The body solved this three billion years ago.

Every cell in your body contains the same DNA. Every one of them. The cell in your retina and the cell in your heart and the cell in the skin on the back of your hand all carry an identical copy of the same complete genetic instruction set. And yet they are profoundly, functionally different from each other. The retinal cell does things the heart cell cannot. The heart cell does things the skin cell has no mechanism to perform. Same DNA. Completely different behaviour. Completely different identity.

The difference is not in the DNA. It is in which parts of the DNA are switched on and which are switched off. Sitting on top of the genetic sequence, there is a second layer of structure. Molecular switches. Proteins that block a gene from expressing itself, or open it to produce the proteins that give a cell its function, its identity, its purpose. This second layer is called the epigenome. It is, in the most precise sense, the context layer. The same underlying information, read differently depending on which switches are active, produces an eye or a heart or a lung.

The complexity of what is happening inside a single cell at any given moment is beyond ordinary comprehension. If you were to scale a single cell to the size of Manhattan, every person in that city would represent one protein. Ten billion of them. Moving between five-hundred-storey buildings, building things together, breaking things apart, running molecular machinery that never stops, never sleeps, never pauses. That is one cell. One second of its operation. Happening simultaneously in thirty-seven trillion cells in your body right now.

And it works. For decades it works. Because the epigenetic switches stay in the right positions, the right genes express the right proteins, and every cell knows what it is and what it is supposed to do.

Then it starts to go wrong.

DNA sustains damage constantly. Radiation. Sunlight. Oxidative stress. The ordinary friction of being alive. The cell is extraordinarily good at repairing that damage. Dedicated repair proteins move in, fix the break, restore the sequence. But every repair carries a small risk. The epigenetic markers, the switches that determine which genes are on and which are off, can be displaced slightly in the process. Reset to the wrong position. Over years, over decades, across millions of repair events, the drift accumulates. Genes that should be active go quiet. Genes that should be silent begin to express. The cell does not lose its DNA. It loses its ability to read its own DNA correctly. It loses its context.

This is ageing. Not damage to the underlying information. Drift in the layer that governs how that information is interpreted. The heart cell starts misfiring. The retinal cell stops regenerating correctly. The skin cell loses the protein production that kept it elastic and resilient. The underlying instructions are still there, intact, unchanged. The context layer that told the cell how to read them has been corrupted by accumulated drift, one small displacement at a time, invisible at the individual repair event and catastrophic at the population level across a lifetime.

Shinya Yamanaka

In 2006, a scientist named Shinya Yamanaka discovered something that reframed the entire question of what ageing is and whether it is reversible. He found four proteins that, introduced into a cell, would move all of those displaced epigenetic markers back to their original positions — resetting the cell completely, returning it to a stem cell state from which it could become any cell type in the body. He won the Nobel Prize for it.

What followed was the discovery that changed everything. Introduced in smaller amounts, those same proteins did not reset the cell to a blank stem state. They moved the markers back to where they were supposed to be for that specific cell type. Not a reset to zero. A restoration of the original context. The retinal cell became a young retinal cell again. The heart cell recovered its correct electrical behaviour. The skin cell resumed the protein production it had lost. In animal models, blindness was reversed. Ageing markers were rolled back. The underlying DNA had not changed by a single base pair. Only the context layer had been restored.

The body had been carrying the information needed to function correctly the entire time. What it had lost was the ability to read that information in the right way. And what Yamanaka's factors did — what the entire field of epigenetic reprogramming is now building on — was restore the record of what the correct reading was supposed to look like.

Now. Consider what you just read.

Not as biology. As an infrastructure problem.

Every regulated institution accumulating consequential decisions over time is running a version of the same system. An underlying record, transactions, relationships, risk assessments, model outputs, human judgements, that is complete, intact, and in principle available. And a context layer sitting on top of it that governs how that record is read, what it means, how it connects to everything else the institution knows.

And like the epigenome, that context layer drifts. The policy version in force when a treaty was priced is no longer the version in the system. The model that produced a credit recommendation has been through three updates since the facility was approved. The trust level assigned to a data source at ingestion has never been reviewed, and the source has changed materially in the intervening years. The cedant relationship that informed an underwriting judgement was understood differently by the two people in the room, and neither interpretation was captured.

The underlying data still exists. The context that gave it meaning at the moment the decision was made has degraded, drifted, and in most cases evaporated entirely. The institution does not have ageing cells. It has ageing decisions. And the consequence is identical. A system that was functioning correctly begins to produce outputs that no longer reflect the integrity of the information underneath them. Not because the information changed. Because the context layer that governed how it was read was never built to hold its position over time.

This is epistemological drift at the institutional level. The same phenomenon Hop 5 introduced as the accumulation of low-trust inputs, stale context, and miscalibrated model confidence across a decision population — visible only when you have infrastructure that captured the epistemological status of every node at the moment it was true. Invisible, and compounding silently, when you do not.

Replay

Replay is the Yamanaka factor for institutional decisions Yamanaka's insight was not that the DNA was wrong. It was that the context layer governing how the DNA was read had drifted from its correct state, and that drift was reversible if you had a record of what the correct state looked like.

Replay is the same insight applied to institutional decision infrastructure.

When the graph captures the context boundary, the permission scope, the model version, the confidence scores, and the human judgement at the moment of every decision, not reconstructed afterward, captured at the time, it is preserving the epigenetic state of that decision. The exact configuration of the context layer at the moment the decision was correctly made. The markers in their correct positions. The switches in the right state.

When a claim is disputed eight months later, Replay does not reconstruct what happened. It restores the original context. Not an approximation. Not a reconstruction from logs and memory. The exact conditions. What was in scope. What was not. What the model saw. What confidence it returned. What the human understood. What they decided. The context layer, intact, at the moment it was true.

The body's repair mechanism introduces drift as an unavoidable side effect of fixing damage. Every institution's operational reality introduces drift as an unavoidable side effect of operating at scale over time. Models update. Policies change. Data sources evolve. People leave and take their contextual understanding with them. The underlying record accumulates. The context layer that governed how it was read at any given moment degrades.

The difference between an institution that can govern this and one that cannot is not the quality of its data. It is whether the infrastructure underneath its decisions was built to preserve the context layer at the moment of capture, and restore it at any subsequent point with the same fidelity that Yamanaka's factors restore a cell to its correct epigenetic state.

Most institutions are not built for this. They capture outcomes. They do not capture the context layer that produced them. And like a body whose epigenetic repair mechanisms have been running without proper oversight for decades, the drift is silent, cumulative, and by the time it becomes visible, it has been compounding for years.

The repair mechanism has to be native

Here is the part of the biology that matters most for the infrastructure argument.

The body's epigenetic repair capability is not a feature added on top of the cell. It is a native property of the cell's architecture. The repair proteins are always present. The mechanism for reading the epigenetic state and correcting drift is built into the system at the foundational level. It does not require an external intervention to activate. It is running continuously, as a consequence of how the cell was designed.

Retrofit does not work here. You cannot add epigenetic repair capability to a cell that was not built to support it. The machinery has to be there from the start. And when it is, the cell has something that fundamentally changes its relationship to time and damage. Not immunity to drift. The ability to correct drift before it becomes irreversible.

The same is true of institutional decision infrastructure.

Adding a logging pipeline to a system that was not built to capture context does not give you Replay. It gives you a record of outcomes that lacks the context layer those outcomes depended on. Adding a documentation workflow to a model deployment that was not built to attach provenance at inference time does not give you the ability to restore the epistemological state of a decision made eighteen months ago. It gives you a timestamp and a model name, neither of which survives the scrutiny of a regulatory review or a coverage dispute.

The context capture has to be native. The permission enforcement has to be native. The provenance attachment has to happen at inference time, not reconstructed afterward. Not because these are difficult things to do in principle, but because they are only trustworthy when they are properties of the architecture rather than layers applied on top of it. A context layer that was captured natively is a fact. A context layer that was reconstructed afterward is a claim. In regulated environments, the difference between those two things is not recoverable once the moment of the decision has passed.

The body figured this out over three billion years of iteration. The cell does not wait for damage to become catastrophic before it repairs. It does not reconstruct its epigenetic state from memory after the markers have drifted. It captures and maintains the correct state continuously, as a native property of its architecture, because that is the only mechanism that works.

The institutions building decision infrastructure now have the same choice. Build the capture natively, at the infrastructure level, from the start. Or accumulate epistemological drift silently, at scale, until the moment arrives when they are required to restore a context they never preserved.

What comes next Hop 6 is the biological argument. The epigenome as the original context layer. Epigenetic drift as the institutional equivalent of accumulated decision debt. Replay as the Yamanaka factor for institutional decisions — not a reconstruction, a restoration.

Hop 7 goes to the boundary between institutions. A reinsurer and a cedant. A syndicate and its follow market. A pharmaceutical company and its regulatory authority. Consequential decisions that cross institutional lines, governed by provenance chains that cannot stop at either institution's perimeter. The infrastructure question that nobody has answered yet, and why getting it wrong at the boundary is where the entire governance argument collapses.

The science behind the argument The biology in this piece is real. This conversation between scientists is where the cellular repair analogy originates. Worth twenty minutes of your time.