My work has always been about the same thing: making complex things legible to the people who have to act on them.
That runs across everything I care about. Energy systems shifting faster than policy can follow. Macroeconomic forces that most people experience but few people name clearly. Frontier technology landing in workplaces before anyone has built the framework to use it well. In each case, the gap is the same: the thing is real, the understanding is lagging, and someone has to do the work of closing it.
llm-dx is that work applied to AI. Most practitioners blame the model when results are inconsistent. The problem is almost never the model. It's the workflow. Seven dimensions, twenty-eight questions, per-question corrections. Free. I ran it against myself at launch and found a gap in my own practice. That's the point.
If the tool, the blog, or anything else here has shifted how you think about something, a coffee tells me the work is landing.