đ§ The Interstitium: What Lies Between
How the Space Between Human Minds Learned to Think
âď¸đ§ They say travel expands the mind⌠this essay has been stuck incubating for about a year now, but suddenly burst into fruit (with the superhuman assistance of Claude Opus 4.6, naturallyâŚ) high in the air somewhere over central Asia between Taiwan and TĂźrkiye. Refined with back-and-forth1 in the airport arrivals concourse and then finished with somewhat jetlagged manual editing and AI illustration in the cafes of the bustling Galata Quarter in Istanbul. A rare burst of clear-minded productivity and AI-enhanced inspiration to finally get these concepts explored in writing! (Clearly I need to be between more oftenâŚ)
Mindfood for the Easter holiday weekend, enjoy!
I. The Space Between
Every thought you have ever received from another human being has been mediated.
This is easy to forget. When you sit across from someone, making eye contact, reading microexpressions, feeling the vibration of their voice in your chest â it feels direct. Brain to brain. But it isnât. Between your grey matter and theirs lies an elaborate chain: motor cortex to larynx to pressure waves to tympanic membrane to auditory nerve to temporal lobe. Even the most âdirectâ human communication passes through layers of encoding, transmission, and decoding. There has always been something in between.
What has changed â what is still changing, with gathering and perhaps irreversible velocity â is the scale, complexity, and now the agency of that in-between layer.
For most of human history, the intermediating substrate was thin: air, parchment, ink. The processing nodes â human brains â were overwhelmingly where the complexity resided. Today, the connective infrastructure between minds dwarfs the minds themselves in raw computational terms. The global internet processes more bits per second than all living human brains combined. The data centres, undersea cables, cell towers, protocols, algorithms, and now large language models that sit between one personâs thought and another personâs comprehension constitute a vast, thickening layer of mediation.
I want to propose a name for this layer, and a framework for understanding how it has evolved and where it is heading.
II. The Interstitium
In 2018, researchers at New York Universityâs School of Medicine announced the identification of a previously overlooked anatomical structure: the interstitium. It is a network of fluid-filled spaces woven through connective tissue throughout the body â surrounding the digestive tract, the lungs, blood vessels, muscles. It had been hiding in plain sight, dismissed as mere âdense connective tissueâ in every anatomy textbook. Once recognised for what it was, it became potentially the largest organ in the human body.
The interstitium does not think. But it connects everything. It transmits signals, transports fluids, cushions organs against mechanical stress. Without it, the bodyâs specialised organs would be isolated, unsupported, and unable to coordinate.
This is the analogy I want to draw. Between human brains â the grey matter of civilisation â there has always been an interstitium: the connective, transmissive, mediating substrate through which thought moves from one mind to another. For a hundred thousand years, this interstitium was almost invisible. A vibration in the air. Marks scratched on clay. Now it is a planetary-scale computational infrastructure, and it is beginning to develop something that looks uncomfortably like agency.
The central claim of this essay is that the interstitium â the total mediating infrastructure between human minds â is undergoing a phase transition. It is shifting from passive conduit to active participant. And the consequences of this shift will reshape communication, cognition, economics, culture, and ultimately what it means to be human.
III. A Brief History of What Lies Between
To understand where things are heading, it helps to understand the trajectory. Each new intermediating technology has altered not just how humans communicate, but what kinds of thought are possible. The medium, as McLuhan insisted, is the message â but more precisely, the medium shapes the space of transmissible and therefore thinkable ideas.
Spoken Language (~100,000 BCE)
The first great intermediating technology was language itself. Before structured speech, primate communication was limited to gesture, facial expression, and vocalisations â rich in emotional content but impoverished in abstraction. Spoken language created a bandwidth explosion. It enabled the transmission of hypotheticals, narratives, plans, and â critically â instructions that could be followed in the speakerâs absence. Language was the original compression algorithm: it reduced the blooming, buzzing confusion of sensory experience into discrete, transmissible tokens.
But it was also radically constrained. Spoken language was ephemeral (no durability), limited to earshot (minimal reach), and one-to-few at best.
Writing (~3200 BCE)
Writing cracked the time barrier. A thought could now persist beyond the lifetime of the thinker. This single capability â durability â transformed everything. It enabled law, bureaucracy, accumulated knowledge, and history itself. Writing also introduced the first significant lossiness trade-off: the richness of vocal intonation, facial expression, and real-time feedback was sacrificed for persistence and portability.
Cuneiform, hieroglyphics, and eventually alphabetic scripts also introduced a new kind of intermediation: the scribe. For millennia, the interstitium required specialised human operators. Literacy was a professional skill, not a universal capacity. The medium itself created a priestly class of intermediaries.
The Printing Press (1440)
Gutenbergâs press did not change the nature of what could be transmitted â it was still text â but it shattered the topology constraint. Writing had been one-to-one (a letter) or one-to-few (a manuscript copied by hand). Print was one-to-thousands, then one-to-millions. The k-factor of ideas exploded. The Reformation, the Scientific Revolution, and the Enlightenment were all downstream consequences of a topological shift in the interstitium.
Legal Contracts and Institutions
An underappreciated class of intermediating technology: the formalised agreement. Contracts, constitutions, charters, and treaties are technologies for binding future behaviour. They sit between human minds not to transmit information, but to constrain and coordinate action over time. They are, in a sense, the interstitiumâs first attempts at autonomous persistence â agreements that endure and exert force independently of the minds that created them.
The Telegraph (1844) and Telephone (1876)
The telegraph collapsed distance. For the first time, information could travel faster than a human being could carry it. The telephone went further, restoring the richness of the human voice â intonation, emotion, hesitation â that writing had stripped away. But both maintained a strict one-to-one topology.
Cinema (1895), Radio (1920s), Television (1950s)
The broadcast era introduced a profoundly asymmetric topology: one-to-millions, with a drastically attenuated return channel. The audience could not talk back in the same medium or at the same bandwidth â but they were never entirely silent. Box office receipts, ratings data, letters to editors, advertising revenue, word-of-mouth reputation, and purchase signals all constituted a trickle of feedback flowing upstream against the torrent of broadcast. The return channel was low-bandwidth, indirect, slow, and took unpredictable routes back to the source. But it existed, and content producers who ignored it did not survive long. The broadcast interstitium was not a one-way valve so much as a massively asymmetric one â a fire hose in one direction, a leaking tap in the other.
This asymmetry shaped the entire 20th centuryâs political, commercial, and cultural landscape. It created the conditions for mass culture, mass persuasion, and the curious phenomenon of millions of people sharing a parasocial relationship with individuals who could not know them in return.
The Internet and World Wide Web (1990s)
The web restored symmetry and created the first truly many-to-many topology. Every node could both transmit and receive at comparable bandwidth. The interstitium became a mesh. Bandwidth increased by orders of magnitude. Latency dropped toward zero for text and, eventually, for audio and video. The web also introduced a new quality: searchability. The accumulated interstitium became navigable.
SMS and Messaging (2000s)
Mobile messaging did something subtle but significant: it made the interstitium ambient. Communication was no longer an event â a letter opened, a phone call answered â but a continuous background hum. The always-on, always-connected pocket device dissolved the boundary between âcommunicatingâ and ânot communicating.â The interstitium became the default state of being.
Social Media (2010s)
Social media platforms introduced algorithmic intermediation: the first time the interstitium began making autonomous decisions about which signals to amplify, suppress, or route. The platformâs recommendation algorithm sits between sender and receiver, and it has its own objective function â engagement, retention, revenue â that is orthogonal to the communicative intent of either party. This was the interstitiumâs first acquisition of something like agency, however primitive.
Videoconferencing and Webcasts (2020s)
The pandemic era normalised high-bandwidth, bidirectional audiovisual communication at scale. But it also revealed the uncanny valley of intermediation: Zoom fatigue, the missing peripheral awareness, the cognitive load of decoding a tiled screen of floating heads. The interstitium was transmitting more data than ever, yet something essential was being lost. The medium was too thin for presence and too thick for efficiency.
Large Language Models and AI Agents (2024â)
And now the present inflection. For the first time, the interstitium can generate. It does not merely transmit, store, amplify, or route human-created signals. It produces novel text, images, code, and reasoning. An AI agent sitting between a human principal and a human counterparty does not function as a pipe or a filter â it participates. It drafts, summarises, negotiates, decides, and acts.
This is the phase transition: the interstitium is no longer connective tissue, it is beginning to think.
IV. The Intermediation Framework
To make this progression rigorous rather than merely narrative, I propose a framework for characterising any intermediating technology along seven dimensions. Together, these dimensions define the memetic propagation potential of a given medium â its capacity to carry ideas between minds and embed them durably in culture.
The Seven Dimensions
The seventh dimension â Agency â is the new variable. For most of human history it was effectively zero. It began to climb with algorithmic content curation. With AI agents, it is approaching values that demand a fundamentally different model.
The Intermediation Matrix
Categorising and quantifying each technology and dimension gives a matrix something like this:
Notes: Bandwidth figures are order-of-magnitude approximations for characteristic use. Lossiness is assessed relative to the full richness of face-to-face human communication. Agency scores are the authorâs estimates based on the degree of autonomous signal modification.
Memetic Propagation Potential
From these dimensions, a composite index can be constructed:
The denominator is significant: lossiness degrades propagation linearly, as expected. But Agency appears as a squared penalty â or, depending on your perspective, a squared modifier. Why squared? Because when the medium itself has agency, it introduces a multiplicative uncertainty into both the content and the routing of the signal. The sender no longer knows what was said, and no longer knows who received it. Each unknown compounds the other.
At low Agency values (~0), the formula collapses to a straightforward propagation model. As Agency approaches 1, the MPP becomes increasingly unreliable as a predictor â which is itself the point. The framework predicts its own breakdown. When the interstitium thinks for itself, the memetic propagation potential of any individual human signal becomes subordinate to the mediumâs own objectives.
This, perhaps, is where things are already heading.
V. The Thickening â A Quantitative Interlude
Consider a rough accounting of where computational âmassâ resides in the global communication system.
The human brain operates at an estimated 10^16 operations per second (by some measures; the precise figure is debated, but the order of magnitude is instructive). With approximately 8 billion human brains, total human computational capacity is roughly 8 Ă 10^25 operations per second.
Global data-centre compute, as of 2025, is estimated at approximately 10^21 FLOPS â still several orders of magnitude below total human brain capacity. But this comparison is misleading in two ways. First, artificial compute is growing exponentially (roughly doubling every 18â24 months) while human brain count grows at ~1% per year. The crossover is a matter of when, not if. Second, and more importantly for this essayâs argument, the interstitial function of artificial compute â its role in mediating between human minds â is already dominant. Most human-to-human communication now passes through digital infrastructure. The interstitium is already thicker than the nodes it connects.
A useful thought experiment: if you were to weigh all the intermediating infrastructure between your brain and every other brain you interact with in a given week â the cables, routers, servers, cell towers, satellites, algorithms, language models â against the biological mass of those brains, the infrastructure would outweigh the wetware by a factor of thousands.
The interstitium was built as plumbing. It has become a mycorrhizal network â grown larger than the forest floor above it, carrying signals the trees did not send, connecting roots that did not know they were connected.
