Information is not neutral. It behaves like a mathematical constant: what the system receives determines what the system returns. Not as metaphor, not as narrative — as physics. Bateson formulated this with a precision time has not eroded: information is a difference that makes a difference. Without difference, there is no information. Without information with density, the system processes noise and returns noise.
The contemporary crisis of the digital ecosystem is not a passing cultural phenomenon, nor the result of misguided decisions by Silicon Valley executives. It is the mathematically predictable output of a decade of massive low epistemic-density input. The period 2012–2025 functioned as an uncontrolled experiment whose initial conditions — the algorithmic democratization of digital reach without epistemic hierarchization — produced chaotically amplified effects, compounding exponentially. Lorenz described precisely this mechanism in 1963 while studying atmospheric systems. What we observe today in search indices and distribution feeds is the same principle applied at industrial scale.
Meta: converting visibility into commodity
The Meta case is the most thoroughly documented historical record of how a platform transforms its internal logic from social network to advertising engine. The data are precise and verifiable. The average organic reach of Facebook pages fell from approximately 16% in 2012 to 1–2% in 2025. Over the same period, the company's advertising revenue grew from $4.28 billion to $196 billion, representing 97–98% of total revenue.
Precision matters here: the data show a strong temporal and structural correlation between the sustained reduction of organic reach and the parallel growth of advertising revenue. Although Meta consistently framed these changes as improvements to user experience — prioritizing "meaningful interactions" — the net result was a massive migration of visibility toward paid content. Free visibility ceased to be a distributed good; it became a control variable.
The influencer did not die with this shift. It professionalized. The global influencer marketing market grew from $1.7 billion in 2016 to $32.55 billion in 2025. Both curves — falling organic reach and rising influencer market — tell the same story from two perspectives. What Meta eliminated as a free good, the market restored as a transactional one. Visibility did not disappear: it was privatized.
Google trapped in its own structural contradiction
Google faces an unprecedented evolutionary trap. To complete its transition from search engine to AI-powered answer engine, it must actively reduce the noise generated by the very ecosystem of massive low epistemic-density content it helped finance and scale for over a decade. This is not a free strategic choice — it is a structural contradiction the numbers make irrefutable.
In 2025, advertising revenue represented 59% of Alphabet's total revenue — $237 billion of $402.8 billion — reflecting the accelerated growth of Google Cloud. Alphabet projects spending between $180 and $190 billion in capital expenditure during 2026, nearly double the $91.45 billion of 2025. Feeding AI models with massive volumes of low epistemic-density content is not merely inefficient — it is thermodynamically unsustainable. A traditional search query consumes 0.3 Wh. An extended-reasoning Gemini query can multiply that cost by a factor of 10 to 1,000.
SpamBrain and the successive iterations of the Helpful Content Update are not user experience improvement tools. They are index-purging mechanisms driven by real economic and computational necessity. Google cannot kill the influencer without killing part of its advertising model. But neither can it feed Gemini with the content that influencer produces. This is the trap: to save the organism, it must progressively eliminate the parasite it raised.
TikTok and the bifurcation no one anticipated
TikTok did not solve the digital ecosystem problem. It reformulated it. Rather than eliminating organic reach, it redefined which signals count toward determining it. The platform shifted weight away from likes — an easily manipulable and socially pressured signal — toward saves and shares, which indicate real utility and intent to return. The completion rate required to trigger viral distribution rose from 50% in 2024 to 70% in 2026. The algorithm is attempting to separate real attention from fabricated attention.
The human response to this new environment has been a bifurcation no platform engineer clearly anticipated. On one side: artificial sophistication — more elaborate scripts, more polished production, more constructed narratives. The problem is that this content has acquired a recognizable texture — it smells like GPT because in many cases it is. On the other: authentic primitivism — rawer, more physical, more instinctive content, as an unconscious response to the excess of synthetic production.
The influencer is not, as has been argued superficially, a trend heading toward its end. The influencer is an anthropological constant: it existed in every culture as the shaman, the orator, the jongleur, the chronicler. What the digital ecosystem did was collapse the barriers of access to relevance. What is observable is that the system's incentives have changed. Every click on "not interested" is a signal that the algorithm failed. The ecosystem is training itself to predict with greater precision — and in that training, volume without difference loses ground.
The allostasis of knowledge
The contemporary crisis of the open web should not be read as a cultural failure, but as an informational thermodynamic collapse. Faced with the entropic dispersion of the mass ecosystem, the survival of knowledge did not occur in the crowd — it occurred in the transition toward structures operating under the negentropy Schrödinger described in 1944. Contemporary dense communities function as open systems that import informational order to resist the thermal death of the digital environment. There is no metaphor here; there is information physics.
Scientific knowledge did not disappear during the influencer decade. It underwent what can be described as systemic allostasis: the maintenance of viability through a change of route, not of objective. In Varela and Maturana's terms, the living system does not alter its organization — it alters the means by which it maintains it. Zenodo, OSF, arXiv, closed Discord communities, niche newsletters, preprint repositories: these are not phenomena of conscious cultural resistance. They are the predictable adaptive output of a knowledge production system that found mass channels had ceased to be functionally useful.
The apparent exception deserves attention: channels like Kurzgesagt, Veritasium, or SciShow reached millions of followers communicating scientific content through mass platforms. These channels do not contradict the thesis — they confirm it by contrast. Their survival in the mass ecosystem required adapting knowledge to the algorithm's format: specific durations, high-cost production, dramatic narratives. They demonstrate exactly how far knowledge had to transform itself to survive in that ecosystem. They operated as negentropic anomalies — injecting real density into a sea of low signal — but at an adaptation cost most scientific knowledge could not afford.
The diet of artificial intelligence
Artificial intelligence does not produce its own content. It consumes differences that make differences. A language model trained on massive volumes of homogeneous content does not acquire reasoning capacity; it acquires statistical reproduction capacity. The difference between the two is not one of degree — it is one of nature.
The extended sandboxes characterizing large language model development — periods of months or years in which the system does not update its knowledge base — are not temporary technical limitations. They are a direct consequence of the signal problem: when the volume of content available for training grows faster than the system's capacity to distinguish signal from noise, the system prefers to operate on what it already has.
This selectivity has a consequence the analysis allows formulating with precision: the future of the digital ecosystem is not AI replacing the content creator. It is AI forming itself alongside those who produced hard real data — verifiable trajectories, operational knowledge derived from documented experience, processed and recorded errors. Not polished scripts. Not flawless edits generated in minutes. The difference the veterinary physician accumulated across 34 FAVN serology cases. The difference the engineer built calibrating hydraulic brakes under real high-altitude conditions. The difference the entrepreneur documented operating three businesses simultaneously under real pressure.
The influencer who explained how to do something in front of a camera, with background music and high-cost editing, did not produce that kind of difference. It produced representation of difference. Epistemic density cannot be simulated statistically. It accumulates.
Positive disintegration at systemic scale
The present times are deferred products of precise initial conditions. The credibility crisis of Google, the bifurcation of the digital ecosystem, the migration of knowledge toward dense low-visibility structures, the thermodynamic pressure forcing mass platforms to purge their own ecosystem — all of this is the predictable output of a decade of massive low epistemic-density input. There are no guilty actors in this analysis. There is physics.
Meta did not reduce organic reach out of malice — it did so because its model required converting visibility into commodity. Google does not face its structural contradiction through design error — it faces it because it optimized for short-term metrics that proved incompatible with its long-term evolution. The influencer did not contaminate the index by intention — it did so because it learned to give the algorithm exactly what the algorithm requested. Every actor functioned rationally within its incentives. The problem is systemic, not moral.
Prigogine demonstrated that complex systems far from equilibrium do not necessarily collapse — they reorganize at higher levels of complexity through what he called dissipative structures. The digital ecosystem, in its current perturbation, is consistent with that model: the disintegration of the previous order is not the end of the system but the condition for its reorganization. The thermodynamic pressure making the massive low-density content model unsustainable is, simultaneously, the force producing the bifurcation toward denser and more sustainable structures.
Those who produced real difference during the decade of noise will not need to explain their trajectory. The trajectory will be the explanation.
