# ABOUT
Twenty years in software architecture, which turns out to be a peculiar form of product design: deciding not only how systems should be built, but what they should become in the first place. A surprising amount of the work has involved removing things other people added with great confidence.
Banking, insurance, energy, media, public sector, e-commerce: the acronyms change. So do the lanyards. The underlying problem rarely does.
Lately, my work involves LLMs, agents, and probabilistic systems. Anything described as “intelligent” receives particular scrutiny. I prefer to reserve judgment until it has survived contact with production.
At Macropoiesis.com, I put that suspicion to work. My research spans quantitative methods, probabilistic valuation, self-learning forecasting systems, agent architectures, and market geometry. Occasionally, a mathematical curiosity acquires an API. At that point, it must answer to evidence.
The principle is simple: a PDF is not a conclusion. Models become implementations. Hypotheses face data. Methods become systems people can actually use. It is there, with nowhere left to hide behind notation, that the interesting failures begin.
I write about the research, the models, and those failures at blog.macropoiesis.com. Failed hypotheses are often more educational than successful ones. They are considerably harder to market.
You can also find me at linkedin.com/in/mrussmann-software, usually returning to a lesson twenty years of architecture has made difficult to ignore.
Most architecture problems announce themselves quite innocently:
“It’s basically just…”