DabbleAgents in Service of Becoming
AI agents that support human becoming instead of capturing attention — parks where the internet built casinos.
The need
Most consumer AI is optimized for retention — time on screen, not time well spent. Shawn started from a different premise: that agents could be built to serve human becoming — helping people notice, choose, connect, and act — rather than capture their attention. Parks, not casinos.
The approach
The work is a suite of built prototypes grounded in three commitments — transparency, choice, and contestability — and a build-to-think method: make the agent real, live with it, see what it does to behavior.
- Agent design oriented around human flourishing
- Calm-technology principles (legible, optional, contestable)
- Rapid, lived-in prototypes rather than mockups
The artifact
Five working prototypes: Shawn’s Café, Breadcrumbs (an e-ink provocation window), Roy_bot (a group-memory agent for a friend group), Snudge (a screen-nudge coach), and Pumbit (with Diane). They’re small on purpose — each one a test of whether an agent can add to a life rather than extract from it.
Why it matters
On the days Roy_bot was present, the group saw roughly a 2× increase in human-to-human interaction — early evidence that an agent can deepen real connection instead of replacing it.
“We can build agents that help people become more themselves — not just keep them scrolling.”
From the thesis
