There is a pattern that humans have run, in different soils and different centuries, every time they have built a single shared system to hold every kind of activity at once. The pattern terminates the same way. The system absorbs the cost of carrying everything. The cost compounds. The cost becomes uneconomic to carry. The shared system is then drained, fragmented, or paved over in favor of a denser collection of single-purpose replacements that cost more, perform worse, and never reconstitute the function the original ran for free. The pattern is older than the internet by approximately ten thousand years.
Wetlands are the cleanest version of the pattern. A wetland is the most catalytic infrastructure a continent can carry: a single biological substrate that filters fresh water, buffers floods, sequesters carbon, hosts seasonal migration, decomposes organic load, and shelters spawning populations whose productivity feeds whatever lives downstream of them. None of these were design intents. They were emergent properties of a system that had to hold every form of biological traffic at once, and could, because the substrate had no notion of which traffic was authorized. Marsh, fen, swamp, bog, peatland: the technical taxonomy is decorative. The functional category is one. A wetland is a unified medium that absorbs everything biological that arrives.
Then a civilization arrives that has decided, on the basis of immediate cost and immediate use, that the substrate is waste ground. The Ramsar Convention's Global Wetland Outlook 2025 documents what happened next at the planetary scale. At least 411 million hectares of wetlands have been lost since 1970, roughly 22 percent of the world's total [1, 2]. The annual rate of decline has held steady at approximately 0.52 percent for over half a century. About a quarter of the wetlands that remain are classified as in poor ecological condition, and that share is increasing in every region the report assesses. Older estimates push the number further: a 2019 IPBES Global Assessment, working with longer historical horizons, concluded that more than 85 percent of global wetland area has been lost since 1700.
The wetland was not abandoned because anyone decided wetlands were worthless. It was abandoned because the unified substrate looked, from the perspective of the civilization upstream of it, like undifferentiated mud. The optimization that drained it was rational at every step. The downstream consequence is the long-running engineering project of the last century: levee systems, water treatment plants, fish hatcheries, flood control reservoirs, migratory bird refuges, peat carbon offset markets. Each replacement performs a single function the wetland used to perform as a side effect of being. None of them, working together, can carry what the wetland carried.
This is the structure of every drained marsh. It is also the structure of the internet at the present moment.
The unified substrate problem
The current internet is a unified substrate operating at planetary scale. It is the only one humans have ever built that comes close to the wetland in functional breadth. Commerce, family communication, criminal coordination, public health information, scientific publication, child exploitation, religious practice, financial settlement, mental health crisis, education, harassment, archive, surveillance: every category runs on the same packet routing, the same identity layer, the same trust assumptions. The substrate has no architectural concept of which category is which. The protocols transmit each one with identical fidelity. The infrastructure was built to be category-blind because category-blindness was, in 1969, the only way to make the system work.
The economic consequence of category-blindness shows up in the moderation overlay that civilization has been forced to bolt on top of it. The European Union's Digital Services Act, in force at full effect since February 2024, requires every Very Large Online Platform to file a Statement of Reasons for every content moderation decision. The publicly accessible Transparency Database has accumulated, by some independent counts, more than 1.58 billion such statements during a single year of political volatility across eight platforms. Each statement is a record of a decision a human or an automated system made after a piece of content had already been transmitted across the substrate. None of them is a record of the substrate refusing to transmit.
This is the cost of a wetland that has been told to also serve as a postal system. A small enclave of researchers, including Stefano Cresci, Amaury Trujillo, Benedetta Tessa, and their collaborators at the National Research Council of Italy, has spent the last two years auditing whether the Transparency Database actually reveals platform behavior or merely renders the inadequacy of platform behavior more legible [5, 6]. The work is unglamorous and methodical. It involves cross-checking 1.58 billion individual moderation statements against platform-published transparency reports, identifying coherence failures, and documenting where the Database's structural design prevents it from answering the questions it was created to answer. Their finding is one a wetland ecologist would recognize. The instrument that documents the decline is not the instrument that prevents it. The Database is a satellite measuring the marsh from orbit while the marsh is being drained.
The instrument is not the same kind of thing as the substrate.
What the wetland actually does
The mistake civilizations make with shared infrastructure is consistent across every domain in which the mistake has been made. The mistake is to read the substrate's lack of differentiation as a lack of value. A medieval cartographer drawing the lower Mississippi as undifferentiated swampland was not making an error of fact. He was making an error of category, looking at a system whose value was its undifferentiated coupling and concluding, because his categories required separation, that the absence of separation indicated absence of function.
The function of a wetland is the coupling. Water filtration depends on the same hydrology that floods depend on, which depends on the same microbial mat that the peat depends on, which depends on the same anaerobic conditions that the carbon sequestration depends on. Pull any one out and the others degrade. Pull two out and the system stops being a wetland and becomes a series of unrelated ditches that no longer talk to each other. The Ramsar Convention's monitoring data, in its most recent harmonized reporting cycle, recorded that the share of countries reporting deterioration in wetland state had risen from 32 percent to 42 percent across the 2011–2021 window, with the share reporting improvement falling from 23 percent to 14 percent. The trend is not toward stabilization. The wetland is being pulled apart, function by function, by civilizations that decided the unified substrate was an inefficiency.
The internet is now where the wetland was around 1900.
The signs are visible at the same locations a wetland ecologist would look for them: the species that depended on the unified substrate are the ones leaving first. Independent journalism, long-form thought, scientific dissent, the kind of stable political conversation that requires participants to remember each other across years. These are the analogues of the migratory waterfowl. They depended on a substrate that filtered without curating, that hosted without selecting, that buffered without enforcing. The substrate stopped doing those things sometime in the last several years, and the species are migrating to the only adjacent landforms still functioning: gated communities, vetted Substacks, the private Discord servers documented in recent peer-reviewed studies of community moderation, the academic mailing lists that survived because they never scaled. None of these is a wetland. None of them performs the unified function. They are the levee districts the people downstream are now constructing in panic.
The cost the substrate carries
When a wetland is asked to carry the load of every adjacent civilization simultaneously, the wetland fails first along the dimensions the civilizations cared about least. The Mississippi delta, before the Army Corps of Engineers' levee project, was carrying the load of the entire continent's freshwater drainage and getting paid for none of it. The function the system performed for the upstream cities was the cost the downstream marsh absorbed. When the marsh could no longer absorb the cost, when it had been so reduced in extent that its filtration capacity dropped below the load arriving, the failure showed up in places the upstream city had no incentive to monitor. Subsidence. Salt water intrusion. The collapse of the brown shrimp fishery. The disappearance of the barrier islands that had been protecting the upstream city from the storms it had spent two centuries assuming would always be absorbed by something it had stopped paying for.
The internet is in the subsidence stage.
The cost the substrate is now absorbing is no longer paid by users in money or attention; those costs are visible and complained about and continuously priced. The cost the substrate absorbs is paid in human cognitive condition. Brain rot, the term Oxford University Press named its 2024 Word of the Year, defined as the deterioration of mental or intellectual condition resulting from overconsumption of low-quality online material, is a clinical observation about the downstream species. A 2025 review in MDPI's Brain Sciences synthesized research from 2023 and 2024 across PubMed, PsycINFO, and Web of Science, concluding that prolonged exposure to high-volume low-quality digital content correlates with measurable deficits in working memory, executive function, and decision-making, accompanied by emotional desensitization and cognitive overload. The review's authors are not policy advocates. They are reading the substrate's discharge and reporting what they find in it.
This is the distillate that should not need a name. The unified substrate, asked to transmit everything with equal fidelity, is now discharging cognitive sediment at a rate the downstream populations cannot metabolize. The condition is being measured the same way wetland scientists measure nitrogen runoff: not as a moral failing of the people downstream, but as the predictable consequence of a system performing a function it was not designed to perform, beyond the threshold at which it can perform it.
The species that quietly walked
There is a separate signal that wetland ecologists watch for, distinct from the gross metrics of area lost and water quality. They watch for the species that have not yet disappeared but have stopped reproducing in the substrate. The breeding pair counts of certain marsh-dependent waterfowl drop years before the visible collapse of the wetland itself. The animals are still present. The next generation is not.
The internet has its own version of this leading indicator. The recent peer-reviewed work on community moderation conducted by Yoon, Zhang, and Seering at the University of Washington, published in the 2025 Proceedings of the ACM on Human-Computer Interaction, found that experienced moderators of large online communities have begun describing the open web as reconnaissance terrain rather than habitable ground. The metaphor is unprompted and consistent across interview subjects. The participants still visit the open layer for material; they no longer reproduce community there. The next generation of trust-dependent activity is being incubated somewhere else.
This is the breeding-pair signal. The presence of the species in the wetland is no longer evidence that the wetland is functioning. It is only evidence that the species has not yet finished migrating.
The reckoning question
Whether the reckoning is coming is a question the wetland record answers in unsentimental terms. The reckoning, in every wetland that has ever been drained, is not a single event. It is the slow recognition that the replacement systems collectively cost more, perform worse, and require permanent maintenance budgets the original substrate did not. The U.S. Army Corps of Engineers' annual operations and maintenance expenditure for levee systems, hatcheries, and flood control infrastructure across the Mississippi basin runs into the billions of dollars per year. None of this expenditure existed before the wetland was drained. The wetland was performing the equivalent function for free. The reckoning was not the moment the wetland disappeared. The reckoning is the permanent line item on every state and federal budget thereafter, doing in fragments what the marsh used to do whole.
The internet's reckoning, if the wetland pattern holds, will not be a collapse. It will be a slow recognition that the bolted-on systems are themselves now the load-bearing infrastructure, and that the substrate they sit on top of has degraded to the point where the bolted-on systems are doing nearly all the work the substrate used to do. The 1.58 billion Statements of Reasons are the first audit [4, 5]. The next audits are coming. Each one is more expensive than the last. Each one performs less of the function the substrate performed when it was healthy. Each one requires a separate budget, a separate institutional sponsor, a separate jurisdictional argument about who pays.
There is a word for the political economy of this stage, borrowed from soil science: anthropedogenesis, the formation of new soil layers under sustained human disturbance. It is not a word that should appear in a paragraph about internet infrastructure. It is the word the situation requires.
The anthropedogenic stage of the internet is the stage where the substrate has been so thoroughly altered by the activities running on top of it that the substrate is no longer the medium the original protocols assumed. The DSA does not regulate the internet the protocols designed; it regulates the internet that has emerged from over half a century of unregulated load on a category-blind transport layer. The audits are accumulating because the substrate is no longer self-correcting. Every new platform, every new modality, every new content category arrives at a substrate that has less capacity to absorb it than the substrate that received the platform before it.
The replacement question
The findings that wetland science has accumulated since the early twentieth century contain one consistent observation about replacements. Replacements do not reconstitute the original. The Ramsar Convention's restoration economics, summarized in the Global Wetland Outlook 2025, identify peatland restoration as among the most expensive forms of ecological recovery ever attempted, returning function over decades and only if the surrounding hydrology has not been irreversibly altered [1, 2]. Restoration is not impossible. It is only ever a fraction of the original, returning function in fragments, on timescales that exceed the political horizon of any administration that might fund it.
Every drained substrate produces a generation of replacement architects. Some of them are sincere; the wetland engineers at organizations like Wetlands International and the U.S. Geological Survey's Wetlands Research Center are among the most rigorous applied ecologists working in any field. Their argument is not that the wetland can be rebuilt. Their argument, made in the technical reports, is that the wetland's functions can be partially distributed across designed alternatives, with the explicit acknowledgment that the distributed system will cost more in perpetuity and recover only a fraction of what was lost. The replacement architecture is not an upgrade. It is a permanent adaptation to the absence.
The internet's replacement architects are already at work, and they are doing the same thing. The lifecycle identity researchers, the absence-aware protocol designers, the federation engineers building portable reputation across vetted enclaves, the small archives running on volunteer labor that exist because the open archive cannot be trusted: each is building a single function the unified internet used to perform as a side effect. None of them is rebuilding the unified internet. They are, in aggregate, accepting the wetland argument. The original substrate cannot be restored, and what comes after will be a fragmented landscape of single-purpose replacements that perform less of the function, cost more to operate, and require permanent institutional sponsorship to keep running.
The pattern that does not announce itself
A wetland's collapse is not announced by any single moment. It is announced by the slow arrival of the engineers who specialize in pieces of what the marsh used to do, by the permanent budget lines that did not exist a generation earlier, by the gradual relocation of the breeding pairs to landscapes that turn out, on closer inspection, to be smaller, more expensive, and more dependent on continuous human attention than anyone had planned for. The civilization upstream does not, at any single point, decide that the marsh has been lost. The civilization upstream wakes up, fifty years after the last serious drainage project, into a built environment of ditches and pumps and seawalls and discovers that the marsh is now memory.
The question the wetland record raises is not whether the reckoning is coming. The wetland record records that the reckoning is structural; it has been arriving in slow installments since the load on the substrate began exceeding the substrate's capacity to absorb it. The question the wetland record raises is the one no civilization that has drained a marsh has ever answered well: whether the species that depended on the unified substrate, including the upstream civilization that did not realize it depended on the unified substrate, can survive on the fragmented landscape that succeeds it, or whether each fragment turns out to have been carrying a function nobody priced until it was already gone.
If the substrate that holds everything is the same substrate that produces the things only an unfragmented substrate can produce, what gets carried in the gaps between the replacements that the replacements were never built to hold?