# The AI Popsicle Index: a human-first way to measure AI

There's a quiet little metric we love. The **Popsicle Index**, coined by former U.S. Assistant Secretary of Housing Catherine Austin Fitts, asks one disarmingly simple question about a community: *can a child safely walk to a local store, buy a popsicle, and come home alone?*

It's brilliant because it measures how a place actually **feels to live in** — trust, safety, connection — instead of a cold number like GDP. A neighborhood can have a booming economy and a terrible Popsicle Index.

AI has the same problem, and nobody's talking about it.

We measure AI by **output**: tokens generated, lines shipped, tasks automated, how "busy" the agent looks. None of those numbers ask the only question that matters: **did the human end up better off?**

So here's our take — an **AI Popsicle Index**, tuned for people building with agents. Five questions, asked honestly:

1. **Is your idea actually shipping** — or stuck in a pile of half-finished pieces?
2. **Are you in control** — building *with* the AI, with no rogue detours?
3. **Is your time respected** — real progress, not re-explaining yourself every session?
4. **Is your compute well spent** — value created, not just tokens burned?
5. **Would you be proud to ship** what you have so far?

Score it. If those are high, it almost doesn't matter how many tokens you spent — you're winning. If they're low, no amount of output volume will save you.

This isn't anti-AI. It's a refusal to confuse *motion* with *progress*. It's why our tools compute this score locally, right alongside your project — the one number we think is worth watching.

Not "how much did we generate?" But "did the human end up better off?"

*The Popsicle Index is Catherine Austin Fitts' concept; the AI adaptation here is our own.*