Reading the ground: information as substrate¶

Substrate is a word for the conditions that constrain a choice without making it: geography, climate, technology, institutions, population density, and the cost of the alternatives. People choose within those conditions and rarely get to choose the conditions. Read a history of organised violence that way, as War follows the ground does, and the rival theories stop competing. Each names a mechanism that dominated wherever the ground made it the decisive one.

The same reading works on information, and it is less comfortable than the historical version, because the substrate in question is still being laid.

What made a life stealable¶

A forager’s possessions could be carried, and the owner could walk away. A harvest could not. It sat in one place, could be counted, taxed, hoarded and carried off, and it came back every season. Agriculture did not invent the idea of taking things. It made taking reliably worth the effort, year-on-year, and institutions grew up around the return.

Behaviour used to be like the forager’s possessions. It happened and was gone, remembered imperfectly by whoever was present, unavailable to anyone who was not. What changed is not that people became more interesting. Ordinary activity acquired the properties of grain: fixed in a record rather than in a memory, countable, storable, and comparable against everyone else’s. A journey to work becomes a row with a timestamp. A hesitation before a purchase becomes a measurement. None of it was worth taking while it stayed unrecorded, and all of it is worth taking once it does not.

The harvest metaphor breaks in one direction, and the break favours the harvester. Grain removed from a store is gone from it. A record copied is still held by everyone who had it, which means exposure accumulates and rarely reverses. Something disclosed in one context stays available to be read in another, years later, by a party who was not in the room and had no interest in the original moment.

Proximity without transparency¶

Some conflict needs no aggressor. Two neighbours, close enough to threaten each other and not transparent enough to reassure, can manufacture a quarrel out of mutual uncertainty alone. Nothing material changes hands. What moves is only what each side believes about the other, which makes it a substrate built almost entirely of information.

The arrangement around a person’s data has the proximity and none of the transparency, and the opacity runs one way. A broker can hold a detailed profile of someone who cannot name the broker. An inference drawn about a household is acted on by parties the household never contracted with, through a chain of onward sales that no participant is obliged to render legible. The asymmetry is not incidental to the system; it is close to being its operating principle, since a market in inference works best when the inferred party cannot audit the inference.

Complexity does defensive work here that secrecy could not do openly. A regime that simply refused to say what it held would invite a fight. A regime that discloses everything, in terms of service nobody finishes and data flows no single participant can map end to end, is technically transparent and practically opaque, and it gets the benefit of both.

Where the exits went¶

Conflict only accumulates power when the losers cannot leave. In open country a defeated group walks somewhere else and the victory evaporates. In a valley hemmed by desert, defeat means incorporation, because there is nowhere to walk to. The feature doing the work in that account is not the fertile valley. It is the desert around it.

Digital life has acquired its own deserts, and they were mostly built for other reasons. Access to a bank, a tax authority, a school, a landlord, a doctor’s appointment or an employer increasingly runs through infrastructure with no offline equivalent still functioning. Where a service is the condition of participating in ordinary administration, refusing it is not an option that a person can price and decline. Network effects close a second set of exits: leaving a platform where everyone is means leaving the people, not the software. Consent, under those conditions, describes a signature rather than a choice.

This is why exhortations to individual responsibility land badly. Being told to read the terms, or to simply use something else, presumes an open landscape. Some of the ground is open, and the practical gains there are real. Much of it is a valley.

Ways to walk off¶

Walking off is still worth attempting, provided the exits get sorted honestly into the ones that change a person’s position and the ones that mainly produce a feeling of having acted.

Some exits are real and cheap. Where an alternative genuinely exists and switching costs little, leaving works exactly as advertised: a different browser, a different search engine, a messaging app that cannot read the messages. Nothing structural changes, but the surface a person presents gets smaller, and a smaller surface is worth having even where exposure cannot be removed. Green’s strategies for people sequence that work.

Some exits are partial and still count. Where departure is impossible, the remaining move is to raise the cost and lower the yield of collection: separating identities so that records do not join, paying in ways that do not resolve to a name, correcting and bounding what is already held. None of this leaves the valley. It makes the ground harder to farm, which is a different and more modest claim than escape.

Some exits are collective rather than individual, and these are the ones that alter the substrate itself. Interoperability requirements, data portability with real teeth, procurement rules, enforcement budgets, and public alternatives to private infrastructure all work by reopening a route out rather than by asking a person to be more careful. Whether that route reopens is argued at systems effects and, where states are the actor, nations and states.

And some exits are theatre. Cookie banners that record a preference and change nothing downstream, opt-outs applied per profile in a market that regenerates profiles, deletion that removes a customer record and leaves the derived inferences intact. Recognising these as scenery rather than doors is not cynicism. It prevents effort from draining into the places designed to absorb it.

The stated reason and the actual ground¶

Ideology dominates how conflicts get narrated, which makes it easy to assume it explains why they happen. The narrative layer turns over every few decades; the substrate changes far more slowly. Gods, crowns, and nations gave way to each other while arable land stayed worth fighting over for ten thousand years.

The narratives around personal information turn over at a similar pace. Personalisation, relevance, security, convenience, safety, and lately the training of models: each has been the stated reason in its turn, and each will be replaced. Underneath, the contested thing has not changed much. It is whether ordinary life continues to deposit a durable, joinable record, who holds that record, and what it costs to be somewhere the depositing does not happen.

Reading any substrate is the same move. The narrative layer records what people say. The substrate records what constrains them. The two rarely agree, and the gap between them is usually where the argument actually is.

Last reviewed: 2026-07-18.