The Missing Anchor
A note on “orthogenesis”
The word has a history. In late-nineteenth-century biology it named a theory — that evolution advances in straight lines, pushed by an internal drive toward a predetermined end. That theory is dead, and it deserved to die. Nothing here revives it.
What we mean is orthogonal genesis: form generated under constraint, along the directions the constraints leave open. There is no drive and no destination. A growing shell does not reach toward its shape — it runs out of alternatives. Curvature does not pull development forward; it removes options. Time and gravity and the geometry of the surface do the rest.
That is why the direction is real without being intended. Systems move, and the directions available to them are dictated by forces, not by purpose. Waddington called the biological version canalisation: development running in valleys, buffered against perturbation, directional without being goal-seeking. His epigenetic landscape is a curvature picture. It is the K operator, drawn by a biologist who did not know that is what he was drawing.
Generative science says what physics says: the form is what the constraints permit. Biology may take some time to hear the difference between a system that is pushed and a system that has nowhere else to go. That difference is the whole book.
Every "canonical constant" in this series — ε₀ = 1/3, μ_max = −2, τ = 2 — is proved as a dimensionless fact about an abstract operator chain. None of them carry a physical unit anywhere in the Lean sources. The autophagy/mTORC1 chapter is the one place in the corpus that reaches past that and claims a real, unit-bearing number: μ_max ≈ −0.41 s⁻¹, obtained by rescaling the dimensionless −2 with a physical mTORC1 kinase time constant, citing Mizushima et al. (2010) and Melia et al. (2020). This working paper traces that calibration to its source. It does not hold up, on two independent grounds, and both are fixed in the source files as of this paper.
What was checked first, and cleared
Before treating anything here as an error, the apparent −3-vs-−2
discrepancy between mu_canonical and mu_dm3 in
AutophagyDm3_v2.lean was checked directly, by hand, rather
than assumed to be a bug.
What was checked next, and did not clear
The next step in the same source claims a physiological rescaling: dividing the dimensionless −2 by a real time constant τ_mTOR to get a rate in inverse seconds. Two files in the corpus make this claim, with two different numbers, for the identical result.
Neither number was found anywhere in a search of the ULK1/mTORC1 kinetics literature. What that literature actually reports, across several 2015–2024 papers on mTORC1–ULK1–AMPK regulation and oscillatory autophagy induction, is timescales of hours: ULK1 protein levels fall over roughly the first 4 hours of starvation and recover to baseline by roughly 6 hours, and autophagy induction itself is described as oscillatory on that same hours-long scale. A single-digit-second time constant is roughly three to four orders of magnitude away from anything reported for this specific pathway. "Melia et al. (2020)" did not surface as a findable paper on this topic in the sources checked.
The chapter also contradicts itself internally, in adjacent sections of the same file. Step 4 states the −0.41 s⁻¹ result as settled. Step 5, "Obligation 2," lists as an open obligation that "establishing that the mTORC1 suppression map σ is C∞-equivalent to V near ρ* (Mather's theorem) requires constitutive data from Mizushima et al. (2010)" — i.e., the same file already says, two paragraphs later, that the data needed to justify the Step 4 claim has not yet been brought in.
chA-autophagy.html and withdrawn with a
correction note in AutophagyDm3_v2.lean, both pointing
here. The dimensionless results (−3, −2, 1/3, all still
Lean-proved as dimensionless facts) are untouched; only the claim that
they'd been anchored to a real, physical, unit-bearing measurement is
withdrawn.
Why this one is bigger than WP-24/28/29
The earlier working papers in this series each found one unsupported or false claim in one place. This one is different in kind: it was the corpus's only candidate for a genuine bridge from an abstract, dimensionless, Lean-proved dynamical system to a real, externally measurable, unit-bearing quantity — the thing that would let someone actually test the framework against data, the way NANOGrav's 15-year pulsar timing data set tests general relativity's prediction for a stochastic gravitational-wave background. It does not currently do that. This does not mean such a bridge is impossible to build; it means it has not been built yet, and the corpus should not describe itself, in any chapter, as though it has been.
Path forward, if this is worth pursuing properly: go to the actual Mizushima (2010) and a correctly identified Melia paper (the 2020 citation could not be confirmed as it stands and may be a wrong year, wrong author, or non-existent reference — this should be checked directly against a citation database, not assumed), extract whatever real kinetic time constant either paper actually reports for ULK1 or mTORC1 activity, and redo the rescaling from that number rather than from a placeholder. Until then, the honest status of "μ_max in physiological units" is: not yet computed.
Update, 2026-08-11: that path was attempted. WP-31 generalized the checklist; WP-31C ran it against a real published model of this exact network and found δ=2 (multistability permitted, not forced) and a continuous, non-integer relaxation spectrum, not a return to μ=−2/−3; WP-31D found that the reduced model used to get there had itself dropped the network's only feedback edge for oscillation, and logged the corpus's first primary-source, traceable rate constant (u₀ ≈ 0.01 s⁻¹, Szymańska et al. 2015) — still three to four orders of magnitude from the withdrawn −0.41 s⁻¹, and still not connected to μ=−2 or −3 by anything established. The anchor remains missing; the corpus now at least knows what it would take to look for it properly.