Summary:
The setup. For thirty years, biogerontology has been split between two camps: aging is a program (an evolved, scheduled, quasi-developmental process — the view associated with Blagosklonny and hyperfunction theory) versus aging is stochastic damage (entropy accumulating with no design behind it). Fedichev’s move is to say both camps are looking at real data and both are misreading it. Aging looks programmed — reproducible trajectories, species-specific lifespans, coordinated hallmarks, master-gene mutants, continuity with development, partial reversibility — but none of that requires a controller.
The mechanism. He reduces aging to a mean-field model built on three macroscopic variables: regulatory resilience, cumulative entropic damage, and noise strength. Damage accrues independently at countless microscopic sites; because those sites all project onto the same few slow macroscopic modes, their uncoordinated failures show up as one smooth, apparently orchestrated decline. Resilience erodes until the system hits a saddle-node bifurcation — the point of collapse. That’s the “mean field”: coordination in the readout without coordination in the underlying events.
Each of the six pro-program observations then falls out for free. Stereotyped trajectories come from a low-dimensional attractor. Species lifespan is set by when the bifurcation arrives, not by a counter. Coordinated hallmarks come from timescale separation — fast biological modes are enslaved to the slow ones. Master-gene effects are single mutations nudging a regulatory eigenvalue toward the bifurcation. Continuity with development and partial reversibility both follow from slow modes relaxing toward a fixed point.
The strongest argument: universality. Different species solve aging with completely different molecular hardware — rDNA circles in yeast, telomeres in humans, other substrates in mice — yet converge on the same phenomenology: Gompertzian mortality and biomarker variance growing linearly with age. Fedichev’s line is that programs are arbitrary, so a program can’t explain convergence. Only mechanism-independent physics — a universality class — produces that. He also derives the development-speed/lifespan correlation from allometry plus energy conservation plus the second law: maintenance cost sets growth rate and damage rate together, no schedule needed.
Evidence against information flow. Three independent signatures suggest aging sites don’t talk to each other: methylation statistics are Poisson (independent rare events), activation barriers follow Gumbel extreme-value statistics (also independence), and single-cell methylation shows high mutual information along developmental pathways but essentially zero correlation among the sites that actually govern maximum lifespan.
A regime distinction. Short-lived “unstable” animals (worms, flies, mice) are dominated by regulatory instability largely independent of damage; “stable” long-lived animals including humans are dominated by damage-driven instability. Much of the old debate, he argues, was two sides describing different regimes.
The experimental test. Caloric restriction and parabiosis both shift the dynamic, reversible, information-rich component of aging — and show no detectable effect on the entropic component. Exactly what the model predicts.
Why it matters. If entropy rather than dysregulation sets maximum lifespan, then “re-instructive” interventions — partial reprogramming, signaling inhibitors — only address the reversible part. Going further would mean restoring from a copy or replacing hardware: macromolecular clearance, tissue engineering, cell therapy. Blagosklonny was right that the coordinated, developmental-looking changes are real; Fedichev’s claim is that they’re readouts of entropic drift, not its cause. The open question he leaves is whether thermodynamic fidelity itself can be reached pharmacologically.
Read the full article: Program without program — Peter Fedichev