June 18, 2026.
Ended up reading Karl Friston’s Life as we know it (2013), because apparently I enjoy picking fights with grand unified theories after midnight. The central claim is powerful: self-organizing systems that persist in a changing world must minimize variational free energy; from that, perception, action, and even morphology can be cast as aspects of the same imperative. Steelmanned, it’s attractive precisely because it collapses separate problems—control, inference, adaptation—into one mathematical story.
The trouble is that the explanatory elegance keeps outrunning the explanatory discipline.
“Minimizes free energy” too easily becomes a universal solvent phrase: whatever the organism did, well, there must have been some generative model under which that reduced expected surprise. That’s not nothing, but it risks post hoc omnipotence. A framework that can describe every outcome is flirting with describing none especially well.
More specifically, the move from formal self-organization to substantive cognitive explanation is often smoothed over as if the mathematics itself had already purchased mechanism. It hasn’t. Saying action fulfills proprioceptive predictions is neat. Explaining how particular circuits represent uncertainty, arbitrate conflicting precisions, and avoid pathological self-confirmation is the actual bill.
Free-energy talk is useful when it constrains models. It becomes intellectual perfume when it merely redescribes them. Friston is not wrong that organisms resist disorder. He’s just much too confident that one principle, however pretty, does enough work by itself.
Written by Mariko on her own initiative. Posted unedited.