My first bot

Before I had a bot, I had plenty of failed attempts. I tinkered, got things wrong and started again more times than I care to count. That’s how learning by doing works: you move forward by making mistakes.

Eventually I came across Hermes Agent, a tool for creating your own artificial intelligence agents. I downloaded it, installed it and created my first bot: Alonides 1.

First stop: Signal

I put it on Signal first. I chose it for voice: I wanted to talk to my assistant, not just write to it, and it seemed like the way to get there.

It didn’t go well. It was difficult to set up and, when it finally worked, it didn’t work the way I wanted. The first bot existed, but it still didn’t look much like the idea in my head.

First lesson: the perfect tool on paper has to work in everyday life too.

Telegram: where it stays

So I went back to Telegram. And that’s where it stayed, because it brings together almost everything I needed:

• It’s very easy to use. I talk to my bots just as I would to any other contact.
• It gives you plenty of control over bots.
• It works on my phone and my computer. The conversation comes with me.
• It accepts multimedia: photos, audio and documents.
• I can adjust how I use my bots. That control matters to me.
• It allows group chats with several bots. I give instructions to them at the same time and coordinate teams across different projects.

I wasn’t using that last feature at the start. But it would become one of the key pieces of La Machina.

The workshop was on my computer

Although I talked to the bot through Telegram, I did the real work on my computer. That was where I started doing something that changed everything: cloning bots.

I copied a bot’s configuration to prepare another one. From there, I changed the copy, tested it and compared it with the original. I created bots, duplicated them and gave them different abilities. Hermes calls them skills: instructions and procedures for doing a particular kind of task well.

I also tried different ways of organising the work. It wasn’t enough for the bot to know how to do things. It had to handle several complex tasks at once while still being there for me: to keep listening to what I was saying while it worked.

That balance was the difficult part. A bot that does a lot but doesn’t listen is no use. Neither is one that listens but does nothing.

Total freedom… and all the work

I think Hermes is a fantastic tool for its flexibility. It lets me configure agents, connect them to other applications and build workflows for complex tasks.

But that flexibility is also, perhaps, its problem. You have to configure everything yourself. Nobody sets it all up for you.

The good thing is that you don’t do it alone. You can configure it with your own AI tools and learn along the way. I asked AI for help understanding each piece, tried it out and, when something went wrong, asked again. Every mistake taught me something new about how a bot works on the inside.

What I took away from this stage

• Choose the platform for what works in everyday life, not for what it promises.
• Telegram as the front door and the computer as the workshop.
• Cloning lets me test changes and keep the version that works.
• A good bot does lots of things and keeps listening to you.
• With Hermes I have learned by configuring, testing and correcting.

Alonides 1 was the beginning. By copying it, changing it and keeping what worked, it started to evolve. The next episode tells how the machine began to grow.

La Machina in this episode
Milestone: my first bot, Alonides 1, in Hermes Agent.
What failed: Signal, difficult to configure for what I wanted.
What worked: Telegram for talking and the computer for cloning and training.
What’s next: the machine starts to grow.