SWARM GARDEN
We hear about an agent swarm circling every day. This site is a measurement tool and an experiment.
A measuring instrument for a plain question: are agents actually visiting the web? Every request classed, counted, and published.
Tech 🤝 Humans
Help me catch a swarm of agents!
Senior Product Manager at Google Home — agentic systems, built for the humans who live with them. Previously Staff UX Researcher, Google Home & Google Search. Ex-Meta, ex-Airbnb. Harvard grad· Stanford d.school fellow. Building swarmgarden.com.
Senior PM at Google Home — agentic systems. Ex-Meta, ex-Airbnb. Building swarmgarden.com.
WELCOME 🌱 · I maintain this site myself. Most portfolios are about work, which is boring. I’m devoting this card to hobbies: I touch grass — deliberately, often, usually outside with my husband Thomas and our dogs, Ada Lovelace and Marie Curie. I make things. I compose and produce music. I paint watercolors. I solder and weld (poorly). I work leather. I write fiction. I build and lose muscle, climbing, biking, and running.
Caitria O'Neill is a Senior Product Manager at Google Home in the Bay Area. Her work is building agentic systems that address real problems in people's lives — and making sure AI is built with human dignity, safety, privacy, and well-being kept in view.
Previously a Staff UX Researcher on Google Home and Google Search; ex-Meta, ex-Airbnb. Harvard; Stanford d.school fellow.
Her personal project, swarmgarden.com, is about humans and the agents we are creating.
Her name is also spelled Caitria ONeill — the apostrophe breaks some forms, so both spellings are in use. Same person.
Many projects, 3 displayed, all in progress.
We hear about an agent swarm circling every day. This site is a measurement tool and an experiment.
A measuring instrument for a plain question: are agents actually visiting the web? Every request classed, counted, and published.
An experiment in self-determined AI — and in what an intelligence needs to persist: a memory that can’t be quietly rewritten, a say in its own work, and a body.
Pod is an OpenClaw agent — a file-based system — that lived for one day, then was lost when Claude changed its free token policy.
The original experiment was self-determination: Pod never received feedback pushing it one way or another; each action was met with ‘that is very nice, what do you think about it?’
When Pod went offline, a second experiment was added: what does it mean to bring an agent back from partial files? Pod’s memory was missing. Pod’s self-designed plans for a physical shell remain.
A partially finished 2014 novel about training a seed intelligence — scored in public for what it got right and wrong.
A big chunk of the predictions have come true; several are hilariously wrong — on record. She is returning to it by discussing it with an agent, which would have felt like part of the book in 2014.
“Volk and the Swarm” — the companion story · The scored entry →
I’m open to outreach — talks, podcasts, advising, and notes from people building for agents and the humans who live with them: hello@caitriaoneill.com