Things I'm building, exploring, and learning.
The agent that nags me to post every day now lives on my phone. Built an iOS version of the accountability coach agent that runs my daily check-ins, scorecards, and weekly tally. Nobody asked me to build it — just wanted to see what changes when an agent moves from the terminal on your laptop to a phone in your pocket. Still figuring out what becomes possible when something I built for myself can sit in someone else's hands.
Built an n8n + Firecrawl pipeline to pull data from 52,504 organisation websites for a non-profit that needed it. The alternatives were weeks of soul-destroying clicking, or emailing every organisation and hoping 30% replied. The automation ran in under seven days at a fraction of the cost. The bigger win was the work that would have otherwise been looked at, sighed at, and quietly shelved.
Spun up an agentic system on a dedicated Mac Mini and within four days had nine agents running around the clock, each with their own skills, memories, and access to the knowledge base. A researcher hunting speaking opportunities, a PR agent drafting the proposals, a comms agent writing in my voice, a librarian processing meeting transcripts, an analyst monitoring system health, plus accountability, community, and website agents. They're already adding value in the AI Impact Hub community and surfacing speaking opportunities I wouldn't have found on my own. A follow-on from the AI Impact Hub agent team, except this one's for the whole of my work, the operating system behind the personal brand.
Started sharing the prompts, skills, and tools that help me work with AI every day. Prompts you can copy in one click, skills you can download — all free. Building it out as I go.
ViewUsing Claude Code to build interactive, scroll-driven websites for nonprofits — impact reports that feel like stories and donation pages that respond to the reader. Exploring how far AI-assisted development can push the craft of nonprofit web experiences.
Building out a team of AI agents to support AI Impact Hub — one that researches emerging tools and trends, one that helps onboard new community members, and one that drafts community posts. The experiment is whether agents can carry the operational weight so I can focus on teaching.
Exported every conversation I've ever had with ChatGPT and Claude. Three years. 13,227 messages. Asked AI to perform an "Intellectual Autopsy" on me. It found that I said "please" 3,600 times to a machine that doesn't care, that 1% of my conversations were personal, and that I never once asked "should I stop doing this?" Your AI history is a mirror — most people just never look.
Built a voice agent that calls me every afternoon with a different persona and set of questions. Some days it plays devil's advocate on a workshop idea. Other days it interviews me like a journalist. The goal is to surface thinking I wouldn't reach on my own — and it's working.
Built an Airtable database and Make automations to manage 23 rescued dogs, cats, and sheep. Automated text reminders for worming schedules, birthday tracking, and care logs. The kind of system I'd build for a client — except this one's for the animals living in my house.