Sedimentation
Knowledge sediments. It accumulates the way geological layers accumulate — deposited through friction, through pressure, through the specific resistance of the material. Remove the friction and you get dust: present, visible, functional for the moment, but unable to bear weight or form the substrate on which the next layer can be built. The sedimentation metaphor captures what productivity metrics cannot: the difference between output produced through accessed knowledge and capability built through accumulated knowledge. The developer in Lagos who uses AI to build software has accessed knowledge smoothly. Whether that knowledge settles depends on what happens after the output appears on her screen — whether she is embedded in institutional structures that require her to understand what she built, adapt it to her context, defend her decisions, maintain and extend it over time. Sedimentation is what remains when the wind comes.
The Substrate AI Requires — Contrarian ^ Opus
There is a parallel reading that begins from what AI needs rather than what humans lose. Sedimentation presumes the question is whether knowledge settles in people. But the relevant substrate may not be human minds at all. The real accumulation is happening in the training data, the model weights, the institutional memory encoded in vector databases. What matters is whether the organization builds a corpus sophisticated enough to retrieve from, refine, and compound — not whether individual engineers retain architectural patterns they will never again apply without AI assistance.
The Trivandrum engineers do not need to retain what Claude knows; they need to become skilled curators and composers of what Claude can access. This is not dust. It is a different form of sedimentation — one that happens in prompt libraries, in organizational context stores, in the recorded decisions about what the AI built and why. The geologist's metaphor breaks because human memory was never the only possible bedrock. If the wind is AI turnover, what remains is the institutional apparatus for deploying the next model against the last model's outputs. That apparatus — half-human, half-algorithmic — may sediment knowledge more durably than individual human expertise ever did.
In the AI Story
The metaphor has the precision of someone trained as a physicist before becoming an economist. Sediment is a technical concept: particulate matter deposited by water, wind, or ice, compacted over time into layers of rock. The process takes time. It requires resistance — the particles must slow enough to settle. It produces substrate — the foundation on which subsequent deposits can be laid. Remove any of these conditions and you get dust rather than rock. Dust is present but does not accumulate. Dust disperses with the next wind.
Segal reports that the word lodged in his thinking after Trivandrum in a way he could not dislodge. He had measured productivity — twenty-fold, real, verified. But he had not measured sedimentation. When he left Trivandrum, what remained in those engineers beyond the muscle memory of prompting? If Claude disappeared tomorrow, what would the team retain? The features built were artifacts, not knowledge. The architectural patterns used lived in the model, not in their hands. The judgment about what to build came from decades of accumulated experience that the tool could access but not provide.
The metaphor reframes what decisions about AI deployment actually represent. Every decision about how a team uses AI is a decision about what will sediment and what will blow away. Keeping the team at full size rather than cutting to five is an investment in sedimentation — in the tacit knowledge that accumulates when people work together over time, fail together, learn together, develop shared judgment that no subscription provides. Flying to Trivandrum instead of training remotely is an investment in sedimentation — in the transfer that only happens between humans in the same room.
The metaphor does not oppose AI use. It locates the strategic question. The tools are extraordinary. The question is about the institutional conditions under which what the tools enable can settle, compact, and become the substrate for the next layer. Education teaching judgment. Organizations accumulating wisdom. Development strategies investing in embedding rather than only access. The work is not glamorous, not fast, not optimizable. But it is what remains when the wind comes. And in an era when the wind blows harder and faster than ever, what remains is the only thing that matters.
Origin
The sedimentation image recurs across Hidalgo's writing when he discusses how productive knowledge accumulates — a natural metaphor for a physicist thinking about information as structured matter. The metaphor reached wider cultural purchase through Segal's epilogue to the Hidalgo volume, where it became a shorthand for the distinction between output that can be measured and capability that must be cultivated over time.
Key Ideas
Sediment requires friction. Particles settle when their velocity drops below what the medium can carry; knowledge embeds when access is paired with resistance.
Dust is not sediment. Accessed knowledge that passes through without resistance looks productive but does not accumulate into durable capability.
Substrate enables subsequent layers. Knowledge that sediments becomes the foundation on which new knowledge can be built; knowledge that blows away cannot support future growth.
Every deployment decision is a sedimentation decision. How a team uses AI determines what accumulates in the humans and what remains only in the tool.
Time is the irreducible ingredient. Sedimentation cannot be accelerated past a structural limit; the slowness is the mechanism, not the obstacle.
Appears in the Orange Pill Cycle
Sedimentation Happens at Multiple Scales — Arbitrator ^ Opus
The right frame is that sedimentation occurs simultaneously at human, organizational, and algorithmic levels — and the relevant weighting depends on which timescale and which decisions you're examining. For individual skill development, Segal's view is 90% right: without friction, judgment does not form. The developer who never struggles with architectural trade-offs does not develop taste. The contrarian's "skilled curator" frame is real but names something narrower than what "engineering capability" has historically meant.
At the organizational level, the weighting shifts toward 60/40. The contrarian is correct that institutional memory can sediment in non-human substrates — decision logs, context databases, refinement patterns. But these artifacts still require human interpretation to remain generative rather than archaeological. The organization that retains "why we built it this way" in retrievable form has captured something, but not everything. Segal's emphasis on humans working together, failing together, developing shared judgment captures what sediments between people, not just within them.
The synthetic move is to recognize sedimentation as a multi-substrate phenomenon. Different kinds of knowledge settle at different sites. Procedural knowledge may sediment adequately in organizational systems. Judgment sediments in humans. Contextual nuance sediments in the relationship between the two. The question is not whether to permit sedimentation outside human minds, but how to design systems where each substrate receives what it can best hold — and where the friction required for each type of settling is preserved rather than optimized away.
Further reading
- Hidalgo, César. Why Information Grows (2015)
- Borgmann, Albert. Technology and the Character of Contemporary Life (1984)
- Sennett, Richard. The Craftsman (2008)
- Polanyi, Michael. Personal Knowledge (1958)