
Teaching children vibe coding starts with one question: what problem do you have?
Today is day 2149 of my daily barefoot-style running streak. I have covered 16,505.98km so far, which leaves 23,569.02km of the 40,075km lap of the world. Every run gives me time to think. Today my thinking went to my three children.
For about a month I have worked deeply with AI, building software through vibe coding with Claude Code. I am not at the front of adoption. I am ahead of the middle. Most people I speak to still use AI as a chat box, where you ask a question and get an answer. I now use AI to build tools which complete tasks for me. The results have surprised me.
My children are 15, 13 and 11. They are growing up in a world different from the one I grew up in. In my experience, schools, the economy and the country all adapt more slowly than technology changes. A head start in understanding AI matters for what they go on to do.
I have not started teaching them yet. I want to be honest about this. Today's episode is about a plan, not a result.
My plan is simple. I ask each child what problems they have in their lives. Many problems have no software answer. Some do.
My 15-year-old son told me he struggles with revision. I saw an opening straight away. Revision apps already exist, so my aim is not to compete with them. I want him to see what is possible when you build your own tool for your own problem, then change the tool until it does what you want.
I have pushed my children before. In some areas of their lives I pushed too hard. The result was lost interest, or a decision to avoid what I wanted for them. Pressure creates resistance. Push harder and the resistance grows, until anything I suggest gets refused.
So I told my son this: I will help you build something for revision. I will not push you. You drive. Your desire to fix the problem has to come first, and I will sit beside you while we do the work together.
I will offer the same to my 13-year-old daughter and my 11-year-old son, when each has a problem of their own.
I will let them do the prompting. They will learn to structure a prompt clearly, test the result, spot what is wrong and go again. Every piece of software goes through this cycle before anyone finds a use for the finished product. A child who learns the cycle at 15 learns more than coding. They learn to define a problem, try a solution, check the result honestly and improve.
This goes past chatting with an AI. Building a tool which does a task for you is a different skill. Agentic AI, where software acts on your behalf, is where this leads, and vibe coding is one of the most advanced ways in.
Two conditions apply. The child needs a problem and enough desire to fix. And AI needs to be able to solve the problem. If either is missing, we build nothing, and I accept this rather than force an idea on a child who does not want it.
Consistency taught me this. Day 2149 did not come from one huge effort. Day 2149 came from one run each day, kept small enough to repeat. I want my children to meet AI in the same way: one small problem, one session, one improvement. No pressure, no deadline from me.
This vlog is part of my fundraiser for children's causes. I am attempting the ultimate ultra marathon of running and vlogging on consecutive days, barefoot-style, until I have covered 40,075km around the world. My target is to raise £1M. With 23,569.02km still to run, the work is long and the days are ordinary. Children sit at the centre of why I run, so thinking about how to prepare my own children for the future fits the same mission.
If you have a child in or near secondary school, ask them what problems they face. Listen to the answer before you suggest any tool. Then tell me in the comments what they said. If you want to follow the streak and the fundraiser, subscribe, share and watch tomorrow's episode. The more people I reach, the more I raise for children.
Watch the full episode: https://youtu.be/ZWPD9hNtVHY





