01What Is the Learn Anything Engine and How It Works

The Learn Anything Engine is a new feature in Hermes Agent that lets the agent learn from any source you give it. You feed it a blog post, a PDF, a folder of code, or even notes, and Hermes distills that information into a reusable skill that it keeps forever.

In this episode, Julian Goldie showed how the engine works inside the Hermes chat. You type /learn, paste a source like a URL, and Hermes reads it. Then it writes a skill file, saves it, and you can use that skill in any future session.

The agent no longer forgets how to do things after you close a session.

02Main Features of the Learn Anything Engine

The Learn Anything Engine has several features that make it powerful and easy to use. First, it can learn from anything: a document, a folder, a job you just did, or even pasted notes. You are not limited to one type of source.

  • One command learning: Just type /learn in the chat and point to your source.
  • Automatic skill creation: Hermes reads the source and writes a clean skill markdown file itself.
  • Skills are kept forever: The agent saves the skill and loads it only when needed.
  • Works everywhere: You can use /learn in the dashboard, command line, or with Hermes Jarvis voice agent.
  • Skills are shareable and editable: You can edit the markdown file or share it with others.

No separate learning engine or extra setup is needed. The skill is stored in a file called a skill MD file, which includes the name, description, version, procedure, and pitfalls.

03How to Set Up the Learn Anything Engine in 5 Minutes

Setting up the Learn Anything Engine is fast. You need the latest version of Hermes Agent. Then you follow these steps:

  1. Run Hermes update to get the latest version that includes the /learn command.
  2. Open a chat with your agent in the dashboard, terminal, or through Hermes Jarvis voice.
  3. Type /learn followed by your source. For example: /learn https://example.com/guide or /learn folder: workflows.
  4. Hermes reads the source and offers a skill markdown file for you to approve. If you have the approval gate on, you say yes to save it.
  5. After approval, the skill is saved. Next time you need it, just ask the agent for the task by name (e.g., /skill agent-operating-system).

Julian noted that the whole process takes about five minutes once. After that, you never have to re-explain the workflow again.

04How to Use the Learn Anything Engine: Real Examples

Julian showed a real example during the episode. He used the /learn command with a guide from his own website about the agent operating system. He typed /learn and pasted the URL.

Hermes read the guide, learned the content, and saved it as a skill named "agent-operating-system".

To use that skill later, you type /skill agent-operating-system. The agent automatically loads the skill and knows how to answer questions or perform tasks based on that guide. Julian also showed that Hermes can detect triggers: if you mention terms like "agent OS" or "mission control", the skill loads on its own.

You can also use /learn with a folder of code, notes from a previous job, or any document. The skill file is written by Hermes, not by you. Julian did not write a single line of the skill file; Hermes created it.

05Results and Comparisons Shown in the Episode

In the episode, Julian tested the engine with a guide from his website. He expected it might not work, but it worked smoothly on the first try. The skill was saved instantly with the message "Learned it and saved it for future sessions."

Julian compared the Learn Anything Engine to bolt-on tools like MCP services. He explained that those tools let the agent call external functions, but /learn is different because it captures your workflows and steps. You cannot download those from anywhere else.

Only your agent can learn them from watching you.

The engine also works with Hermes Jarvis, the voice-activated agent. Julian showed that you can use the same /learn command with voice, making it even faster. He noted that a free local model completed the learning in about 25 seconds.

06Tips, Limits, and Mistakes to Avoid

Julian shared several tips and warnings based on his testing. First, always update Hermes to the latest version before using /learn; the command won't work on older versions. Second, if you have the approval gate turned on, the agent will ask you to approve the skill before saving.

Julian had it off, so his skill saved immediately. If you keep the gate on, you can review the skill file before it is stored.

  • Use clear sources: If your source is vague, the skill might be less accurate. Point it to a specific document or folder.
  • Stronger models for hard tasks: For tricky jobs, you can route /learn to a stronger model.
  • Bad skills are fixable: You see the file before it saves. If it's wrong, edit one line or run /learn again with a clear source. Nothing is locked in permanently.

Julian also addressed common mental blocks. Some people think teaching AI requires coding, but it does not. You just say one word and describe the thing.

Others worry their agent forgets, but that is exactly what this engine fixes. The skill is kept forever.

07Who the Learn Anything Engine Is For and Why It Matters

The Learn Anything Engine is for anyone who uses Hermes Agent and repeats the same workflows every week. If you spend time re-explaining steps to your agent each session, this feature saves you time. Julian said that the old way meant you explained a workflow, the agent did it, then forgot everything the moment the session ended.

Now you teach it once, and it keeps the recipe.

This matters because the agent builds muscle memory. Every skill you teach makes your agent more capable over time. Waiting for perfect tools means you fall behind.

Julian recommended starting immediately because the gap grows between people who teach their agent today and those who wait.

The engine also makes skills shareable. You can publish a skill file and let others install it with one command. Julian shared his own skill files with his community.

This turns your best workflows into reusable tools that anyone can benefit from.