Murmur is a long bet on one question: can a system organize knowledge by itself, the way living things do? This page is who's asking, and why from such an odd angle.
I'm David Smit, Dutch, based in Berlin, twenty-plus years in games. Creative and art director, game designer, and cofounder and Chief Product Officer of a VC-backed studio: I raised the money, built the teams, and took projects from a blank page to launch.
Shipping games teaches you to make hard calls fast, on incomplete information, across art, code, design, story, and economy at once. Excellent training for research, mostly because nobody warns you it isn't. Twenty years of it also teaches exactly one research skill: noticing, quickly, that you're wrong.
I'm not a career machine-learning researcher. The career machine-learning researchers will have noticed. It cuts both ways, so I use both edges: I look in corners specialists walk straight past, and I stay honest about what I don't know yet.
Most machine intelligence is trained once, then frozen. I'm betting on the opposite: systems that never stop organizing themselves. Nothing alive gets a training phase and a deployment phase. I've checked.
The best clues live mostly outside AI, so I read sideways. Ant colonies dividing up work with nobody in charge. An embryo building a body from one cell. A physical system settling into its own pattern. When something over there rhymes with the problem in front of me, I build the smallest version that could work and see if it survives. It usually doesn't.
four habits (the “we” is me and the machines)
Stuck? Read another field. Colony behavior, developmental biology, physics. The borrowed idea usually fits better than the obvious one. Ants have been doing distributed systems longer than we have.
Every idea ships with the experiment that would kill it. Most die in the first week. The good ones take longer.
Push a few hundred runs, or a few million items, through the question and watch the pattern. One tidy example mostly proves you found one.
Most of a system decides nothing. Find the one or two places where it lives or dies, and spend the boldness there.
The second one is the load-bearing habit. More on that discipline →
See where the bets have landed.
Reach me at info@murmurproject.eu
murmur · the lab notebook · updated when something breaks or works. Usually the former.