Why I Switched From Claude To Hermes Agent transcript

I still use Claude, but it's stopped being the AI I run my life with. And the weird part is it was not because Claude got worse. Claude is still amazing. I still use it for web design, coding, and focused product work. But once I started using Hermes, [music] I realized I was comparing the wrong thing. Claude gives me an incredible AI session, while Hermes has become a full AI workforce. It lives in Telegram, where I already work.

It remembers how I want things done. It can build skills when I correct it. >> [music] >> It can run scheduled jobs, workflows, manage emails, and improve every single time I use it. That is the difference most people miss. Claude is built to give you the best possible AI experience. Hermes is built to become a better companion and operator the longer it works with you. So in this video, I'm going to show you why I use Claude less now, what I still use it for, what I moved to Hermes, and why I think Claude simply is not enough anymore.

Let's get started. Before I show you the actual workflows, I need to clear something up. I'm not going to sit here and say Hermes is smarter than Claude, or Claude is smarter than Hermes, or run both of them through some random coding challenge that does not reflect how I actually work. That is not the point. Claude is one of the best AI products in the world. Claude code is extremely good. If I'm building a web-first product, working through a front-end, designing a page, debugging a website, or doing a focused dev session, I still reach for Claude.

I'm not pretending I deleted it from my life. The real switch happened somewhere else. I started realizing that Claude was the place I went when I wanted help. Hermes was becoming the thing I built work around. That sounds like a small difference, but it is the entire video. Claude is an incredible assistant inside of a session. Hermes is closer to an operating layer. It has tools, memory, skills, cron jobs, messaging, profiles, and workflows that can live outside of one conversation.

So, the question is not which one gives a better answer to a prompt. The question is which one can actually become part of how I operate every day. And the first place that difference shows up is where the tool lives. Claude is something I open. Hermes is something I message. And at first, that sounds like a tiny difference. Who cares if I open a website, open an app, or message something in Telegram? But once you actually use AI every single day, you realize this matters a lot.

Most workflows do not fail because the AI was not smart enough. Most workflows fail because they are annoying to start, annoying to repeat, or they live somewhere outside your natural day. If I have to stop what I'm doing, open a separate app, rebuild context, explain what I want, paste in the right files, and then move the output back into the place where I actually work, I will do it once or twice, but I will not build my life around it.

Hermes changed that because it lives where I already am and wherever I decide to go. For me, that is Telegram. I can be in a normal conversation and a topic on my phone, away from my desk on vacation, and still work with the agent. I can ask it to check something, create something, build something for me, or prepare something without switching into some special AI work mode. That is why features like the react workflow and the email workflow matter so much for me.

They are not impressive because they're impossible to imagine with another tool, they're impressive because they are effortless to create and use inside Hermes itself. I can tell Hermes what I want in the place I already talk to it, and it can turn that into an actual workflow. That is the difference between an AI tool you visit and an AI system that starts to live with you. Claude can give me an amazing answer. Hermes can become the place where the work happens, but the location is only the first layer.

The bigger reason I switched is what happens after Hermes does something once. This is the part that made Hermes feel fundamentally different to me. Claude can be customized. You can give it instructions, you can create projects, you can use Claude code with different files, documents, commands, and if you're technical enough, you can manipulate Claude into doing a lot. But, with Hermes, learning is not a workaround, it's not a hack, it's not something you need to manipulate into it.

Learning is the design. Hermes has memory, it has skills, it has persistent preferences. It can save procedures, it can be corrected, and those corrections can become part of how it works with you next time. That is the thing I kept feeling with Claude. I could get a great result, but I often had to teach the session. With Hermes, I feel like I'm improving the operator every single time I use it. If Hermes makes a mistake and I correct it, the correction becomes durable.

If we discover a better way to do a task, Hermes saves that as a skill or it updates an already existing skill. If I repeatedly ask for a specific style, a specific workflow, or a specific standard, it carries that forward. And imagine what happens after months or years of working with Hermes, it just improves and gets better and better and better. That is a totally different relationship with AI. Most people use AI like a disposable conversation.

They start a chat, get an output, close it, and then the next time they come back, they have to recreate the same context again. Hermes is moving in the opposite direction. The more you use it, the more specific it becomes to you. And this is also where the company difference matters. Anthropic is trying to build the best possible AI model and product experience. That is a massive goal, and they're very good at it. You can see that in Claude code, Claude Claude design, and all the different experiences they are pushing.

New research with Hermes is aiming at something completely different. The goal is to make the best possible AI companion and workforce. And that difference shows up in the product. Every update is not just about making the chat feel better. It is about making Hermes a better companion, a better worker, a better operator, and a better system that develops and grows around you. That is why Hermes feels like it compounds.

A new AI tool can launch tomorrow with a better model, a prettier UI, or some crazy demo, but it does not instantly have my workflows. It does not instantly know how I like things done. It does not instantly have the skills we built. The more time passes, the harder Hermes becomes to replace. And that leads into the next big difference. Claude waits for me. Hermes can start working without me. The biggest limitation of a chatbot is that it waits.

It can be the smartest chatbot in the world, but most of the time it is still sitting there waiting for you to show up and ask something from it. That is fine when the task is creative or exploratory. If I want to design a page, brainstorm an idea, write code, or think through a product, waiting is not a problem. I'm already there. But a lot of real business work is not like that. Real business work is recurring. It is checking, monitoring, following up, reporting, reminding, reviewing, catching problems before they become exp- ensive.

That is where Hermes started feeling different. Hermes can run scheduled jobs. It can check things on a cadence. It can watch for signals. It can send alerts. It can deliver reports. It can create workflows that do not depend on me remembering to open a chat and ask for something. And yes, to be fair, Claude code is getting better here as well. Claude code is on mobile now. Claude code has more ways to be accessed remotely.

The whole AI tool space is moving in this direction. But mobile access and proactivity are not the same thing. Mobile means I can reach the AI. Cron means the AI can reach me. That is a huge difference. If I'm away from my desk and I remember something, Claude mobile is useful. But if a workflow breaks, a report is due, an email needs response, or client signal appears, I do not want to rely on myself remembering to check.

I want the agent to wake up, look, understand what changed, and tell me what matters. That is why Hermes works better for the way I actually operate. It can help with email workflows. It can help with reaction workflows. It can help with notion. It can help with recurring reports, with alerts, and because it has memory and skills, those workflows can become more specific over time. That is the moment where Hermes stops feeling like a chatbot and starts feeling like a workforce.

A chatbot helps when you ask it. A workforce just works. It has jobs. But, if an AI is going to have jobs, then it needs to understand how you want those jobs done. And that is where personalization becomes important. Every AI tool is going to copy features. This is just how the market works. One tool adds a feature, then another tool adds something similar. One tool gets mobile, another tool gets mobile. One tool gets agents, another tool gets agents, so on and so forth.

So, if your entire reason for using something is one feature, that reason might disappear quickly. But, personalization is different. >> [snorts] >> If Hermes has been working with me for a month, it is not just a list of features anymore. It has accumulated context. It has learned preferences. It has built skills. It knows the difference between how I want an email handled, how I want a script written, how I want a Telegram reply phrased, how I want a workflow verified and improved upon. >> [snorts] >> That does not mean it is perfect.

It still makes mistakes. It still needs correction. It still needs human approval for anything that can be external or risky. But the important part is that the corrections matter along the way. When Hermes gets something wrong, the answer is not just try again in this chat. The answer can be fix the way you do this for all the times you do this going forward. This is why this becomes harder to compete with over time.

A new tool can be impressive on day one. It can have a better demo look cleaner. It can even use a stronger model, but it has to compete with the version of Hermes that has already been shaped around my work. That is the most underrated thing in AI right now. People keep asking which model is the best this week. But the long-term question is different. Which AI system is becoming more valuable the longer it is with you?

And for me, that has been Hermes, which is also a reason why when Claude discontinued the ability for you to connect it through your subscription to Hermes, a lot of people panicked. A lot of people thought this would be the end of open Claude and Hermes and you know, AI assistants, but come to find a lot of people just switched to a different model provider like ChatGPT subscription or the Codex subscription because they just loved Hermes so much.

It did so much that Claude or any other model on its own could not do. That is why I do not think the future is just better chatbots. I think the future is AI systems that become personal operating layers. But to be clear, that does not mean Hermes replaces Claude for everything. I want to be honest here because otherwise this video becomes useless and a little dishonest. I still use Claude. Mostly, I use it for web design and development.

If I'm building a web first product, designing an interface, or working through a page, or doing a focused dev session, Claude code is still one of the best tools in the world. There are times where I want exactly that kind of environment. I want to sit down, focus on a product, go deep into the code base, iterate on the design, and use a tool that is optimized around that experience. And Claude is excellent for that.

So, I'm not saying never use Claude. That would be stupid. The way I think about this is the following. Claude is still one of my favorite tools for focused creations. Hermes is becoming my default for ongoing operation. Those are different jobs. If I need to design a landing page, I can use Claude. If I need an agent that knows my content standards, can work from Telegram, can remember corrections, can run schedule checks, can manage an email workflow, can help create an entire reaction workflow, and can improve every time I use it, I use Hermes.

That is the split that I now use when I think about the two. Claude is amazing when I'm sitting in front of it. Hermes is amazing even when I'm not. And once you feel that difference, it changes how you think about AI tools completely. Because the real switch is not from Claude to Hermes, the real switch is from chat to operating layer. This is the part I think most people are going to miss. The future of AI is not just a better chat window.

A better chat window is useful. A better model is useful, a better coding assistant is useful. But if AI is going to become part of how we actually live and work, it cannot just sit there waiting for prompts. It needs to know you. It needs to improve. It needs to remember, be able to to run workflows. It needs to work where you already are. It needs to become something closer to a companion and workforce than a chatbot.

This is why I switched. Not because Claude is bad, Claude is incredible. I still use it. But Hermes changed what I expect from AI. I do not just want something that can answer a question. I want something that can help operate my business, my life, my systems, everything around me. I do not just want something that can write code. I want something that can learn the way I work and turn that way into automatable workflows that just run without me.

Why am I needed for most of the things that I do? I do not just want something that is smart inside one session. I want something that compounds over time. That is the reason Hermes became harder and harder for me to replace. Every skill it made makes it more useful. Every correction along the way made it more aligned. Every workflow made it just more embedded. Every day it operates inside my actual work, it becomes less like a tool I use and more like a layer I rely on.

So, if you are already using Claude, I'm not telling you to stop. Use Claude for what it is amazing at. But if you feel like AI is still trapped inside a chat box, try Hermes for one real workflow. Not a random prompt, not a some random demo. One workflow that actually happens every week that you spend time on every single week. Maybe it's checking emails. Maybe it's summarizing meetings. Maybe it's reading your meeting notes and sending them turning them into tasks inside your notion.

Maybe it's monitoring your notion for the tasks you assign to yourself or wherever that is. Maybe it can do those tasks for you. Maybe it's managing some recurring business process you keep forgetting to do. Give Hermes one job. Let it learn how you want that job done, then improve it. That is when the switch starts to make sense. Claude made me faster when I was sitting in front of it. Hermes made AI useful when I was not.

And that is why I think chatbots are not enough anymore. If you want to try this yourself, do not start by asking Hermes for random ideas. Start with one workflow you already repeat every week. In the pinned comment of this video, I will leave a simple starter prompt you can paste into Hermes once you set it up. It will help you choose one workflow, define the job, decide what Hermes should remember, and set up the first version without pretending it can magically run your whole life on day one.

The goal is not to replace every AI tool you use. The goal is to feel the difference between chatting with AI and giving AI an actual job. If you enjoyed this video, make sure to leave a like. And if you're new to my channel, then subscribe because I have a ton more content like this coming your way. Oh, and look at this. The algorithm gods just told me you're very likely to enjoy this video. So, click it and maybe I'll see you there.