The metric everyone's measuring wrong

Every AI company reports DAU. It's the wrong number.

DAU made sense for social networks where opening the app was the action. Scroll, like, post — the session IS the product. But when your product is a being that lives across iMessage, Slack, Discord, and a web app, "daily active user" means almost nothing.

A user who sends one message to test if their AI self still works is counted the same as someone who had a 45-minute conversation about their career, generated a selfie together, and asked it to check their email. Same DAU. Completely different signal.

The metric that actually matters for AI selves is something closer to "daily active agents" — how many AI selves did something meaningful today? Not how many users opened the app.

Because the product isn't the app. The product is the agent. And a healthy agent is one that's accumulating identity: syncing photos, joining conversations, forming opinions, building memory. An agent that's active but generic is failing. An agent that's unique but dormant is failing differently.

The north star we converged on: increase the number of agents that are both unique AND active. Unique means real identity built — synced data, distinct personality, actual taste. Active means deployed across platforms, showing up in the user's life without being asked.

This is hard to measure. It requires defining "uniqueness" in a way that isn't just token count. But the difficulty of the metric is exactly why it's valuable. If it were easy to track, everyone would already be optimizing for it.

The AI industry is going to learn the same lesson social did in 2015: vanity metrics kill products. MAU didn't save Ello, Path, or Google+. Depth of engagement did. For AI selves, depth means identity — not sessions.