01What Minimax M3 + Hermes + Ollama Is and How It Works
Minimax M3 + Hermes + Ollama is a stack of three free tools that turn your computer into an AI worker. MiniMax M3 is the brain that thinks and makes choices. Ollama runs that brain on your machine.
Hermes is the worker that does the actual tasks. Together, they let you point AI at your own to-do list.
In this episode, Julian Goldie explains how these three pieces fit together like Lego. He calls the setup a brain, an outlet, and a worker. MiniMax M3 is the brain.
Ollama is the power outlet you plug the brain into. Hermes is the worker that reads files, sends messages, and uses the brain to figure out each step.
All three pieces are free. The model MiniMax M3 came out on June 1st and is open. You can run it through free tools on your own computer.
That means the same model that worked by itself for almost 12 hours can work for you.
02Main Features of MiniMax M3 and Hermes
MiniMax M3 can hold one million words at once. That's like a filing cabinet big enough for a whole year of your notes. It still knows where everything is.
Older AI tools had tiny drawers and forgot the start of a chat.
M3 also uses a trick called sparse attention. Instead of flipping through every page, it only pulls the pages it needs. That keeps it fast even when you give it a mountain of stuff.
M3 can look at pictures and watch video. It can also click around a computer screen like a person. So it sees, reads, and acts.
Hermes comes with over 70 built-in skills. It can run in the background, not just in one window. You can hook it up to Telegram, Discord, Slack, or WhatsApp.
You can text your AI worker from your phone like it's a teammate.
Hermes can also do things on a timer. You set it up in plain words. For example, tell it: "Every morning at 8, check my community for new questions and send me a short summary." Then you forget about it.
03How to Set Up Minimax M3 + Hermes + Ollama
- Install Ollama on your computer. Ollama is the tool that runs the brain.
- Download MiniMax M3 through Ollama. This makes M3 available on your machine.
- Install Hermes. Hermes is the worker that will use the brain.
- Launch Hermes pointed at M3. Type one short command, and the worker, brain, and outlet snap together.
- If your computer isn't powerful enough, use the free cloud version of M3 instead. Your data isn't kept.
- Test with a simple task, like asking it to summarize a community post.
Julian says the teams behind Ollama and MiniMax set this up on purpose. They wanted it to be easy. And it is.
You don't need to build an app or write code.
You can run a smaller model on your machine for simple jobs. Save M3 for the heavy ones. Hermes lets you switch brains with one little command.
You're in control.
04How to Use the AI Worker for Real Tasks
The episode walks through several examples. One is teaching Hermes once how you write your welcome message for new members. From then on, it drafts that message for every new person without you explaining it again.
Another example: tell it each night, "Look through my community posts and write me three short tutorial ideas based on what people are stuck on." You wake up to ideas.
Because M3 can hold a million words, you can feed it big stuff. Drop in your whole library of past content. Ask what you've covered too much and what you've never touched.
It will read all of it and tell you.
Here's a one-sentence example from the episode: You could drop in every tutorial and every comment from your last 50 videos. Ask, "What question do people keep asking that I've never answered?" It will find it.
Julian also describes a daily routine. You wake up, open Telegram, and there's already a message from your AI worker. It read every new comment overnight, found the three questions people keep asking, and wrote you a note about each one.
Free, on your own machine.
05Results and Tests Shown in the Episode
The episode starts with a test. MiniMax M3 was handed an award-winning research paper. It was told to redo all the experiments in that paper on its own.
No one helped.
It ran by itself for almost 12 hours straight. It wrote the code, ran the tests, and kept going when things broke. It came back with the work done.
On hard coding and step-by-step tests, M3 beats some of the big paid models. It gets close to the very top ones. In a few tests, it even wins.
Julian is honest about limits. The very top paid models are still a step ahead on the hardest jobs. But close to the best in the world for free on your own computer is the headline.
He also notes that running the big brain on your own machine needs a decent computer. If yours is older, the free cloud version saves you. Nobody gets left out.
06Tips, Limits, and Mistakes to Avoid
Julian gives practical warnings. Don't treat everything online like magic. These free open models are amazing, but the top paid models are still ahead on the hardest jobs.
Running the big brain on your own machine needs a decent computer. If yours is older, use the free cloud version of M3. Your data isn't kept.
You can run a smaller model on your machine for simple jobs. Save M3 for the heavy ones. Hermes lets you switch brains with one command.
Don't skip the setup steps. Julian says the only thing standing between you and your own AI worker is knowing the steps. He built a roadmap inside his community to help.
Start small. Give your worker one clear task, like summarizing comments. Then add more.
The tools are free, so the only cost is your time learning the steps.
07Who This Is For and Why It Matters
This is for anyone who runs a business or wants to. Julian says the gap between "I have an idea" and "it's done" is shrinking fast. You don't need a big team or coding skills.
You need a free brain, a free worker, and a few clear instructions in plain words. People who learn to set this up now will look like they have 10 employees.
People who wait will wonder how everyone got so far ahead. The tools are free, so the only thing stopping you is knowing the steps.
Julian built the AI Profit Boardroom to help. It has a 30-day roadmap for this exact stack, live coaching calls, daily tutorials, and a prompt library. He says 2,800 owners are already testing this stack.
He also mentions the AI Success Lab, a free community with 75,000 people sharing what works. The notes from this episode are available there.
