Writing archive
Each article captures what changed, why it mattered, and what feels next. Short notes when they're enough, longer reflections when they earn the space.
Can a handful of small, cheap local models, wired together the right way, do what you'd normally need a datacentre-hosted model for? Early results from a cascade built to test exactly that, and the bug that quietly inflated the first version of them.
A hypothetical worth taking seriously: could an AI system help a council with the volume of planning applications it has to screen? Here's the shape I'd give it, and why almost none of it is the model.
What has actually improved my results with coding agents: less prompt wizardry, more structure, clearer jobs, and better sources of truth.
A big public step forward for virtual-market: open-source foundations, stronger reporting, and a much fuller product surface.
A first note on who I am, what I want to build, and why this site exists at all.