Meet us at the next idea.

AI agents are great at figuring things out on the fly. But when it comes to shipping software, reviewing code, or responding to production incidents, "figuring it out as you go" isn't always what you want.

What if you could give your agents the freedom to reason while keeping them on a well-defined path?

Meet Outpost.

Outpost brings deterministic workflows to non-deterministic AI agents. Build teams of agents, called Sherpas, and dispatch them on Expeditions along predefined Routes. Each route defines the steps, permissions, and human approval points needed to get the job done. Your agents can reason and solve problems along the way, but the workflow remains structured and predictable.

Whether it's your custom software development lifecycle, incident response, support tickets, or shipping software, Outpost lets you define how work gets done.

And because Outpost is built on Atomic, every change carries a verifiable history of who did what, including the agents involved.

Join us for a technical walkthrough of Outpost, where we'll explore the architecture, the ideas behind agentic workflows, and demonstrate a team of Sherpas working together to build, review, deploy, and respond to a production incident.

No hands-on setup required. Just come see what happens when AI agents have a map to follow.

Git has spent two decades teaching us to think about software change through ancestry. Commits point backward, histories form, and the repository tells us how one state descended from another.

Atomic starts from a different premise. Its causal graph represents change through patches and dependencies, allowing us to reason about what a change needs, what it affects, and how it can be composed independently of a single line of ancestry.

On paper, those models look like they belong on opposite sides of the source-control universe.

But peanut butter and jelly looked pretty different too.

In this workshop, we'll explore what happens when ancestral and causal models of change meet. We'll look at where their assumptions differ, where they unexpectedly complement each other, and what becomes possible when you stop treating them as mutually exclusive.

And then we'll show you what we've been building.