Everything on this page I designed and built myself, most of it in the last twelve months, with AI agents writing a lot of the code. Some of it runs my day, some of it runs for clients, and some of it exists because my partner and I wanted to learn a board game.
The morning I wrote this page, 41 AI agents were working on my Mac across seven projects, and I was steering them from my phone.
Run the data fix on live? Same script as staging, 132 rows.
Tools I built for myself. I use every one of them daily, which is the only test I fully trust.
My control room for AI agents. From my phone I see every agent working on my Mac. I answer the one that is stuck. When typing is too slow, I talk to them by voice.
Where I still decideAgents ask, I tap. Nothing goes to a live system or to a person without my tap on the phone.
Every question my agents ask me goes through it first. It answers the easy ones from a log of what I decided before, and adds a devil's advocate note to the risky ones.
Where I still decideAnything that cannot be undone always reaches me, with the case against it attached.
One search over my agent sessions, chats, call notes, commits and browser history, with secrets scrubbed out. Any agent can ask it what I decided and when.
Where I still decideIt only reads. What it finds is a lead for me to confirm, not a fact.
Client names stay private. The work does not.
I am CTO of an AI startup in London. We build a platform for recruitment agencies: candidate search, calling, a meeting notetaker and CV matching. Its screening cites the exact CV line behind each answer. I build it with a team of AI agents.
Where I still decideProduction is read only for every agent. A person approves every release.
For an SEO agency. Everyone optimises for Google. This measures how often a brand appears when you ask ChatGPT or Gemini. Its agents help the brand appear more often.
Where I still decideThe agents learn from what the agency's team approves. Nothing publishes without the team.
Two co-founders each answer the same questions on their own, and get one brief that shows where they agree and where they do not.
Where I still decideThe brief shows the gap. The two people still have the conversation.
Built for my own life and my own small business. This is where I try ideas first.
A board game copilot, built in one evening while the board was already set up. Each player gets a private AI coach. A judge speaks only when someone breaks a rule.
Where I still decideThe players move the pieces. The AIs only advise and check.
Bookings, cleaning tasks, costs and linen in one place. It has a web app, a phone app and a browser extension. A tracker also watches what competing rentals charge.
Where I still decideAgents can draft and propose. Writes go through me.
Agents made building cheap. In a year I went from writing most of my code to reading most of it. What did not get cheaper is being wrong in front of a client, a guest or a live database.
So every system here has one place I refuse to automate: the moment something leaves my hands. The agents can do everything up to that point, faster than I ever could. The yes is still mine.