Hermes AI Just Learned to Read Books

Imagine handing your AI agent a book and then 10 minutes later, it knows the whole thing forever. That's exactly what Hermes AI agent can now do. So a free new update just drops it and lets your agent learn from any book, any PDF, any document you give it with one simple command. You can feed it, for example, the best sales book ever written and your agent writes emails using those exact frameworks.

You can feed it your own business documents and your agent understands questions the way you would answer them. So you're not just using an AI anymore. You're building AI that gets smarter every single week. And later in this video I'll show you the first book you should feed it because this one choice makes the biggest difference.

Stick with me until the end and you'll walk away with an AI agent that knows more about your business than most of your team. Let's get into it. Hermes AI agents can now learn any book. So that means with one command you can put in a PDF your agent reads the whole thing, breaks it down and turns it into a skill that you can use forever.

This just dropped yesterday. Technium the co-founder and lead engineer of Hermes agent and news research posted it and his exact words were just forward slash learn point to any PDF or book that you have. Now, let me show you what actually means for your business because on the surface, it sounds small, a new command, but when you understand what's happening under the hood, you'll see why this changes how you build AI agents from now on. Here's the problem this solves and you've probably felt it yourself.

Let's say for example, you got a book, maybe it's a sales book, a marketing book, maybe it's your own company documents, your brand guidelines, your standard operating procedures, and you want your AI to actually know that stuff. What did you do before? Well, you dumped the PDF into the chat and then one of two things either happened, either the AI choked because the book was too big for its context window or it read it once, answered your question and forgot. If in the moment the chat ended, then the next time you have to upload the whole thing again and again and again, every single conversation, you're having to refeed the same book or documents back and that's the old way and it's broken in a way that most people never even noticed.

The new way is what I call the book brain engine. Here's how it works. So instead of dumping a book into AI short term memory, the book brain engine converts a book into a permanent skill. It reads a whole thing once it pulls out the frameworks, the decision rules and mistakes to avoid.

It builds a small index file, one file per chapter, a glossary of every key term, a patterns file and a cheat sheet of quick reference rules. Then your agent carries that brain around forever and when you ask a question, it doesn't reread the book. It looks at the index, opens the exact chapter it needs and answers from the real content. The book stops becoming a file that you upload and it becomes part of how your agent thinks.

Now let me give you the actual numbers because this is where it gets interesting. So the whole thing is built on an open source project called book to skill made by a developer and it's got over 18,000 GitHub stars, which for a tool like this is pretty huge and the projects own testing, measured on real books found it uses 24 to 51 times fewer tokens than dumping the book into context to answer one question. So if you think about what that means in plain English, every question you ask your AI costs tokens. Tokens are expensive and they cost time and if your agent needs 50 times fewer tokens to answer from a book, your agent just got dramatically cheaper and faster at using knowledge.

Same answers, just a fraction of the time and cost. So Hermes have built that straight into Hermes agent. So now if you run forward slash learn point to any book or PDF on your computer, it does the whole thing for you. Now someone actually asked him on X if there's a size limit.

He said the main skill file caps at 100,000 characters. When I say him, I mean technically of the founder of Hermes. Uh, so 10,000, 100,000 characters is about 33,000 tokens, but the system is built to create reference files. So the main skill just points the way to small compact references for each table in the book.

The brain has a table of contents and only opens the page it needs. You might also wonder if it works on EPUB files. So it does. Any book file you have on your machine works with this.

Now, quick pause here. Inside they are profitable. And we've got over 3,800 business owners. A lot of them are running Hermes agent right now as part of their setup.

Some of them had never touched AI before joining. And this week we're covering exactly how to use this forward slash learn command so that you can get the most out of it and how to build an AI agent that actually knows your business inside out. You get daily step-by-step tutorials for coaching calls every week where you can ask questions live about your Hermes setup and the 30 day roadmaps. You're never guessing what to do next.

Link in the comments description, or go to the AIprofitbottom.com to get access. Now you might wonder, okay, why does this matter so much? Well, let me explain the hidden costs that nobody talks about. The Bootskill project calls it the discovery loop tax.

And once you hear it, you can't unhear it. So when an AI agent reads a big PDF the normal way, it doesn't just read it once front to back. It navigates. It pulls up the table of contents.

It jumps to a section. It realizes that the wrong section is opened. Then it backtracks. It rereads and it does all of that again on every single conversation.

So the book never sticks. It's like hiring an assistant. It reads your entire company handbook every morning, forgets it by lunch and reads it again the next day forever. And the book brain engine basically fixes out once.

So at the conversation point with this skill. So after that, every question only costs what the answer costs, the structure is already built. And that's a shift. It's from like renting knowledge to owning it.

Now, actually someone asked like, wasn't it like this before? Is that actually like any change? You can actually see that's not the case. So Hermes skills previously had to stay small and contained maximum like 200 lines, no reference files allowed.

So skills were built for little tasks. Now, when Hermes receives a large document like a book, it's allowed to be expansive with full chapters, glossaries, cheat sheets, deep knowledge, properly organized. And that's a real headline because Hermes agents just went from learning tricks to learning subjects. Now you might say, okay, what do you actually do with this?

Well, let me get practical because this is where it stops being tech news and starts being a business tool. Everyone here is like feed it a book and thinks about reading books, but the tool works on any structured document. The original project spells this out internal documentation, for example, brand and voice guidelines, research papers, specs and standards, anything that you reopen often enough that you wish you memorized it. So picture this, for example, you run an agency and you feed your client onboarding document so that now every agent task follows your exact process without you explaining it each time.

Or for example, let's say you're a coach and you turn your own program materials into a skill. So your agent answers client questions the way you would answer them. Or let's say for example, you're in e-commerce and your brand voice guide becomes a skill so that every product description and email your agent writes actually sounds like your brand. It's just one document, one command with permanent knowledge.

Now some of you might be thinking, this sounds technical. I'm not a coder. This isn't for me. Here's the honest answer.

The command is literally just forward slash learn followed by the file location. That's it. So you don't need to write code. You don't build anything.

The hardest part is knowing which file to point it out. So if you can attach a file as an email, you can do this. And the people getting the most out of tools like this right now aren't developers, they're regular business owners who just started before everyone else. Some people say, you know, AI makes stuff up.

How do I trust it with my most important documents? And that's a fair concern. And this is exactly why this approach matters. So the whole point of the book brain engine is that the agent answers from the actual content and the project is explicit about this.

So when your agent loads a skill, it reads the real chapter file and answers from the real text. The structure exists specifically to stop the guessing. When an agent has no access to the source, it fills gaps with guesses. When it has an index brain built from your document, it looks things up.

The update reduces hallucination on your material. It doesn't add to it. Now some people say, I'll wait until this stuff settles down. It changes too fast and I get it, but honestly, some things are worth waiting on, but look at what waiting costs here.

This feature shipped yesterday. The people who set it up this week will spend the next few months building elaborate skills, their books, their recipes, their niche knowledge, whilst everyone else is still copying and pasting PDFs into chat windows. Knowledge compounds. A skilled library you start building today is an asset that grows every single week.

And the gap between the person with 50 skills and the person with zero doesn't close on its own. It widens. Let me paint the picture of the old way versus a new way, just so you understand the contrast here. Like the old ways you find a great book on sales, you read it three months later.

You remember maybe like 5% of it. Your AI never saw it at all. And that knowledge is gone. The new ways that you find a great book on sales, you run forward slash learn.

10 minutes later, your Hermes agent has a permanent structured brain built from that books. Every followup email, it drafts every objection handling script. It writes every piece of outreach. It can pull from that book's actual frameworks forever, and you can just stack the next book on top of the next.

So the old way knowledge leaked out of your business constantly, the new way knowledge actually accumulates. And here's what makes the Hermes version specifically interesting compared to just using the original repo. Hermes agent isn't just a chat window. It can run on your phone, discord, CLI.

It has scheduled tasks so it can run tasks on a schedule without you being there. It has persistent memory across sessions. It can spawn sub-agents for parallel work. So if you combine those, you've got an agent that runs on a schedule, remembers everything and can permanently learn from any document you give it.

So you can feed it your industry's key reference material, set a scheduled task to draft your weekly newsletter using it. And that's the knowledge worker that studied your exact field working whilst you sleep. And that's the direction all of this is heading. Agents that don't just do tasks, but agents that build up expertise and get better over time.

The same way a good employee does. One more thing worth noting. So the underlying book skill format follows an open agent skills standard. That means the same skill format works across Cloud Code, GitHub, Copilot, CLI and AMP too.

So the skills you build aren't locked into one tool. The knowledge is portable. You're just building an asset and the asset travels with you. If you think about where this goes over the next few months, right now it's one book at a time.

But as someone in the replies already asked Technium like whether skills could cross check multiple books in a directory. So for example, finding contradictions between two books and merging the sources. That's the obvious next step so that you can have whole libraries folded into one unified brain updated as new material lands. The original project already supports folding new documents into existing skills.

So your agent's knowledge stops becoming a snapshot and starts becoming something that grows. So here's what to do with this this week. If you want to move on this first, if you don't already have Hermes agents set up, install it, it's free, it's open source built by Noose Research. Second, pick one document, just one.

The book you always recommend, for example, or your most used internal document run forward slash learn on it. Third, test it so you can ask your agent questions. You know the answers to watch it, pull from the actual content. And fourth, once you actually trust it, one skill a week, you can start skill stacking.

So in three months, you've got a specialist. And if you want the fast path and all of this inside the air profit boarding, this is exactly the kind of update we'll be able to play books around the moment it ships. So we've got daily tutorials showing you how to set up Hermes, how to use forward slash learn to turn your business documents into agent skills and which documents give you the biggest wins. First, you get four weekly coaching calls where members who are already deep into Hermes answer your questions live about your setup, a 30 day roadmap built around getting agents working in your business.

You never stop wondering what's next. And you also get the agent operating system, the mission control dashboard, where you can plug in all your agents. So your Hermes, your Claude, your open claw or work in one place with your shared memory. So the skills Hermes learns from your books, feed into your workflows across your whole stack.

You also get the full zip file for installing it, a video tutorial and daily updates. As we improve new versions, over 3,800 business owners are inside right now. Plenty of them were side with zero AI experience. Plenty of them are running Hermes agent today, link in the comments description, or go to the AIprofitboard.com.

Your agent can now learn from any book you hand it. The only question left is which book you hand it first. I'll see you in the next one.