Normalisation is the unglamorous but essential work of turning messy human questions into clean, predictable input that an LLM can…
A side‑project turned laboratory, this chapter dives into how I stopped copy‑pasting my way through AI work and built a…
Part II dives into the messy reality of turning human questions into something an AI can actually understand. It’s where…
This chapter kicks off the real build: turning messy family data into something an AI can reason about without inventing…
Genealogy data is elegant; GEDCOM is… whatever the opposite of elegant is. This chapter explores why your family tree deserves…
This project began with one simple genealogy question I couldn’t answer and the realisation that our tools store data beautifully…
Aptly named after https://en.wikipedia.org/wiki/Lemmings_(video_game). Why the Lemmings reference? An ever-increasing number of companies have adopted chatbots, following the fad like…
In this post I cover training AI to throw a basketball into the hoop from any angle / distance.
How do I reduce the time taken to learn? Which is better ADAM or not, what learning rate optimisation wins?