I tend to understand a new technology by giving it a real problem. These are a few cases where I used AI to make sense of decisions I was actually facing—not demos built in search of a use.
The original tools contain personal information and keep changing as the decisions change. So this is a portfolio of what became possible, how I approached it, and what I learnt along the way. The dates are also a useful record of how my use of AI is changing.
August 2026
A car’s price is obvious. Its cost is not. I built an interactive model for comparing used versus new, petrol versus EV, finance offers, running costs, resale and the opportunity cost of taking money out of a home-loan offset account.
July 2026
I knew when I liked a sofa or a rug in a shop. I had almost no idea whether they would like each other—or my house. This became a way to test colours, proportions and complete rooms before buying anything.
May 2026
I had to compare property types, suburbs, contracts, valuations, inspections, personal taste and finances—often all at once. I built one working system to hold the research and still help me make a decision.
October 2025
A rapid Airbnb-style listing generated with AI prompts, then tuned with real photos, pricing, and context to market our Abbotsford apartment sublease.
September 2025
How a coding assistant and an LLM help me build and maintain a minimal Jekyll site: small, auditable changes; simple CSS; and GitHub Pages for deployment.
September 2025
An AI‑generated autobiography from my tweet export (5,384 tweets, 2011–2025). A creative reconstruction exploring how narrative and “voice” emerge from a long public feed.
September 2025
Field notes from a data‑labeling “rubrics bootcamp”: where labeling fits in the AI value chain, limits of binary rubrics, and small evaluation experiments to try.