NL · EN · DE · FR · SV · DA · PL · ES · IT · NO · PT

Hermes Agent makes the difference in AI use

If you only use AI through chat, you're leaving a lot of AI potential on the table. Chat AI remembers little, waits for your input and only then gets to work. Hermes Agent from Nous Research is, in my opinion, next level AI use. Why?

First of all, Hermes Agent works as an agent. That means Hermes is capable of performing actions independently. For example, it can look up new information on the internet on its own. Where chat AI is limited to the information it was trained on, Hermes can find up-to-date information by itself.

Secondly, Hermes remembers what it has already done. This gives you the option to switch between assignments or tasks without having to enter a whole new prompt.

Thirdly, Hermes learns from the mistakes it makes. Reflecting happens automatically and works surprisingly well, although you can't count on it always going right. It generally detects that it makes mistakes, but the corrective actions aren't always appropriate. That's where human supervision is necessary.

Hermes Agent uses a Large Language Model (LLM), and even though it seems as if the system really thinks, it is simply put still a statistical system. It calculates the probability of the next word (token). Because it has made a mistake, the probability increases that the solution found also goes in the wrong direction.

That is why it is vitally important that you know what the AI is doing. If the AI knows more about a subject or action than you do, you are actually no longer able to supervise, and the AI will start doing the wrong things. Be very aware of that.

Fourthly, I consider the free choice of different LLM providers a major plus point of Hermes Agent. I can choose between various AI cloud providers (OpenAI, Anthropic, Mistral). I can switch between them without losing my "chats" or sessions. Those stay in Hermes Agent.

If I have a heavy task, I can choose a strong model for it. For simple tasks, I pick a smaller or specialized model.

Hermes Agent is open source and can use local AI. This also makes Hermes suitable for companies that process confidential data or want to deploy Hermes for work suited to an AI agent.

What can you use Hermes Agent for?

Certainly, and I'm making use of that possibility more and more. For example, I let my Hermes Agent perform recurring tasks. Think of renewing security certificates and checking incoming e-mails. The important thing is that the tasks are worked out in great detail. Sounds like a lot of work, although in my experience it's comparable to explaining a task to a new employee.

What I use Hermes Agent for:

  1. Converting articles to MP3 files
  2. Renewing certificates
  3. Spam and phishing filtering and reporting to the relevant internet provider
  4. Checking my texts
  5. Translating texts
  6. Working out ideas
  7. Building prototypes
  8. Deploying software
  9. Version control of source code
  10. Fetching local weather and public transport information

That last one was an experiment, and it worked so well right away that I rely on it. When I go to the office, I first cycle for 20 minutes and then take the train. It's nice to know whether the train is running or delayed.

I use my Hermes Agent every day. It keeps an eye on all kinds of things for me and often gives suggestions that we then work out further.

Hermes Agent runs perfectly 24/7 on an old computer. I personally recommend Linux, because Linux is more stable than Windows for this purpose. You can also use a machine in the cloud where Hermes Agent runs. There are instructions for that on YouTube.

My advice: install Hermes Agent locally. Less security hassle and you get much more out of it there. Especially if you also do home automation (smart plugs and lighting). Hermes could, for example, track your location per room based on your phone and turn off the lighting where nobody is.

← Back to the blog overview