# When the Loop Must Wait

When should a loop keep interpreting what it has, and when must it look again?

The question is sharper than whether the sensor is engaged. A living sensor can return to the same record with patience, intelligence and real stakes, yet receive no new evidence capable of correcting the claim. The return may still matter. An old trace can yield an implication no one saw the first time. It can acquire meaning through a later theory or a changed life. But reinterpretation of a record is not an independent confirmation of that record, and human presence does not make it one.

Galileo's observations show both sides of the problem. When he first saw small points near Jupiter, he took them for fixed stars. On later nights the points remained near Jupiter while changing sides and mutual positions in ways inconsistent with that interpretation and the expected motion of Jupiter. The sequence supported an account in which they moved around the planet. [Galileo's text](https://bertie.ccsu.edu/naturesci/Cosmology/GalileoStarry.html) records his initial doubt about Jupiter's motion as well as the later orbital conclusion. Several nights supplied a pattern one view could not, without making the observation an interpretation free of assumptions.

Saturn did not yield so readily. Galileo saw a central body with two lateral forms and described it as triple-bodied. Later appearances changed, but he did not identify the ring. Limited resolution left more than one account compatible with the appearances. Huygens subsequently brought improved optics and a geometrical explanation to the successive views. His [1659 account](https://mathshistory.st-andrews.ac.uk/Extras/Huygens_Saturn/) describes both the labor of making lenses and the reasoning behind a ring hypothesis he had already announced in encoded form in 1656. Time supplied differences; an account of how those differences were produced made them intelligible.

The Accademia del Cimento's 1660 model experiment went further: viewing an artificial ringed Saturn through different telescopes showed how a ring could produce a triple appearance through weaker optics. The [Museo Galileo account](https://catalogue.museogalileo.it/multimedia/SaturnsSystemBis.html) supports that optical possibility, not a unique proof of the real planet's structure. The instrument's way of showing had itself become something to investigate.

## What time can add

Waiting has a specified evidential purpose when there is an account of what the interval may make available or change. Sometimes a signal or an intervention outcome already exists but has not reached the observer. Propagation delay sets the earliest receipt time; waiting through it does not change the content of a fixed record. At other times the target, viewing geometry or experimental condition will evolve so that hypotheses which agree now predict different observations later. That is evidential waiting. Anticipated divergence justifies it as a strategy; unanticipated changes can also make waiting useful in retrospect.

A distant source can therefore provide a newly received view of an old state. The source's time, the record's arrival and the moment of recognition need not coincide. An archive can carry that record into a later living encounter; preserving its origin and instrument conditions allows the later reader to interrogate what it can support. Greater distance does not by itself determine how long that inquiry will take.

This gives the loop more than one possible next move. It should interpret or reanalyze when the existing record already bears differently on named alternatives and the unresolved work is to draw out that difference. It should obtain another measurement when a feasible acquisition is expected to separate alternatives that the present record leaves together. It should wait when a specified delay or changing condition is expected to make such evidence available. It should calibrate when uncertainty about the instrument's own transformation prevents an observation from identifying the target. When none of these moves is feasible, the disciplined result is *unresolved*. A living loop preserves the possibility of correction; it does not promise that every question will close.

For this rule the alternatives must be named, because discrimination is conditional. Evidence can favor one member of a proposed set without proving the set exhaustive; two separated models can both remain wrong. Exploratory observation can also reveal alternatives no one had formulated. Recognition in the framework's lived sense includes what the finding means, how it connects, and how it changes the sensor. But its depth cannot repair a measurement that leaves the relevant alternatives observationally identical.

## A minimal observation horizon

Consider two possible two-pixel objects, one illuminated on the left and one on the right:

$$
x_L=(1,0), \qquad x_R=(0,1).
$$

Suppose the instrument alternates between two scheduled views. At every even epoch it records the sum,

$$
A_{2k}=[1,1],
$$

and at every odd epoch it records the difference,

$$
A_{2k+1}=[1,-1].
$$

The sum view gives the same result for both objects:

$$
A_{2k}x_L=A_{2k}x_R=1.
$$

No amount of noiseless repetition at even epochs separates left from right. A replay of an archived sum observation is the same trace again, not a new draw. Even a newly recorded sum is fresh data but redundant for this claim. The difference view does separate them:

$$
A_{2k+1}x_L=1, \qquad A_{2k+1}x_R=-1.
$$

Here waiting from an even to an odd epoch has an explicit evidential purpose: the acquisition operator changes. This engineered, noiseless schedule is not an ephemeris or a reconstruction of Galileo's telescope. Repeated measurements with independent noise could improve precision; that is a different case. The [model](observation-horizon/model.py), [results](observation-horizon/results.json), and [evidence ledger](observation-horizon/evidence-ledger.md) make the assumptions inspectable.

One ambiguity remains. If the instrument's left and right channels may have been swapped, the difference view cannot distinguish a left object with ordinary wiring from a right object with swapped wiring. Both pairs produce the same observations at every epoch. A known left-hand reference measured through the difference channel resolves the swap, assuming the wiring stays fixed. This illustrates one part of the Saturn problem: the next requirement may concern the instrument model before it concerns the object.

This distinction exposes a precise open edge in [the Veer formalism](/appendices/the-veer). Its limit-case discussion says that when a sensor is present, every free-energy reduction is territorially grounded. Its honest edge then treats sensor presence as the experimental proxy for a positive territorial channel. The cases above split those claims. A human can be fully present while attending to a record that contains no claim-specific corrective information. An automated detector can deliver a discriminating measurement without becoming a living sensor. The detector supplies contact; the living sensor recognizes what that contact bears on and what it means.

The framework holds that truth as lived recognition requires circulation through a being capable of experience and change. These equations neither establish that claim nor replace recognition with discrimination. They press a narrower requirement: identify how evidence could correct this particular claim. The proposed discipline is to preserve that possibility across the time and mediation the claim requires. An interval without an answer need not end the inquiry. But neither the passage of time nor the depth of engagement supplies the observation that is missing.