Where I put AI and code to work - on finance problems and a few personal ones. Some are live, some are in progress. Demos arrive as each one matures.
The site you're looking at now - designed and built from scratch with Next.js, TypeScript, and Claude Code, then deployed to Vercel.
Why it matters: The proof is the product. Rather than claiming I can build with modern tools, I built the thing you're reading on - start to finish.
A website I built as best man to organise a 22-person stag do - schedule, logistics, and key details in one place (password-protected for the group).
Why it matters: Long before the finance automation, this was me reaching for code to solve a real organisational problem that people actually relied on.
View codeAn AI assistant that pulls my running data via Strava's new MCP and uses Claude to spot trends, flag plateaus, and suggest what to train next.
Why it matters: The same measure-analyse-act loop I apply to financial performance, pointed at my own running - proof the analytical toolkit travels beyond the finance function.
A personal knowledge vault in Obsidian wired to Claude - capturing and connecting what I read across AI, Bitcoin, and finance, inspired by Andrej Karpathy's second-brain approach.
Why it matters: Staying genuinely current in fast-moving fields is hard. This turns scattered articles and videos into a structured, searchable base that compounds over time.
AI tooling to streamline invoicing and payroll for a family-run small business, cutting the manual admin a bookkeeper handles each month.
Why it matters: A real-world test of AI on everyday finance admin - starting with family as a proof of concept, with an eye on growing it into a side venture.