs a model to think and then some form of computer or runtime to do the actual work, whether it's on your local laptop or in the cloud. And then when it comes to file management, tools, plug ins, memory, all of them have some semblance to this.
It's just a matter of how configurable, how malleable, and how many layers there are. One company calls them bots, and others call them projects, plugins, skills, and even gateways.
So you pretty much have the same names but different jobs. I'm gonna do a quick overview over the top five and then not only will we rank them, but we'll take a look at all the functionalities that they share. Now the first one I'm gonna go over is the new kid on the block called GrockBot.
Now GrockBot has seemingly gone viral on certain platforms versus others, but the core thing that it promises is a turnkey experience for someone who is nontechnical and just wants to spin up an app, set up a bunch of agents, and get them to work.
And the core feature that people are focusing on is the fact that all of your agents technically have their own sandboxes in the cloud. So they essentially have their own computers that they can work on even if you shut down your laptop and go on the road.
Now this concept sounds new, but it isn't. You've been able to run things like OpenClaw and Hermes agent on a virtual server for months now. The real difference is is that Grokbot doesn't make you think.
It gives you that exact same experience without you having to work for it. Now if you were to take an X-ray machine and take a look at what Grock Bot looks like underneath the hood, this is what you'd see. First and foremost, you would see the persistent cloud virtual machine, and this is the sandbox that all of these agents are able to use to be able to run things like research, use plug ins in the cloud, and take advantage of not depending on your computer.
Next up, you have parallel bot screens. So you essentially have all of these different agents that have access to the exact same cloud, but their own screen, which is a blank slate through which they can use the browser, plug ins, connectors, and anything you enable for them. And this overall setup makes it so that if you shut down one bot and it needs to continue its work with another agent, they all share the same workspace.
So one agent will be able to access the files and folders used and worked on by another. And similar to other agents that you might have used in the past, they have access to things like skills, routines, and cron jobs.
Jobs. So a few reasons why Grok bot seems to be winning the spotlight and gaining a lot of attention, and one of them is more speculation on my part.
Number one is that it's very easy to use. Now reason number two is because Grok and Cursor have now merged into one entity, you can now have an entire brand new ecosystem for a single price.
So for a price tag of $300, which is very steep, but only a $100 more than something like a Claude Code or Codex Max plan, you get access to the whole suite of Xpremium, Grokbot, Cursor at a very high level and being able to take advantage of all of them working in unison.
Now one speculation I have is that if you look at most creators on YouTube that have made a video on how Grokbot is amazing, insane, or changed their lives in two days, you'll notice that most of them have underperformed. Now if you head over to x, you see a completely different story. And x is the very platform that integrates well with this whole ecosystem and theoretically could be giving an artificial push for all the posts and testimonials on this product.
Now if you think this is outlandish, go back to the very beginning of 2026 when Google's anti gravity came out. Not only was it exceptionally popular, but every single video on it on YouTube seemed to go viral overnight.
Even smaller creators were getting three to five times the reach on their videos. Now could it have been because of that new car smell? Sure.
But it's possible that since YouTube and Gemini are owned by the same family that the algorithm gave an extra bit of love, a bit of incentive for anyone to make a video on that product. But again, that's all hypothetical. Now if Grock Bot sells the finished experience, then Hermes agent exposes the machinery underneath it.
Hermes is an open source agent that seemingly was born out of all the chaos, the craziness, and the hype that came with OpenClaw. One of the major areas where Hermes agent differs from Grok bot is that in Grok bot, you are stuck with Grok models. You can't change the effort.
You can't even change the model as of today. I'm sure that will change in the future. But with Hermes, you can swap any model you want, and you can even have a setup where you have 10 different bots with 10 different underlying models.
And like I said, if you wanna be able to give all of the agents their own screen that's connected to their own workspace in the cloud, that's been something that you could vibe code just by asking Claude Code or Codex to make it for this Hermes agent. But the real claim to fame for this assistant is the underlying harness.
Harness being the ability to manipulate different tools like writing, editing, creating, searching, using skills, and plugins all within one scope that work really well irrespective of whatever model you choose. Especially if you use the newest version, you're gonna notice that the way that Hermes agents manage prompts, context, skills, learn skills through conversations, and understand exactly when it might make sense to be able to store a conversation as a memory and or a skill and keep iterating on it over time and self heal and self improve.
And just like you can with Grok bot, you can create cron jobs, schedule triggers, routines in the cloud, and spawn parallel agents. But because you have access to all the underlying code, you have access to the entire plumbing.
So if you want Grok, Cloud Code, or Codex to look at it and inspect and change the way that the Hermes agents behave, you can.
And luckily, we're getting to the point where you can use models that are sophisticated enough where you just give the command in plain English on what you want it to do or not to do, and you can start to manufacture and structure the harness every which way you like. But naturally with freedom also comes burden and responsibility.
So you're still on the hook to manage things like setting up those servers in the cloud, being able to manage your execution, your profiles, your security, your environment secrets, and everything else that would keep the system running and making sure that you maintain it moving forward.
So this makes sense if you want a DIY, but not if you want the model provider or the vendor to make all those decisions for you and all you have to do is just use it. Now what if you're not a fan of Grok or Elon or both and you don't want a DIY and you want something that just works out of the box and it's kind of a middle ground between something like Claude Code and Claude Chat.
Claude Cowork was born by being fully vibe coded using Claude Code and it went viral for all the right reasons. Before Anthropic made the Claude Code tab on the desktop app bearable enough for someone who's non technical to actually use it and actually go the distance with it, Claude Cowork was your only option. So it's either a terminal or Claude Cowork.
So for nontechnical folks who just wanted to do basic and straightforward knowledge work, Claude CoWork was amazing because you got one connector button, you could click on it, access any service that typically was used in your organization, and you were on your way. And Anthropic was really one of the first companies to come up with the concepts of skills, projects, MCPs, so it made a lot of sense to invest in this ecosystem since you could use the cloud models.
You could have a very straightforward experience where you could embed your prompts, add knowledge files, and go from there. And before Grokbot made it cool, co work was also running on a cloud environment where all of your agents and all of your sessions would run there as well. You shouldn't have that experience of having the different screens of computer use to really give you that wow factor.
And as this product went viral, the entire team kept shipping almost every single day a brand new feature or addition. So you eventually got the ability to take clawed co work on the go with something like dispatch. So you had that OpenClaw experience without all the baggage, the burden, the updates, and all the security night mares.
And one of the best parts was and still is that you can manage your sessions either with local files on your computer or with assets you have in the cloud. If you're a solopreneur or a small business, then you might be nimble enough to be able to jump from Claude co work to chat GBT work as soon as all the Claude models recently became hard to work with.
Claude slopus, if you will. But if you're a company that where you had to really dig your heels in and choose Claude as your ecosystem, there are a couple concerns that come to mind.
Like I said, at some point the Anthropic team was shipping updates to co work daily. And we went from daily to weekly to monthly to now once every while you might accidentally see a brand new update. Update.
So the very pattern that I saw here, I get worried about when I look at something like Rockbot because today it is very fashionable. It is very hype worthy, but we don't know how many updates are going to be shipped this product before the companies need to pivot again to a brand new direction for whatever it is that is gonna be the hottest thing in two to three months.
Now OpenAI saw Claude Cowork and said, will raise you a chat GBT work. Initially, when this came out, it seemed very random because Codex was doing a great job, and there was a good bifurcation between ChatGPT where you chat versus Codex where you work.
So identical to Cowork, this product is meant to sit in between chatting on the web versus using codecs. So riding a bicycle, if you will, versus driving the Lamborghini where you can retrofit it and make it do whatever you want with codecs. And whereas OpenAI used to be horrific at making assets like docs, slides, Excel files, nowadays, it's on par to sometimes better than Claude Cowork at those same tasks.
And as is true with all of these options, you still have the ability to upload files, research, create scheduled tasks, and every single thing that you can do with Cowork and Rockbot. So assuming that you're not locked into a vendor, then using Chat 50 Work versus Cowork would really be a difference of what model do you like more.
If you had to make a bet, I'd probably pick co work just because it's been in the market longer and the longer a product has existed, the more feedback it's gotten, the more bugs it's gone through. So since ChatGPT work is still relatively new, if you were to make a bet on which product, I'd probably take the former versus the latter.
And last but certainly not least, we have Open Claw, needs no introduction and barely needs a description of what it does. And while barely anyone I know uses this today, it was the initial seed that was planted that basically gave rise to the agent revolution.
Now if you look at Open Claw today, there's actually vast improvements compared to where it was in February. But because it lost favor and it fell out of style and Hermes agent took advantage and capitalized on all of its weaknesses at the right time, you'll barely find most people actively using OpenClaw unless they really dug in and made that ecosystem their own.
Now behind the scenes, the most impressive part of OpenClaw wasn't the harness, but it was more so all of the different services, all the different MCPs, the CLIs, they were able to be stitched together and hooked up to create this Frankenstein Jarvis system. It was equivalent to the four minute mile where all of these different services had never been combined together in this way, but once we saw it, we couldn't unsee it.
So although it might be a good infrastructure and architecture to study, most likely you have much better options out there, especially if you want a turnkey experience with a product that is continually being updated. Then before we get into ranking all of them and just looking at a final comparison to make sure you understand exactly what might make sense for you, it's important to talk about the alpha option, which is building your own system, which is exactly what myself and my community have done.
To build your own system today is easier than ever because we have the smartest models that will continually get smarter and all you have to do is take the GitHub link of a series of different repos, let's say OpenClaw, Hermes, other ones that are derivative that you can find, throw them into a brand new session, Codex, Clawd, Grok, and just tell it, I want to be able to build my own assistant that does x y and z.
And like I said at the start, all of these products have the identical six core pieces. You have a file structure, a file system, you have tools, you have memory, you have routines, permissions, and interfaces.
And especially when it comes to file management using tools and most importantly creating and managing your own memory layer. These are very special snowflake features.
So if you use something like a Grok bot, you can never change the way the memory system works. You can't add your own layers. It's not meant to be broken apart and you can swap in and out different components.
So you could theoretically make a full wish list of every single feature that you wanna be able to create and clone and give it to your model of choice whether it's a codec sole or a Claude code fable and have it build your own harness that you can keep improving day after day. Because I'm telling you now, even though I showed you the top five, there's gonna be a sixth.
There's gonna be a seventh. There's gonna be another new product that changes the world and makes you $10,000,000 in fifteen hours. And I'm gonna tell you right now that it's gonna have a series of elements that are very similar to what you've already seen, but slightly repackaged.
Maybe some quality of life features that you can adapt to your own setup. And this is the exact same thing that I've been using ever since March of this year when I finally went away from the OpenCLaw and built our own version that we called ClawedClaw. And the whole point of the system is that every single time we see a brand new feature, if it's open source, we will just look at the repo, see what's good, extract that good, bring it into our ecosystem, and the same thing if Grokbot has this amazing orchestration feature that feels so turnkey, I will use it, observe it, and then reverse engineer it.
Well, this doesn't give you the polished experience, you own your infrastructure and as things come out, which they will never stop coming out, you'll never feel stressed to feel like you're moving houses from one bot to the next and spend more time moving and learning than actually using this to drive business value. And by the way, if the sound of building your own system sounds daunting but extremely appealing, then you might wanna check out the first thing down below for my early AI adopters community.
We initially built our own system inspired by OpenClaw all the way back in April that we called ClawdClaw. And ever since then, we've made over 500 updates to that product to get it to a place where it's more than enough to use your existing Codex or Clawed subscription with this product to not pay additional fees and use it the same way you would Hermes Open Claw but on steroids.
So if that resonates to you, then check out that link and maybe I'll see you inside. Alright. Back to the video.
Before we rank them, we have to ask a specific question. Are we ranking them based on the turnkey experience or the potential of pushing them to become really powerful. If we ask on the turnkey experience, then I would put Grock Bot at the very top and then followed by Claude Cowork.
And then from using it, just because it's not fully there maturity wise, I would put Chatuchitee work in the a tier, and then you'd have Hermes agent in the b tier, and then we'd probably throw Open Claw in the c tier. And then I'm not gonna mention any one of these other providers. I just wanted to show that there are many other options.
Let's say Manus was something that I used to use back in the day, now not so much. Now if you were prioritizing independence, your ability to push a system to its limits and have full transparency as to what happens underneath the hood, the ranking would look very different. Now if you value making your own system and building it your own way on your own terms with the models of choice, then s tier would be building your own system like I said.
A tier would be Hermes agent. B tier would be Claude Cowork and Grok bot. Probably give an inch to Grok bot right now just because all the focus is on it from the team.
They wanna make it amazing and pretty much every single day similar to co work, you're seeing brand new updates. Chachity work would be c tier and unfortunately, I'd have to plop open claw in the f tier. To bring everything together, if you take a look at this map, this is the very reason why I can't tell you why there's no definitive winner.
If someone is telling you that there's a winner for you, they're lying or they're shilling a brand new product. If you have always on execution, files and memory, scheduled routines, local device reach, reusable skills, model routing, browser and web, app connectors, messaging channels.
All it takes is hovering over either the closed source or open source options, and you'll see that for the most part, they share the exact same elements. And where they differ is the ability to add things like model routing, messaging channels, operator ownership, etcetera.
So hopefully this video gives you the clarity that you need to make the decision as to what is the best personal assistant setup for you. You ideally wanna lower the amount of time you spend learning and keeping up with the brand new thing versus just digging your heels in and investing and getting leverage over whatever setup you end up choosing.
And to make your life infinitely easier, not only am I gonna give you the web page that I walked through in this video so you can use it, share it with whoever, and find the clarity you're looking for, but I'll also give you a throwaway skill. Meaning, it's a skill that you can use once to help you really understand based on a series of questions that it will ask you what is the best setup for you.
You'll be able to find both of those in the description in the second link down below. And as always, if you want a place to take refuge from all the hype and focus on the very things that will move your business forward and adapt our own Claude Claude system that we've built over months and made it work with both Claude, Codex, and any model you can imagine, then you wanna check out the first thing down below for my early AI adopters community.
For the rest of you, if you found this helpful, I'd super appreciate a like on the video, share it with someone that might benefit from it, and leave a comment if you so choose. I'll see you on the next one.
The Hook
The bait, then the rug-pull.
The creator opens by naming five products by name in one breath, then admits the twist up front: he thinks they're all pretty much the same underneath the branding, and by the end of the video you'll know exactly which one fits you, when it's worth switching, and when it makes more sense to build your own.
Frameworks
Named ideas worth stealing.
00:45list
The six core pieces of any AI agent
Model
Runtime (local or cloud)
Files
Tools
Memory
Routines
Every agent product, no matter the branding, is assembled from the same six layers. The video reuses this checklist to evaluate every product that follows, then re-lists it as files, tools, memory, routines, permissions, and interfaces when describing a DIY build.
Steal forEvaluating any new SaaS or agent tool before switching: map what it actually replaces layer-by-layer instead of reacting to the launch hype.
16:37model
The 12-category capability map
Always-on execution
Model routing
Browser and web
App connectors
Files and memory
Scheduled routines
Agent orchestration
Messaging channels
Local device reach
Reusable skills
Finished deliverables
Operator ownership
A single grid used to score Grok Bot, Hermes Agent, Claude Cowork, ChatGPT Work, and OpenClaw against each other on the same 12 dimensions at once.
Steal forBuilding a like-for-like comparison table before adopting or dropping any agent/automation tool.
CTA Breakdown
How they asked for the click.
VERBAL ASK
17:38product
“Check out the first thing down below for my early AI adopters community... and grab the free AI Assistant Decision Kit, the second link down below.”
Soft-pitched twice (once mid-video at the build-your-own section, once at the very end), tied to a genuinely free lead magnet (the Decision Kit) rather than being the only ask, so it reads as a resource drop rather than a hard sell.
A creator walks through the local tool he built that routes to 37 image and video models, then reverse-engineers exactly what a subscription platform like Higgsfield is doing so you can rebuild it yourself.
A breakdown of why maxing out effort settings on Claude, GPT, Grok, and Gemini rarely makes the output better — and the framework for picking the right level every time.
The same update that let Claude Code and Codex turn a screen recording into a skill quietly taught both of them to watch raw video, no plugin required.
A Claude Code creator mines his own chat history into a personal test pack, then builds a slash-command benchmark that tells him in one run whether a new model release is actually worth switching to.
A 44-minute walkthrough of the five-layer stack behind running open-weight AI on your own hardware — and the trick of using Claude Code itself to build the whole thing for you.
00:15
00:33
00:59
01:09
01:16
01:41
01:54
02:27
02:43
03:00
03:15
03:32
03:49
04:02
04:20
04:32
05:02
05:19
05:29
05:43
05:57
06:05
06:24
06:32
06:51
07:18
07:32
07:46
08:04
08:17
08:27
08:36
08:54
09:08
09:36
09:48
10:05
10:15
10:26
10:41
11:07
11:19
11:27
11:50
12:16
12:25
12:40
12:54
13:09
13:23
13:38
13:53
14:07
14:22
14:36
14:46
15:02
15:19
15:45
16:01
16:31
16:58
17:09
17:27
17:44
17:58
18:05
Video of the Day13:01
16:17
08:38
13:12
08:16
44:41