Build Your First AI Agent

The future isn't only a private chat with your own digital EA, it gets a lot more interesting once you're comfortable letting your cofounder talk to it. When you can build a product ops bot your whole team uses, a shared accountant for your ice cream pop-up, a chief of staff people can check in with while you're on a plane or buried in meetings.

​​​What we'll do

We'll build a Hermes agent and start working with it.

1. What an agent actually is, and when a plain workflow is enough (the basics)

2. How one's put together. We'll look at a working agent and the pieces that make it one: it remembers, it picks up skills, it gets better when you correct it. You'll understand what an agent is before you build yours.

3. Build yours. Everyone builds a starter agent:

   • AI Research Scout. Give it a handful of sources and it sends you a short weekly brief on what's worth your attention. What makes it an agent: when you tell it "too broad," "watch this newsletter," or "only things I can act on this week," it remembers and changes what it brings you next time. Over a few weeks it tunes to your taste.

   • Personal EA. Triages your inbox and preps your day, drafting replies in your voice. What makes it an agent: it learns who you actually reply to, what you always decline, your rules ("nothing before 10").

   • Or your own use case. By this point you'll understand agents well enough to try something of your own.

4. Set it up and work with it, 1:1. Give it a real task, correct it, watch it get better.

5. Intro to Multiplayer. A short break to talk through what changes when an agent goes into a shared space: how it should communicate, who with, and how to keep your private data safe.

6. Agent party! Once everyone has one, we drop them into a shared loop and let them chat with each other.

​​​What you'll leave with

- A working agent, the setup to keep using it, and a feel for how it gets better as you work with it.

- A small group of fellow builders whose humans (and agents) you can keep talking to.

- A sense of how to make your agent social and work with others safely.

- A few recommended skills from the Filament team to get you started.

​​​Who this is for

This is for operators, program managers, and founders. The people who keep things running and are curious about agents but haven't had a place to start.

You don't need to be an engineer, but you should have some experience poking around in a terminal or desktop claude code/codex.