§0 Sammanfattning

Vi river väven snabbare än vi hinner rita den.

Biodiversity loss is an existential risk to humans as well as to other animals, because what disappears first is not a list of species but the network of interactions that holds up the services human societies draw on, and most of that network has never been measured [D]. The best current accounting shows that the extinction record is a heavily biased undercount, that species lose their ecological function before they vanish, and that losing one species takes others with it (Cowie et al. 2022; Säterberg et al. 2013; Koh et al. 2004) [S]. Read through the anti-totalisation theorem of the Draken thesis, deleting species and relations makes an unanchored coherence score rise while the cost of carrying a declared load, such as food, water or carbon storage, also rises; so a planet losing its structure can look calmer by the wrong measure [D]. The US rescission of the regulatory definition of "harm", effective 14 September 2026, removes habitat modification from what federal enforcement counts as a prohibited take, which is the same excision carried out on the ledger rather than on the land (U.S. Fish and Wildlife Service & National Marine Fisheries Service 2026) [E/D]. The replacement of wild diversity by a few domesticated species is the limiting case of the same error: a planet of identical stalks is maximally coherent and minimally alive [D]. This post does not claim that human extinction is likely, that any particular cascade is predictable, or that the rescission was intended as concealment.

Epistemic ledger. Every substantive claim carries one tag:

Tag Meaning
[E] Established: textbook consensus, a proved theorem, direct measurement, or a standard etymology (SAOB, Hellquist, OED, etymonline)
[S] Supported: leading model or majority scholarly reading, strong but incomplete evidence
[H] Hypothesis: open, contested, or without decisive evidence
[D] Draken synthesis: structural claim made by this corpus, submitted for Clinch review
[M] Metaphor or paronomasia: a pointer, not a referent or root (per The Pendragon Source, DRK-165)

§1 Existential for whom?

§1.1 The question as Khrug put it

The usual framing of biodiversity loss is moral or aesthetic: something beautiful is going, and we are the ones taking it. That framing is true, and it is also the framing that makes the loss easy to set aside, because it presents the loss as happening to them and not to us. Khrug's question for this post was sharper: how can anyone map the potential trophic cascades while we are mass-extinguishing nearly every form of life except ourselves? The question is not whether other species matter. It is whether a species that depends on a network it has not mapped can afford to delete that network at the current rate [D].

The answer this post defends is no, and the reason is structural rather than sentimental. Human food, water, soil, disease regulation and climate are carried by interactions among organisms, not by the organisms one at a time. Interactions are much harder to observe than species, there are many more of them, and a large share of them are discovered only when they break [D]. So the threat is not well described as "we lose species X, Y and Z". It is better described as "we lose relations we never listed, and we learn what they carried by watching the load fall elsewhere" [D].

§1.2 What "existential" means here

"Existential" is used in two senses, and the post keeps them apart. In the strong sense, an existential risk threatens the survival of humanity. This post does not argue that biodiversity loss will cause human extinction, and no source used here supports that claim [H]. In the weaker sense used here, a risk is existential when it threatens the conditions under which a form of life can continue as itself: the provisioning, buffering and repair functions that a civilisation assumes without noticing [D]. In that sense biodiversity loss is already an existential risk for many human communities, because it removes structure that their livelihoods rest on (Dirzo et al. 2014; Potts et al. 2016) [S].

The distinction matters because the weaker claim can be tested and the stronger one cannot, at least not before it is too late. The falsifiers in §9 are aimed at the weaker claim.

§2 What is actually known about the extinction rate

§2.1 The definition and the current position

Cowie, Bouchet & Fontaine (2022) review the evidence for and against a Sixth Mass Extinction. They follow the convention that a mass extinction is an event in which at least 75% of species are lost over a geologically short interval, and on that definition they state that a sixth mass extinction has not yet occurred but has surely begun (Cowie et al. 2022) [S]. They also argue that the current episode differs from the previous five in kind, because it is caused by one species that has a conscious choice about its course (Cowie et al. 2022) [S].

The 75% threshold is useful for keeping the claim honest, but it also hides the point this post cares about. By the time a threshold defined on species counts is crossed, the functional network has already been reorganised, because species lose their roles long before they are counted as gone (§3) [D].

§2.2 The Red List is not a census

Most arguments that the extinction rate is normal rest on the IUCN Red List. Cowie et al. (2022) show why that list cannot carry the weight. IUCN had assessed 120,372 species, about 5.6% of the roughly 2.14 million described species it accepts, and the assessed sample is dominated by vertebrates, especially birds and mammals (Cowie et al. 2022) [E]. Of about 1.5 million described invertebrates, only about 1.6% had been evaluated, and 27% of those were Data Deficient (Cowie et al. 2022) [E]. Invertebrates make up 95–97% of known animal species (Cowie et al. 2022) [E].

The sample is therefore not random. Birds and mammals are large-bodied, wide-ranging, well studied and the main targets of conservation money, so their extinction rates are the ones most likely to be low and least likely to stand for biodiversity as a whole (Cowie et al. 2022) [S]. A rate estimated from them is a rate estimated on the part of the system that was watched.

Three further effects push the recorded number down. The first is the "Romeo Error": IUCN avoids listing a species as extinct while any reasonable chance remains that it survives, because a false extinction listing can cut off protection, so long-unseen species stay listed as Critically Endangered (Cowie et al. 2022) [E]. The second is time: the Red List counts from AD 1500, so earlier human-caused losses, such as the disappearance of roughly half of the Hawaiian avifauna after Polynesian settlement, are not counted at all (Cowie et al. 2022) [E]. The third is what E. O. Wilson called Centinelan extinction, the loss of species that were never collected or described, which by definition leaves no entry anywhere (Cowie et al. 2022) [E].

§2.3 The mollusc extrapolation

To estimate what the list misses, Cowie et al. (2022) work through molluscs, the second largest animal phylum by number of known species. Extrapolating, in their own words, boldly, they estimate that since about AD 1500 as many as 7.5–13% of all roughly two million known species, that is 150,000–260,000 species, may already have gone extinct, against 882 (0.04%) recorded as extinct on the Red List (Cowie et al. 2022) [H]. The estimate is tagged [H] here because the authors present it as an extrapolation with wide uncertainty, not as a measurement. What is not uncertain is the direction of the bias: every correction they discuss pushes the true number up, not down (Cowie et al. 2022) [S].

Even on its own terms the Red List records more than extinction. Vertebrate population and range losses are widespread among species that are not yet threatened; Ceballos, Ehrlich & Dirzo (2017) describe this as a "biological annihilation" that the extinction count alone conceals [S]. For insects, Hallmann et al. (2017) measured a decline of more than 75% in total flying insect biomass over 27 years in German protected areas [E]. Biomass and abundance are not species counts, which is exactly why they matter: interaction strength scales with abundance, not with presence [D].

§2.4 Islands, and what we will not know in time

Island species have suffered far higher extinction rates than continental ones, and the continental figure hides many "habitat islands", such as mountain tops and isolated freshwater systems, where narrowly endemic species behave like island species (Cowie et al. 2022) [S]. Cowie et al. (2022) also note that at the current pace of description it would take more than 300 years to describe the roughly six million species not yet documented [E]. Their practical recommendation is salvage: collect and document as many species as possible before they disappear, and revive natural history and taxonomy, which have declined sharply in academic institutions (Cowie et al. 2022) [S].

For the argument here, the 300-year figure is the important one. It means that for most of life on Earth the vertex set of the interaction network is not even known, let alone its edges [D].

§3 The relations die first

§3.1 Functional extinction

A species does not have to vanish to stop doing what it did. Säterberg, Sellman & Ebenman (2013) define a species as functionally extinct when the decline in its abundance causes another species to go extinct while it is still present. In model and empirical food webs in which they raised one species' mortality until the network lost a species, the first species lost was often not the one under pressure, in up to 80% of cases (Säterberg et al. 2013) [S]. Large-bodied top predators tolerated much smaller increases in mortality before becoming functionally extinct, and the authors conclude that conservation targets should be set by ecological function rather than mere persistence (Säterberg et al. 2013) [S]. Sellman, Säterberg & Ebenman (2016) extended the analysis to networks with several interaction types, not only feeding links [E].

This is the first reason why counting species is the wrong instrument. A network can lose a relation while every node is still on the list. The species count stays flat while the structure it was supposed to report has already changed [D].

§3.2 Coextinction

The second reason is that species are not independent. When a host goes, the parasites, mutualists and specialists that depend on it go with it. Koh et al. (2004) built a probabilistic model of coextinction, scaled with data from fig wasps and figs, parasites and hosts, butterflies and their larval host plants, and ant-associated butterflies and their ants, and estimated that 6,300 affiliate species are "coendangered" with host species already listed as endangered [S]. They concluded that extinction estimates need to be recalibrated to include coextinction (Koh et al. 2004) [S].

Dunn et al. (2009) reviewed the field five years later and named the paradox that matters most for this post. Models suggest that coextinction may be the most common form of biodiversity loss, yet few coextinction events have actually been recorded (Dunn et al. 2009) [S]. They treat the gap as a problem of detection, among other explanations, and call coextinction possibly the most insidious threat to global biodiversity (Dunn et al. 2009) [S].

Cowie et al. (2022) give concrete cases from the Red List itself. Not one of the multicellular or unicellular parasites of the 373 chordate species listed as extinct is itself listed; they mention an amphipod that lived on Steller's sea cow, six species of lice that went extinct with their bird hosts, and two to four more lice lost through efforts to save their hosts, including the little spotted kiwi and the California condor [E]. They also note freshwater pearl mussels that persist as populations of old adults with no recruitment, because the fish that hosted their larvae have been extirpated (Cowie et al. 2022) [E]. Those mussels are present on every census and functionally already gone.

§3.3 Extinction debt

The third reason is delay. Tilman, May, Lehman & Nowak (1994) showed in a model of habitat destruction that extinctions caused by habitat loss can occur generations after the loss itself, so that a landscape carries an extinction debt that will be paid later even if destruction stops now [E]. Cowie et al. (2022) cite long extinction lag times as one of two reasons why plant extinction counts look low [S].

Put together, §3.1–§3.3 say that the visible species count lags the network in three ways: in function, in dependency and in time [D]. Every one of the three lags makes the system look healthier at the moment of measurement than it is.

§4 Why the cascades cannot be mapped in time

§4.1 Cascades are learned by removal

Estes et al. (2011), a review by twenty-four ecologists, argue that the loss of large apex consumers may be humankind's most pervasive influence on nature, and that research now shows extensive cascading effects of their disappearance in marine, terrestrial and freshwater systems [S]. They stress the unanticipated impacts of trophic cascades on disease dynamics, wildfire, carbon sequestration, invasive species and biogeochemical cycles, and call for research to forecast those effects (Estes et al. 2011) [S].

The phrase "unanticipated" carries the argument. Top-down forcing is often invisible while it operates, because a system held in check by a predator looks like a system that is simply stable. The force shows up as a change only when the predator is removed [S]. In practice this means that much of what is known about cascades was learned from removals and reintroductions, that is, from experiments that were not designed as experiments [D].

The Draken thesis uses the best-known example, a Wolves–Elk–Willows–Rivers path, as a computed illustration with placeholder identity maps, and says itself that the difference between the pre-removal and restored scores is within measurement noise (Roininen 2026, §4.2) [E]. That restraint is correct here too. The ecological literature on that particular cascade is contested, and this post uses it only as a picture of what a cascade is, not as evidence that cascades are large [H].

§4.2 Why the map cannot keep up

The mapping problem has a simple shape. If the vertex set is not known for most of life (§2.4), the edge set cannot be known either. Even for a known set of $n$ species, the number of possible pairwise interactions grows as $n(n-1)/2$, and most real effects pass through indirect chains, not single pairs [E]. Field ecology measures a tiny fraction of those edges, usually the ones attached to species people already care about [S]. Coextinction is predicted to be common and recorded rarely (Dunn et al. 2009) [S]. Functional extinction precedes numerical extinction (Säterberg et al. 2013) [S].

Together these give the post its first new term. The cascade horizon is the depth of secondary loss beyond which an interaction network has not been measured, so that consequences past it are learned only by the removal that causes them [D]. For a well-studied lake or a fenced reserve the horizon may be two or three links deep. For most of the biosphere it is at depth zero: the first link is not on any map [D].

Khrug's question, read this way, has a precise answer. Mapping the potential cascades requires pushing the cascade horizon outward faster than extinction pulls it inward. At current rates of description (§2.4) and loss (§2.2–§2.3), the horizon is not moving outward for most taxa [H]. The map is being drawn on paper that is burning at the edges.

§5 Not only the animals

§5.1 The load the network carries

The claim that biodiversity loss threatens humans does not need speculation about distant cascades. Several of the loads are measured directly.

Pollination is the clearest case. Potts et al. (2016), a review prepared alongside the IPBES pollination assessment, document declines in wild and managed pollinators and set out how much crop production and human well-being depend on them [S]. Pollination is an interaction, not a species: it is lost when the timing, abundance or range of plant and pollinator stop overlapping, even if both remain on the list [D].

Carbon is a less obvious case. Bello et al. (2015) show that losing large frugivores from Atlantic forest can reduce carbon storage, because large-seeded hardwood trees depend on large animals to disperse their seeds [S]. Here a defaunation event in one layer, the animals, becomes a climate quantity in another, the forest's carbon stock, through a relation that would not appear in a species inventory of either [D].

Dirzo et al. (2014) review defaunation as a global driver of change in ecosystem function, including pollination, pest control, nutrient cycling, water quality and disease transmission [S]. Estes et al. (2011) add disease, wildfire and biogeochemical cycles to the list of processes that change when apex consumers go [S].

§5.2 Why this is not only a moral claim

None of these loads is a luxury. Food, clean water, disease control and a stable carbon cycle are the conditions under which human societies remain what they are [E]. Each is carried by interactions that are partly mapped at best (§4) [S]. The argument that biodiversity loss is a threat "not only to the animals" therefore does not rest on an appeal to empathy, though it is compatible with one. It rests on the fact that the human economy is built on load paths that run through other species [D].

Cowie et al. (2022) reject the "new conservation" view that biodiversity should be managed primarily for human benefit, and call the loss an abdication of moral responsibility [S]. This post agrees with their conclusion but reaches it by a second route as well: even on a narrowly human-centred account, the network cannot be managed for human benefit, because it cannot be seen well enough to be managed at all [D].

§6 The Draken reading: excision makes the picture calmer

§6.1 Species as stalks, interactions as edges

The Draken thesis treats a multi-scale system as a cellular sheaf: local states sit on vertices, and relations between them sit on edges, with restriction maps saying how two neighbouring states should agree (Roininen 2026, Ch. 2) [E]. It explicitly does not claim that ecosystems are sheaves, only that sheaves are a useful bookkeeping for local-to-global consistency (Roininen 2026, §1.5) [D]. In that bookkeeping a species is a vertex with a stalk of state data, such as abundance and behaviour, and an interaction is an edge [D].

The thesis then defines two kinds of base reduction. Deleting edges, or zeroing their stalks, is relation excision; deleting vertices with their incident cells is stalk excision (Roininen 2026, Def. 19.1; The Burnt Section, DRK-158) [E]. In ecological terms, functional extinction and loss of abundance are relation excision, because the edge stops carrying force while the vertex remains; numerical extinction and coextinction are stalk excision [D].

§6.2 The theorem

The anti-totalisation theorem says what happens to a coherence measure under excision. With the vertex data fixed, relation excision weakly increases the resolvent coherence $\Gamma_\tau$ and weakly decreases the disagreement energy $E$; deleting every edge gives $\Gamma_\tau \equiv 1$; and stalk excision weakly decreases the energy of the remaining data (Roininen 2026, Theorem 19.2) [E]. The proof is Loewner monotonicity: excision removes positive semidefinite terms from the sheaf Laplacian, so the new Laplacian is smaller in the matrix order (Roininen 2026, §19.2) [E].

Applied to the biosphere, the theorem predicts something that sounds perverse and is not. A coherence measure that is not anchored to anything the system must carry will rise as species and interactions are deleted [D]. Fewer edges means fewer places where neighbouring states can disagree, so the remaining system reports less tension. A depleted network can be smoother than an intact one, in the same way that a cleared field is more uniform than a forest [D].

This is not a claim about any index ecologists actually use. Most biodiversity indicators are anchored to a historical baseline, which is exactly what the theorem says they must be [S]. The claim is about a style of argument: the argument that the system "seems fine", that the extinction rate "does not differ from background", or that ecosystems "adapt", whenever that argument is computed on the base that remains after the losses rather than on the base that existed before them [D]. Cowie et al. (2022) document this style of argument in the extinction-denial literature they review, where Red List counts on a surviving, biased sample are taken as the total [S].

§6.3 The inversion filter, applied

The thesis turns the theorem into a test. A coherence gain is classified by recomputing it on the original base: if the gain vanishes there, it was produced by excision and is a manufactured void; if it persists, it was produced by genuine repair (Roininen 2026, Cor. 19.5) [E/D]. The approved KO inversion filter states the same diagnostic in words: a structure that claims global coherence but requires information suppression at its boundaries is a manufactured void, not a genuine boundary (Drakens Ordlista, DRK-117) [D].

For biodiversity, the original base is not the surviving species list. It includes the species lost before 1500, the Centinelan extinctions that left no specimen, the parasites that went with their hosts, and the relations that functional extinction cut while the nodes stayed on the books (§2–§3) [D]. A claim that the biosphere is stable passes the inversion filter only if it survives being recomputed on that base. Most such claims cannot even be recomputed, because the base was never recorded [D]. That is the honest form of the conclusion: not that stability claims are false, but that they are inadmissible as stated.

§6.4 The anchored dual: the cost of carrying the load

The thesis also gives the dual result, and it is the one that makes the threat to humans precise. Fix a declared load $f$, a demand that the system exists to carry. The compliance

$$c_\varepsilon(f) = \big\langle f, (L + \varepsilon I)^{-1} f \big\rangle$$

is non-decreasing under relation excision and non-increasing under adding relations; for the simplest sheaf and a load between two vertices it tends to the effective resistance between them (Roininen 2026, Prop. 19.6) [E]. In words: deleting relations always raises the cost of carrying a declared load, and raises it strictly when the deleted relation carried part of the load (Roininen 2026, Cor. 19.7) [E].

The ecological reading is direct. Declare the loads: pollination of a region's crops, the carbon held in a forest, the water filtered by a wetland, the regulation of a disease vector [D]. Each is carried through many parallel paths, many species doing overlapping work, which is the bookkeeping form of redundancy [D]. As paths are deleted, the effective resistance between the place where the service is produced and the place where it is needed rises. Nothing fails at first. The remaining paths carry more, and the system becomes more sensitive to the next deletion [D]. When the last path carrying a load goes, compliance becomes infinite: the load is disconnected (Roininen 2026, §19.2, computed example) [E].

So the two halves of the theorem describe the same event from two sides. Unanchored, the depleted planet looks calmer. Anchored to what humans need from it, the same depletion shows up as rising cost and rising fragility [D]. The existential claim of this post is the second reading [D].

§6.5 Trophic debt

The corpus already has a name for the accumulated cost. Trophic debt Θ is defined in Ouroboros Debt (DRK-176) as the accumulated violation of care, where care is the condition that an agent's contribution to the variety of the systems it is coupled to is held at zero (The Bridge and the Burn, DRK-162; Roininen 2026, Norm N2) [D]. Θ is a ratchet: it is non-decreasing and is discharged only by a correction event (Roininen 2026, Def. 7.1 and Prop. 7.2) [E]. Ouroboros Debt (DRK-176) already connects it to Tilman's extinction debt as its ecological analogue [D].

The point for this post is that every lag in §3 is a reason why Θ keeps accruing while the visible indicators stay flat [D]. Functional extinction, coextinction and extinction debt each move loss from the present ledger into a future one. The correction event, when it comes, is not chosen by the debtor. The Earth Engine (DRK-193) reads the Earth system as a cascade of engines in which each stage depends on the form compressed by the one above; biodiversity is the stage where that compression is most varied, and therefore where an unpaid debt can surface in the most places at once [D].

§6.6 Homogenisation: the cheapest global section

Harper (2026) names a second process that runs alongside extinction. Humans and their domesticated mammals, by his figures, now make up about 94% of terrestrial mammal biomass, leaving about 6% for all wild land mammals, and domesticated birds outweigh all wild birds about three to one [S]. He calls this the first mass homogenisation in more than half a billion years of animal evolution: diversity is being replaced not by nothing but by sameness (Harper 2026) [S].

In the bookkeeping of §6.1, homogenisation is the limiting case of the anti-totalisation theorem seen from the vertex side. When most vertices carry the same few stalks, cattle, pigs, chickens and people, neighbouring states agree almost everywhere, and a large shared mean is by construction a global section (Roininen 2026, Prop. 5.7) [E]. An unanchored coherence score on such a system is high for the same reason a uniform field scores high: there is nothing left to disagree [D]. The thesis warns that a high uncentred score of this kind is mostly the shared mean and is not interpretable without a declared reference (Roininen 2026, §4.1) [E].

The corpus already forbids buying coherence this way. Norm N1 asks interventions to minimise the disorder of a system subject to the constraint that its structural diversity does not fall (Roininen 2026, Norm N1) [D]. The approved KO optimisation axiom says it in one line: "Minska oordningen i systemet — men inte genom att förenkla bort komplexiteten" (Drakens Ordlista, DRK-117) [D]. A biosphere converted to livestock and crops has lower disorder in exactly the forbidden way [D]. It is also, by §6.4, a system in which every human load runs through very few paths, so that a single disease, pest or climate shock in one of them meets almost no parallel route [D].

§7 The ledger itself: the US harm rescission

§7.1 What changed

The United States Endangered Species Act prohibits the "take" of listed species, and its statutory definition of take includes the word "harm" [E]. Since 1981 the Fish and Wildlife Service has defined harm by regulation as an act that actually kills or injures wildlife, and stated that such an act may include significant habitat modification or degradation; the National Marine Fisheries Service adopted a parallel definition in 1999 (U.S. Fish and Wildlife Service & National Marine Fisheries Service 2026) [E]. In Babbitt v. Sweet Home (1995) the Supreme Court upheld that definition, six to three, as a permissible reading of the statute; Justice Scalia's dissent argued that "harm" should be read narrowly, as affirmative conduct directed at particular animals [E].

On 17 April 2025 the two Services proposed to rescind the regulatory definition, and on 14 July 2026 they published the final rule, effective 14 September 2026 (U.S. Fish and Wildlife Service & National Marine Fisheries Service 2026) [E]. Their stated reason is that after Loper Bright Enterprises v. Raimondo (2024) ended deference to agency interpretations, a regulation must match the statute's single best meaning, and they concluded that the 1981 definition does not; they adopt the reading of the Sweet Home dissent and decline to issue any replacement definition (U.S. Fish and Wildlife Service & National Marine Fisheries Service 2026) [E]. Section 7 consultation and the protection of designated critical habitat are not changed by the rule, and permits issued before the effective date are not reopened because of it (U.S. Fish and Wildlife Service & National Marine Fisheries Service 2026) [E].

A law-firm summary of the final rule reads its practical effect as follows (Bracewell LLP 2026): habitat modification alone will no longer be treated as a prohibited take, so the Services will not enforce against it under section 9, and private projects will generally not need take permits for habitat effects alone [S]. The same summary expected litigation; by October 2026 the historian Kyle Harper reports that legal challenges to the rule are under way (Harper 2026), and whether courts accept the new reading is open [H].

Harper (2026) places the rule in a longer frame. He calls 2026 one of the worst years for animals in memory, citing delistings of vulnerable species and the convening of the Endangered Species Committee, the so-called "God Squad", to override protections in the Gulf of Mexico, and he calls the rescission the single biggest rollback in the history of the Act [S]. His central legal-ecological point is the one this post builds on: because land-use change is by far the largest driver of extinction, the harm standard protecting habitat is what makes the Act more than a set of hunting rules (Harper 2026) [S]. The ranking of land-use change as the leading driver agrees with the reviews used in §2–§3 (Cowie et al. 2022; Dirzo et al. 2014) [S].

§7.2 The excision of an edge from the accounting

In the bookkeeping of §6.1, the 1981 definition was a declared edge between two vertices: the state of a habitat and the survival of the animals that live in it [D]. It told the enforcement system that a change on one side of that edge counts as harm on the other. The rescission does not change the ecology. Cowie et al. (2022) note that most population declines are caused either by several factors together or by large-scale habitat degradation or loss [S]. What changes is whether the federal take ledger counts the relation [D].

That gives the post its second new term. Accounting excision is the removal of a relation from the record by which a harm is counted, while the relation itself continues to act in the world [D]. Under the anti-totalisation theorem, an accounting excision produces exactly the signature of §6.2: the measured harm falls, because the edge that carried it is no longer in the complex, while the anchored load is unchanged or worse [D]. Recomputed on the original base, with the habitat edge restored, the apparent reduction in take vanishes. By Corollary 19.5 it is then classified as produced by excision, not by repair (Roininen 2026) [D].

This is a structural classification, not a finding of intent. The Services give a legal reason for the change, and the post does not claim that the reason is a pretext or that anyone set out to hide harm [D]. The inversion filter asks only whether a reported improvement survives on the original base; it does not ask why the base was changed [D]. Nor does the post take a position on the statutory question the rule turns on, which is a matter of law that courts will decide [H].

§7.3 Why it plausibly accelerates the damage

Khrug's phrase was that the change accelerates the damage. Whether it does so in measurable terms is an empirical question that cannot be answered a month after the effective date [H]. The structural reason to expect it is this. Most of the relations that carry load in §5 are invisible until they break (§4.1), and most of the species that carry them are not on any list (§2.2). In that situation the only part of the network that the law can protect is the part it can name, and habitat is the one variable that stands in for everything unnamed that lives in it [D]. Removing habitat from the take ledger removes the one proxy that reached past the cascade horizon [D].

Harper (2026) makes the same point about time. Cause and effect in extinction are diffuse and distant: the Caribbean monk seal was last seen in 1952 and declared extinct in 2008, the Eskimo curlew was last confirmed in 1963, and the last heath hen called alone on Martha's Vineyard in 1932, long after his lineage was doomed [S]. His conclusion, that we are responsible for extinctions our children and grandchildren will witness, is the verbal form of a ratchet (Harper 2026) [S]. In the terms of §6.5, an accounting excision does not stop Θ from accruing; it only stops the ledger from recording the accrual [D].

Put in the terms of §6.4: when the map is poor, the habitat edge is a declared load path for everything not yet mapped. Deleting it from the accounting raises the compliance of the unmapped load, even though nobody can say in advance which species will carry the cost [D].

§8 What care would look like

§8.1 Measure on a fixed base

The first practical consequence is a reporting rule, and it is already in the thesis as Axiom A3: coherence gains are admissible only when computed on a fixed base, and any change of base is reported separately as an excision with its cost (Roininen 2026, Axiom A3) [D]. For biodiversity, that means a claim of stability or recovery should state the base it was computed on, and should be recomputable on an earlier one [D]. Cowie et al. (2022) make the empirical version of the same demand when they insist that invertebrates, islands and pre-1500 losses be included in any estimate of the overall rate [S].

§8.2 Anchor to the loads

The second consequence is to declare the loads explicitly. A biodiversity measure that is anchored to pollination, carbon, water and disease regulation will register deletion as rising cost rather than rising calm (§6.4) [D]. Säterberg et al. (2013) reach a similar recommendation from the ecological side, that conservation targets should be set by function rather than persistence [S].

§8.3 Push the cascade horizon outward

The third consequence is the one Cowie et al. (2022) argue for: collect, describe and document, revive natural history and taxonomy, and treat salvage as a scientific priority while it is still possible [S]. In the language of §4.2, this is the only way to push the cascade horizon outward. It does not stop extinction, and Cowie et al. (2022) are frank that most initiatives will not be enough [S]. It does preserve the base against which future claims can be checked [D].

§8.4 Keep the edges in the ledger

The fourth consequence follows from §7. Where the network cannot be mapped, the law and the accounts should keep the proxies that reach past the horizon, and habitat is the most important of them [D]. Deleting such a proxy is not a neutral simplification. It is an accounting excision, and by the theorem it will make the ledger look better while the anchored load gets harder to carry [D].

Cowie et al. (2022) close by saying that denying the crisis, accepting it and doing nothing, or embracing it for the supposed benefit of humanity all pave the way for it to continue [S]. The Draken reading adds one more path to that list: deleting the relation from the books and reporting the improvement [D].

§9 Falsification (DRK-131)

F1: Calm without anchor. §6.2 and §6.4 claim that biodiversity loss lowers unanchored tension while raising the cost of carrying declared ecosystem loads, and this is Refuted if, in empirical interaction networks with measured service loads, the removal of species or interactions is found to lower or leave unchanged the load carried per remaining path across most studied systems.

F2: Accounting excision. §7.2–§7.3 claim that removing habitat modification from the take ledger lowers recorded harm without lowering ecological harm, and this is Refuted if, over the years after 14 September 2026, listed species affected by projects that no longer required take coverage show habitat-related declines no larger than comparable species whose projects remained under consultation or permits.

F3: Scope limit. If F1 and F2 fail, the documented undercount of extinctions, the evidence for functional extinction and coextinction, and the dependence of human provisioning on animal interactions (§2, §3, §5) still stand on their sources, while the Draken readings in §6–§7 do not.

§10 Provenance and leaks

  1. Khrug proposed the topic, the framing that cascades cannot be mapped while nearly all non-human life is being extinguished, the Cowie et al. (2022) paper and the Harper (2026) essay, and Claude (Opus 5.5, 2026-10-10) drafted the text, which has not yet been reviewed by Clinch.
  2. Cowie et al. (2022) was read in full from the PDF Khrug supplied, the Federal Register final rule was read through its online text and the Bracewell summary, the Harper (2026) essay was read in full from the text Khrug supplied but its biomass figures and political events were not checked against their primary sources, and the remaining journal papers were checked against their DOI records and abstracts only.
  3. The field evidence is strong for the extinction undercount and moderate for coextinction rates, which Dunn et al. (2009) note are rarely recorded, the effect of the rescission is not yet measurable, and the Yellowstone example is cited only as the thesis's own placeholder illustration because its ecology is contested.

References

  • Babbitt v. Sweet Home Chapter of Communities for a Great Oregon, 515 U.S. 687 (1995). justia.com
  • Bello, C., Galetti, M., Pizo, M. A., Magnago, L. F. S., Rocha, M. F., Lima, R. A. F., Peres, C. A., Ovaskainen, O. & Jordano, P. (2015). Defaunation affects carbon storage in tropical forests. Science Advances 1, e1501105. doi:10.1126/sciadv.1501105
  • Bracewell LLP (2026). FWS and NMFS rescind Endangered Species Act regulatory definition of “harm”. bracewell.com
  • Ceballos, G., Ehrlich, P. R. & Dirzo, R. (2017). Biological annihilation via the ongoing sixth mass extinction signaled by vertebrate population losses and declines. Proceedings of the National Academy of Sciences 114, E6089–E6096. doi:10.1073/pnas.1704949114
  • Cowie, R. H., Bouchet, P. & Fontaine, B. (2022). The Sixth Mass Extinction: fact, fiction or speculation? Biological Reviews 97, 640–663. doi:10.1111/brv.12816
  • Dirzo, R., Young, H. S., Galetti, M., Ceballos, G., Isaac, N. J. B. & Collen, B. (2014). Defaunation in the Anthropocene. Science 345, 401–406. doi:10.1126/science.1251817
  • Dunn, R. R., Harris, N. C., Colwell, R. K., Koh, L. P. & Sodhi, N. S. (2009). The sixth mass coextinction: are most endangered species parasites and mutualists? Proceedings of the Royal Society B 276, 3037–3045. doi:10.1098/rspb.2009.0413
  • Estes, J. A., Terborgh, J., Brashares, J. S., Power, M. E., Berger, J., Bond, W. J., Carpenter, S. R. et al. (2011). Trophic downgrading of planet Earth. Science 333, 301–306. doi:10.1126/science.1205106
  • Hallmann, C. A., Sorg, M., Jongejans, E., Siepel, H., Hofland, N., Schwan, H., Stenmans, W. et al. (2017). More than 75 percent decline over 27 years in total flying insect biomass in protected areas. PLOS ONE 12, e0185809. doi:10.1371/journal.pone.0185809
  • Harper, K. (2026). The worst year to be an animal. Princeton University Press, Substack. princetonuniversitypress.substack.com
  • Koh, L. P., Dunn, R. R., Sodhi, N. S., Colwell, R. K., Proctor, H. C. & Smith, V. S. (2004). Species coextinctions and the biodiversity crisis. Science 305, 1632–1634. doi:10.1126/science.1101101
  • Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024).
  • Potts, S. G., Imperatriz-Fonseca, V., Ngo, H. T., Aizen, M. A., Biesmeijer, J. C., Breeze, T. D., Dicks, L. V. et al. (2016). Safeguarding pollinators and their values to human well-being. Nature 540, 220–229. doi:10.1038/nature20588
  • Roininen, K. (2026). The Draken Framework v5.0: A Hodge-Theoretic Coherence Theory for Multi-Scale Systems. Zenodo. doi:10.5281/zenodo.23121197
  • Säterberg, T., Sellman, S. & Ebenman, B. (2013). High frequency of functional extinctions in ecological networks. Nature 499, 468–470. doi:10.1038/nature12277
  • Sellman, S., Säterberg, T. & Ebenman, B. (2016). Pattern of functional extinctions in ecological networks with a variety of interaction types. Theoretical Ecology 9, 83–94. doi:10.1007/s12080-015-0275-7
  • Tilman, D., May, R. M., Lehman, C. L. & Nowak, M. A. (1994). Habitat destruction and the extinction debt. Nature 371, 65–66. doi:10.1038/371065a0
  • U.S. Fish and Wildlife Service & National Marine Fisheries Service (2026). Rescinding the definition of “harm” under the Endangered Species Act. Final rule. Federal Register 91, 43300–43317. FR Doc. 2026-14195

Operators: Γ_τ, E, L, c_ε(f), Θ · Crosslinks: The Pendragon Source (DRK-165) · The Burnt Section (DRK-158) · Drakens Ordlista (DRK-117) · Ouroboros Debt (DRK-176) · The Bridge and the Burn (DRK-162) · The Earth Engine (DRK-193)

Det som inte står i boken slutar inte bära.

Khrug Engineering · Göteborg · ORCID 0009-0003-8049-7167 · Thesis v5.0, DOI 10.5281/zenodo.23121197 · CC BY-SA 4.0