You are inside the thing you are measuring.

Wherever you stand, you are at the centre

What can an observer reach, from inside the world it measures?

Not because you are special. Because here is where you measure from. Every observer finds the same thing, and that is exactly what makes it a law rather than a conceit: different observers read different raw values for the same state, and all of them agree on the distances.

The rule this sets up is a discipline, not a slogan: a quantity the observer cannot reach cannot appear in a law the observer is supposed to use. A true underlying value, an absolute rate, a variable nobody can intervene on — when one of those turns up in a model, a view from nowhere has been smuggled in.

Almost everything has to be thrown away

What must be kept, so the rest of the past can be forgotten safely?

A state is a summary of the past that is sufficient for the future — and the test is sharper than it sounds. A summary that predicts the future when you leave the system alone, but not when you push it, is a description, not a state.

This is where most measurement goes wrong, and it goes wrong the same way every time:

A measurement merges histories that have different futures.

See this state question in a concrete, conditional radiation model: two exposure orders and one common future challenge.

Which pushes behave the same way from everywhere

Which interventions act the same way from every state?

A state only means something relative to a set of interventions, so the set has to be named — and not everything qualifies. The gate: an intervention's endpoint from rest must fix its action from every state. That is testable in an afternoon.

Here is the version that matters in a laboratory. Two drugs, each with an effect. What is the effect of both? If the combination law is projective, the answer lies on a one-parameter family — a dial with named landmarks.

And here is where this programme lost a bet

The dated local analysis plan predicted: drugs sharing a mechanism should sit near the Loewe end; independent mechanisms near Bliss. The repository does not have an independent timestamp for that plan. The test used 210 dose-response blocks over 36 drug pairs.

Refuted. ρ = −0.12, p = 0.76 — the wrong sign, and nowhere near significance. Same-axis pairs were more synergistic than distinct-mechanism pairs. An external reviewer reran it and reproduced the failure to five decimals.

Part of the reason is structural and worth knowing: Loewe additivity only sits on the dial when the dose-response curves have Hill slope 1 and full efficacy. At slope 2 the true Loewe surface is at α = −3.5; at slope 3 it leaves the dial entirely. In that dataset the median slopes were 2.0 and 1.3. The prediction, as stated, was only ever right for a minority of real drug pairs.

It stays in the record, with its local plan, code and data. The atlas has the row.

The law, and the edge it cannot cross

How do bounded changes combine, and can the edge ever be reached?

Take any real quantity with a ceiling, a neutral state, and a lawful way of combining changes — where lawful means continuous, associative, and strictly increasing. Then there is a coordinate, its rapidity, in which combining is adding. And in that coordinate the ceiling sits at infinity.

How many such laws are there? Three. Not three that people happen to use — three that are possible.

So why do real things hit their limits?

Populations do go extinct. Genes do fix. Cells do commit. If the theorem says the edge is unreachable, something must give — and exactly one measurable number decides which.

And the horizon is not the end of the space. Push past it and the law is still there, acting on the other side — with the value reciprocated and the rapidity turned by a quarter. In spacetime this is not a metaphor: inside are observers' velocities, outside are the slopes of their lines of simultaneity, and one law moves both.

The room inside is bigger than the room outside

What shape is the room inside the horizon?

In one dimension, the rapidity makes the interval infinitely long and that is all. In two or more, something else happens — and this is the part that changes how the world looks.

This has been measured, by other people, in places nobody was looking for geometry: the structure of complex networks, the space of natural smells, the rat hippocampus, and the branching lineages of single cells. Branching histories need this room, and flat space of fixed dimension does not have it.

The second thing that happens in two dimensions is stranger. The order in which changes are made stops being forgettable.

The geometry is not a matter of taste. It is decided by the algebra of the observer's channels — and by one distinction that is easy to miss: whether an update can be undone.

The same mathematics describes two observers coming into agreement. Couple two rapidities and the shape of the coupling decides everything: an unbounded coupling always locks, a bounded one locks only while the drive is small enough, and a periodic one slips.

Does any of it win on data it has never seen?

The only gate that can go wrong in a way the mathematics cannot repair.

The cited mathematical statements have their stated assumptions. Whether they describe a particular measured system remains a separate empirical question. Here is the instrument built to test one application — the same six gates running on a living mind.

The scoreboard, as it stands

Every one of those has a stated test and a stated loss condition — the observation that would make it wrong. A claim without one is not a hypothesis, and it does not get a row.

Open the atlas Read the papers