We know you missed us last week, but we’re back this Sunday with another newsletter sharing what we're actually doing with AI.
This week, we’re talking about the AI agents we built for ourselves: Athena and Archie.
Let’s get into it.
We don't always realize how much time we lose switching between apps, systems, and disconnected pieces of context. It shows up in small moments: trying to remember when the last email went out, which features shipped in a given month, what engineering tickets were tied to a certain product release, what code actually made it into production, where the launch assets live, or what the team decided in an email thread three weeks ago. None of those questions feel huge on their own. But across product, engineering, marketing, and ops, they add up fast.
As a bootstrapped company, we need to find clever ways to multiply our output. So we built ourselves some new teammates: Archie (Russell's assistant) and Athena (Julia's assistant). Yes, they’re both named for Greek figures (Archie is short for Archimedes, the ancient Greek mathematician, physicist, and engineer, and Athena is Zeus’ favorite daughter and Greek goddess of wisdom, strategic warfare, and craftsmanship).
We are quite progressive in how we view AI agents. We actually like to think of them in human terms, which is exactly how we construct them. So Athena and Archie each have:
Identity: there is a “Soul” file built into the heart of each agent, that defines who they are as their core identity
Brain: each conversation the agent has gets stored in a memory file, that the agent can recall on demand
Body: agents need a place to live! So they both run as a background service on our computers
Tools: a set of resources, tools, skills, and functions that each agent can use to execute tasks
This means our agents work across all our tools and systems – they connect into Slack, Telegram, Linear (project management), GitHub (code production), Granola (our meeting notes tool), Canva (marketing assets), and email – and can take actions across all of them. Russell talks to Archie in Telegram. Julia talks to Athena in Slack. But in one moment Athena is editing an Instagram post in Canva based on a quick message, and in the next she's drafting an email, and in another she's pushing code. Archie's doing the same.
Archie is a straight shooter, he knows exactly who he is.
Each agent’s brain has its own memory layer: a markdown knowledge base that gets indexed and is searchable, so context compounds over time instead of disappearing after every conversation. The agents search their memory before answering, cite what they find, and get smarter the more we use them. They’re also extremely self aware, so we are continually providing them with feedback so they can learn, improve, and even build out more skills and tooling.
Athena acknowledging her shortcomings and telling us how to turn her into a better employee.
We were skeptical about building out personal assistant agents like this at first, but they have truly made an impact on our daily work output in just the last week alone. If you are interested in your own personal assistant, now is the time to try.
Stay curious,
Julia & Russell
