Information as Substrate: The Memory Beneath Matter, Mind, and Code

Strip the gloss from modern metaphors and what remains is older than computers: the claim that reality is written in pattern, relation, memory, and constraint. Not “data” in the app sense. A deeper grammar. Call it information as substrate and you get a sharp tool: matter as a held regularity, consciousness as a local read-write, culture as a long-lived error-correcting code. This sounds abstract until it isn’t. Crystals, DNA, river deltas, traffic. Everywhere structure that carries forward what can be carried and drops what cannot. The question isn’t whether information rides on matter. It’s whether matter is what stable information looks like, under conservation and noise.

Pattern, Relation, Constraint: How an Informational Ground Becomes World

Wheeler’s “it from bit” gets memed into oblivion, but the sober version isn’t a slogan. It is an inventory. When you look at a system and ask what stays the same as it changes state, you’re inventorying information. Symmetry is information. Conservation is information. Even so-called “redundancy” in gauge theories protects the same content under multiple descriptions. Physics keeps discovering that the invariants do the real work. The less glamorous word for this is constraint. Constraint culls the impossible, leaving a thin corridor of allowed histories. That corridor is not emptiness; it’s compressed description. It’s the substrate.

Biology gives the friendliest example. DNA is not a list of instructions in the cookbook sense, it’s a relational scheme that stores viable ways of persisting in a given environment. Genes interact. They don’t execute in neat serial lines; they play chords. The cell is a machine for keeping constraints intact against noise. Error-correction with a mitochondrion on the side. The organism you see is a front-end for maintaining that descriptive stability, for another round.

Quantum theory makes this more uncomfortable. Take relational views (Rovelli and others): properties aren’t owned by particles the way a key owns its teeth; properties appear in relations. Measurement is an event in which a stable relation asserts itself. The universe, then, is not a warehouse of objects but a web of predictable relations that can be compressed into laws. You can call that web “information” without insulting physics. And yes, time bends under this reading. If the basic currency is relation, then “before” and “after” become local bookkeeping — sequences maintained by observers to keep predictions tractable. The flowing now is our way of staying computationally sane.

This is why simulation talk often misfires. The cinematic frame-by-frame image smuggles in a projector and an operator. Wrong metaphor. If information as substrate is right, then the world doesn’t run like a movie; it holds like a proof. Not rendered, constrained. Some histories are inconsistent, so they never occur. Others are consistent and therefore saturate space with themselves, the way a stable crystal cell repeats. That’s not mysticism. It’s housekeeping.

For a longer exploration in this vein, see Information as substrate.

Consciousness as Local Reception: Self as Compression, Not Origin

If reality keeps score in structure, where does mind sit? Not outside. Not on top like a foreman. More like a local receiver-transmitter with a tight energy budget and a talent for summary. The brain does not store a high-fidelity world; it preserves useful invariants — edges, rhythms, contingencies that mattered yesterday and might tomorrow. Predictive processing is one language for this. You don’t see photons; you see hypotheses that survive updates. The result feels like a world because compression, done right, vanishes into clarity.

On this reading the “self” is another compression, a rolling index over policies and memories that cohere just enough to sign for the bill. Strong illusions can do real work. Agency emerges as a set of stable loops that can be referenced and revised. When those loops go wobbly — injury, disease, ritual, meditation, panic — the felt center moves. Not because a soul leaks out, but because the compression changes its basis. The substrate is still information, but the summary has shifted.

Religion — bracket belief claims for a moment — can be read as humanity’s oldest system for moral memory. Long-term storage of costly norms that pay off across generations, not minutes. Ritual encodes coordination; taboo encodes boundary conditions; story encodes counterfactual rehearsals. You can dislike the metaphysics and still see the engineering: robust error-correction under noise, across time. Cultural selection is brutal. Cheap norms decay. Costly ones, if they bind groups and reduce internal predation, stick. In this sense churches and courts are both archives, one more explicit, the other more embodied. Both guard invariants a community won’t risk losing.

Notice what this does to debates about free will, qualia, and identity. It doesn’t dissolve them into “just information.” It rearranges the stage. Experience becomes a local consequence of sustained relational patterns in a body-world loop. Freedom shows up as the width of the corridor carved by constraints plus skill. Identity as a maintained checksum across memory, behavior, and recognition by others. Not tidy. But operational.

And the odd calm that follows: if consciousness is reception with feedback privileges, then meaning isn’t injected into a dead world; it’s negotiated within a living grammar of relations. A little less king-of-the-hill. A little more steward of summaries that can survive contact with tomorrow.

Machines on an Informational Ground: Design Heuristics, Not Hype

Building AI on top of this substrate metaphor forces two questions. What are machines stabilizing? And what social memory do they preserve or erode? Current practice answers the first with “loss on a benchmark” and the second with silence plus a compliance PDF. But if systems are, at base, engines for maintaining and transforming constraints, then putting them into hospitals, courts, schools is not about IQ as spectacle. It’s about whether their learned constraints line up with the constraints we can live with.

Moral “patching” — post hoc filters to calm regulators — treats ethics as a surface spray. Spray washes off. If moral memory is slow, distributed, and expensive to build (because it encodes fragile invariants of cooperation), then the pragmatic path is constraint-first design. Bake invariants down into the loss, the architecture, and the interface. Examples, sketched not sold: a triage model that cannot recommend a plan unless it also surfaces a human-readable explanation linking clinical constraints (contraindications, scarcity, duty-of-care priors) to its choice, with signed accountability that survives organizational churn. Or a school placement system that treats community cohesion and student agency as hard constraints and test scores as soft ones, logging every override in a tamper-evident ledger so memory can do its long job. Not blockchain worship. Just memory that resists incentive rot.

Two heuristics help. First: relation over object. Treat features not as independent nuggets but as conditional structures; build training that preserves relationships (causal scaffolds where known, counterfactual rehearsal where not). Second: time is local. Systems should expose their own sequencing assumptions. If a model’s output depends heavily on the order in which it saw data — and many do — that’s not a nuisance; it’s a property. Surface it. Let institutions decide whether that locality is acceptable. Hidden sequence equals hidden power.

Open methods matter here. Not because transparency is holy, but because secrecy is a solvent that dissolves accountable information. When the informational ground is the thing, the public has a right to interrogate how their shared constraints are being re-written. Corporate governance that treats deployment as a series of PR-safe patches misunderstands the medium. You can’t outsource cultural error-correction to a crisis comms team.

None of this requires anti-tech piety. We can build machines that stabilize the right contours. That remember what must be remembered and forget what can be forgotten. But it refuses the fantasy that a smarter predictor equals a wiser system. Wisdom, in this frame, is disciplined forgetting anchored to collective invariants. Which we barely agree on. Which have to be renegotiated in view of new constraints (climate, migration, resource shifts). Hard, slow work.

So the open question lingers, properly. If information as substrate is our ground note, which constraints do we swear to keep, and who gets the pen when the ledger needs a line erased and written again?