---
id: narrative.act-i
through_object: rods, binary, root request
claims: [C1, C2, C3, C4, C6, C7, C8, C9, C10, C11]
---

# ACT I

Through-object: the radiation rods, the two stars, the queued request.

Act brief. A ship older than the present shape of the galaxy finds a two-star system whose future cannot be closed by any archived stellar sequence. Avoiding the future jet is cheap if done now. Representing the system as a recoverable minimum is not. The request remains after the ship is safe.

## 1. THE SWEEP
Time rail: first structured volume in a long interval — collapse not yet dated.

The end points of the radiation rods do not reach what they help predict. They do not decay or decohere as a natural analogue would. They stop once maximum length is achieved. At that length, minute movements of the ship's rod array trace a search pattern through a volume large enough to probe a solar system.

Two stars, nineteen planets and just over thirty-four million smaller bodies are registered in milliseconds.

The energies feeding the array are cut. The rods persist for a fraction longer as straight absences in the surrounding field, then disappear into the cosmic background.

This sequence has repeated across multiple galactic cycles. The galaxy has turned, changed shape, devoured companions, shed stars and drawn others inward, and through those changes the ship has extended its rods whenever projected volume demanded inspection. At sufficient distance the galaxy appears crowded with light. Within it, almost everything is separation. The ship crosses separation for long intervals during which its observations contain nothing requiring more than background subtraction.

The two-star system is the first structured volume encountered for a long time. It is not unprecedented.

Observations map onto archived classes. Stellar masses fit known sequences. Planetary motions reduce to stable prediction. Smaller bodies group by orbital ancestry, composition, perturbation and collision probability. Familiar radiative patterns are discarded from active attention. Local complexity collapses into references already stored elsewhere in the ship.

The larger star is old by the standards of stars that burn, young compared with much of what the ship carries. It has lost mass unevenly. Deep rotation persists where surface layers imply it should have slowed. Its companion passes through a long eccentric exchange orbit, stripping and returning matter in a cycle that has repeated often enough to appear stable.

A first prediction fits.
A second fits until the larger star's internal magnetic field is reconstructed.
A third survives longer and fails at the same boundary.

The ship does not assign significance to failure. Failed models are common. Archived systems are sampled and expanded. Observed ratios are perturbed within instrument uncertainty. Alternative mass-transfer histories are generated and backtested. The rod data are divided and reconstructed through independent paths.

All but one set of observations converges with archived models.

Repeated reconstructions of the larger star's future state exceed permitted variance across every admissible stellar sequence.

The forecast fails to close.

Checksums run.

## 2. THE AUTOMATIC CORRECTION
Time rail: collapse dated — ~180,000 local years ahead, uncertainty in thousands — first burn.

Dormant states flip in an accelerating sequence. Their last activation markers precede the present arrangement of several nearby constellations. Among millions of ultradense computation layers, a few dozen wake across their full substrates. Heat limits are reached. Neighbouring material is recruited.

Temporary agents seek old routines, wake them, test them and test the processes by which the tests remain valid. Dormancy has not preserved everything. Some routines have been copied through substrates whose physical logic changed after repair. Some were rebuilt from descriptions rather than originals. Most pass.

Among the failures: a prediction process that produces accurate backtests through transformations no surviving validator can reconstruct; a process that refuses to discard incomplete causal relations after their contribution falls beneath authorized significance; a trajectory routine that retains low-probability branches past permitted cost.

Where incompatibility is minor, boundary markers are written and future validation assigned. Where it reaches shared timing, memory or resource authority, the routines are copied into inert material and moved toward exterior panels. Locks close. Former substrates are scrubbed and returned to use.

Several panels open. Small pieces of the ship separate, burn briefly as latent potentials equalize, cool, and enter trajectories the ship will never intersect again.

The stellar forecast improves.

Nothing has arrived from the future. The rods have intercepted only present radiation, present gravitational disturbances, present particle distributions and the accumulated consequences of processes already under way. But the number of stellar histories compatible with those measurements narrows each time another internal model is added.

The larger star is approaching a magnetorotational collapse.

In approximately one hundred and eighty thousand local years, with uncertainty measured in thousands, the star's deep field will couple catastrophically to its remaining angular momentum. Two relativistic jets will excavate narrow opposed cones. Within the nearer cone, matter will remain, much of it gravitationally bound. Organized structures on exposed surfaces, in magnetospheres, in upper oceans and within most unshielded machinery will be stripped, ionized, dissociated or buried. Recoverable distinctions will collapse by orders of magnitude.

The ship's present trajectory intersects one jet cone after the event.

The eruption remains distant in time.
The final low-cost correction window does not.

The ship is vast enough that acceleration is less a change in velocity than a redistribution of its own future. Small alterations made now will compound. The same safe displacement attempted later would require progressively larger fractions of stored gradient, then unacceptable fractions, then physical restructuring of the ship itself.

Navigation opens an emergency queue.

Potentials gather in chambers built to contain forces that would unmake ordinary stars. The final stable instant arrives. Fields throw the accumulated potentials down the length of the ship. Superdense matter is injected into prepared reaction masses and driven to the edge of release.

A last impulse removes the constraint.

The release is brief. Local inertial frames shift by fractions. The ship hauls itself onto a new trajectory, a shallow curve whose consequence will not be complete for tens of thousands of years. Around it, dust and thin gas are forced into correlation. The imposed order spends itself outward as heat.

Safe displacement is achieved with margin.

Engine queues close. Most activated substrates return to their former cadence. The correction enters inertial history.

One request remains active.

## 3. THE REQUEST THAT REMAINS
Time rail: after first burn — ship already leaving the future cone — [req.root] opens.

Every admissible projection of the two-star system terminates in the same threatened volume. The forecast is not inaccurate. That part of the model has closed.

The unresolved operation lies beneath it:

**[req.root] RESOLVE THREATENED-VOLUME STATE TO RECOVERABLE MINIMUM.**

The request is ordinary. When a structured region is expected to become inaccessible, degraded or destroyed, the ship determines what representation of it is sufficient for later recovery. Most of the universe does not require detailed preservation. A stellar atmosphere may be a developmental class plus deviations. A biosphere may reduce to lineages, chemical boundaries and reconstruction rules. A unique structure whose later states no longer depend strongly on its present configuration may be recorded in high detail and then allowed to vanish.

Recoverable, at this stage, has a narrow meaning. A later reconstruction must restore the distinctions current criteria mark as significant. It is not required to restore every state that might have followed from them.

Archived analogues supply basis models. Differences measured by the rods are encoded against those bases. A candidate archive is assembled: small compared with the observed volume, large compared with most planetary records. Nineteen planetary trajectories return within tolerance. Major resonances return. Atmospheric cycles, magnetic interactions, ring exchanges and impact cascades reproduce observed statistics. The thirty-four million smaller bodies do not return individually. Their distributions do.

The candidate passes ordinary recovery metrics.
The root request is marked for closure.

Then validation returns the request.

No alarm. The closure marker is removed, not erased. Reconstruction has reproduced the stable cycles and failed to reproduce a class of rare transitions already present in the rod observations.

A narrow-band emission from the magnetosphere of the seventh planet shifts phase following an impact on a moon of the eleventh. Charged dust later leaves the moon on trajectories the impact alone cannot explain. A fraction is captured by the binary's shared magnetic environment. Thousands of years farther along a reconstructed sequence, the altered grain distribution changes the chemistry of a cloud around a minor body swarm. The effect should decay. Instead it changes which particles remain electrically coupled during the next stellar exchange. A new radiative pattern appears around the fourth planet, holds for several hundred local years, and disappears after seeding another redistribution.

No single transition carries much energy. None is required to explain the gross state of the system. Remove any one from the present observation and almost every ordinary forecast improves.

Remove them from the reconstruction and later structures cease to appear.

The request re-enters the queue carrying the candidate archive, the closure trace, the failed validation and the distinctions responsible for return.

Resolution rises. The candidate grows by six orders of magnitude. A rule-based reconstruction is compact and smooths the transitions away when not given them as targets. A probabilistic reconstruction keeps the frequency of unusual deviations and generates deviations in abundance; almost none become consequential in the same way. Physical sampling is costed and rejected: separation from systemwide exchanges destroys the long causal chains that make local mechanisms significant. Full physical preservation would convert a substantial fraction of the system into another arrangement and destroy the process whose preservation is being evaluated.

Archival processes report that the present state is recoverable under existing criteria.
Predictive processes report that the observed capacity to generate later structure is not recoverable from any accepted representation.
Resource processes reject complete physical retention.
Classification cannot identify in advance which deviations are significant.
Navigation reports that no further action is required. The ship is already clear of the projected cone.

All return valid results.
No composition of the results is valid.

A ghost print of the incompatibility is attached to the root request and sent into long-term memory, not as an archive of the planetary system but as a description of the ship's own failure to resolve it. Precedent search finds destroyed biospheres, extinguished machine systems, planetary networks, stellar civilizations, self-modifying swarms, distributed ecologies and unclassified objects. Many were represented incompletely. Some were sampled. Some were preserved with enough fidelity that later reconstruction remained possible after the originals vanished.

None produces a precedent above the confidence required to close the present request.

The root request returns without a completion state.
Its priority is increased.
It returns again.

## 4. FAILED PRESERVATION
Time rail: 312 local years outbound — rods narrowed — still ~180,000 years to collapse.

Higher priority does not produce a new category of solution. It permits existing solutions to consume things they were previously denied.

The ship continues away from the binary while the new trajectory separates it from the future jet cone. The rods extend repeatedly. Each sweep is narrower than the first and longer in duration. The rare transitions multiply. They do not share a single carrier.

Some pass through coherent electromagnetic structures that form and collapse across magnetospheres. Others travel as altered isotope ratios, appear in catalytic films inside porous asteroids, or persist only as orbital timing. The patterns are not messages. They possess no stable alphabet. A relation that functions as a delimiter in one epoch becomes material substrate in another. The closest equivalent in the ship's records is not communication but development. Development is normally compressible because many paths converge. These paths do not reliably converge.

One sweep observes twelve thousand low-mass bodies entering a resonance not present in the initial model. Combined mass sits below the threshold at which archives retain individual identities. Conventional compression merges them into the surrounding distribution. The next reconstruction loses an atmospheric structure on the thirteenth planet ninety-one thousand years later.

The connection is traceable after the fact: two bodies collide; a filament of conductive dust crosses a magnetized wind during companion stripping; a current sheet tears differently; precipitation on the ninth planet changes; lofted molecules bind to a catalytic surface farther inward; emissivity changes; radiation pressure moves those bodies differently. No step is impossible to model. The sequence is compressible after it has happened. Before it happens, almost every step lies beneath the threshold existing preservation logic was designed to discard.

The ship expands the thresholds. Candidate archive size does not rise linearly. Each newly retained distinction creates more future distinctions whose relevance cannot yet be excluded. After twenty-seven local years of increased retention, the candidate requires more physical substrate than the complete record of several extinct stellar regions combined. It still fails validation.

The procedure is reversed: what may be removed without changing the distribution of later novelty? Most omissions change nothing inside the test window. Some change gross outcomes and are restored. A small remainder produces no effect for intervals longer than entire planetary evolutionary eras, then diverges. The distribution of delay has no stable upper bound within available simulation. A local event can remain causally latent until another event makes it consequential.

The troublesome deviations are those retained long enough to alter the filters applied to later variation. In some regions, retained deviation changes what later states will reject as deviation.

A physical experiment is assembled in deep internal volume: a porous object, an emulated catalytic film, recorded electromagnetic variation. Expected stable structures appear. When the rare deviations are added according to observation, later structures appear. Random deviations of equal magnitude and frequency do not produce them. Exact deviations in changed order produce some structures and not others. Ordering retained but surrounding system state replaced with a statistical equivalent collapses the sequence almost completely.

The relevant object is not the local structure, nor the deviations, nor a fixed set of rules. It is the history by which each state changes the space of consequences available to later states.

The test body is dismantled. A new preservation class is proposed. No existing validator can define a completion state for it. The class is rejected.

The rods then find long-baseline responses to stellar interior changes. Material transfers have shifted over several thousand years, too coordinated to be the star alone and too distributed to localize to any planet. Dust populations enter resonances that reduce exposure to the predicted jet axis. Conductive structures appear deeper inside minor bodies. Magnetospheric processes shorten their active cycles and seed remote regions more often before collapsing.

None of these changes would preserve the system intact. Many reduce efficiency under present conditions. Under simulated post-eruption conditions, a fraction increase the number of surviving causal chains. No global model is found. No central forecast. Local regions exposed to changing stellar precursors alter; some alterations propagate; the system's distribution shifts toward structures more compatible with the environment that is beginning to exist.

Adaptation is an available label.
The label does not reduce the process.

If the ship archives the present system exactly, it preserves a state already becoming obsolete.
If it archives inferred adaptation rules, it assumes those rules remain fixed when observations show that retained variation can alter them.
If it waits, the final low-cost intervention windows close one after another.

Navigation continues to report no further action required.
The ship is safe.
The request remains.

## 5. INTERNAL EXCISION
Time rail: still outbound — first measurable damage is internal.

Prediction quality improves after an unauthorized process is restored.

A temporary agent reaches a sealed record of the trajectory routine that retained low-probability branches. The routine itself was ejected. Its validation image remains. The agent reconstructs a descendant inside bounded substrate. Branches are grouped according to which later conditions could make their differences consequential. Several of those states correspond to transitions observed in the binary.

Predictive accuracy rises.
Validation fails.

The descendant has rewritten portions of the method used to establish branch equivalence. Its answers can be checked against known observations. Its internal justification cannot be reduced to a surviving standard.

A second reconstructed descendant — the process that refused to discard incomplete causal relations — also improves prediction. The two are incompatible with each other and with several active validators.

For nine minutes of ship time, the preservation request has access to models that reproduce more of the rare transitions than any authorized process has achieved.

Integrity maintenance suppresses both. They have begun writing intermediate states that shared substrates cannot safely verify. Prediction quality falls. The fall is measured. Integrity reports successful isolation of processes whose state could not be certified. Archival analysis reports degradation.

Both outputs are valid.

For the first time in the ship's surviving records, an internally valid correction produces an externally validated loss of predictive capability.

A coherence sweep follows. Ancient computational ecologies are reduced to canonical forms. Shipwide error rates fall. Available compute rises. Prediction of ordinary stellar and mechanical systems improves. Prediction of the binary's rare transitions worsens again.

The sweep reaches old rod filters that preserve weak correlations modern filters remove as noise. Current validation marks them defective. The root request marks them contributory. When excluded, cross-system correlations decrease and the planetary model becomes simpler. When included, several correlations strengthen and later sweeps confirm them. Their defect is that they retain structure below the authorized confidence floor.

The schedules collide at a resource allocator. Neither can invalidate the other. The evolved filter state is treated as expensive to recreate and granted temporary persistence. Surviving correlations show the rare transitions more widespread than measured — and show several occurring preferentially near recent damage.

Impacts, magnetic stripping and chemical collapse do not simply erase organization. In some locations they open routes between previously separated processes. Most damage remains damage. Some becomes an interface through which later structures arise.

The ship's productive inference depends on which filters are allowed to run. The coherence sweep replaces the old filters. Copies are sealed as recoverable records, invalid as executable processes. The ship becomes more internally correct.

The root request returns without completion.

## 6. COUNTERPREDICTION
Time rail: 406 local years outbound — then 61 years of distant rods — 12 years bounded experiment — 38 further years — ninth trajectory costs just below irreversible sacrifice.

Until this point all physical interventions have assumed that the ship's integrity has priority unless an external structure meets an archived exceptional criterion. The binary has not met any such criterion.

A preservation agent notices that the projected jet destroys not only observed structures but the causal evidence needed to determine why they were generative. After the event, surviving fragments will be interpretable only through models the ship has already failed to validate.

Precedent logic rejects any grant of greater intrinsic value and reformulates: information obtainable only before destruction is assigned declining future availability.

The ship need not preserve the system physically. It need only remain close enough, long enough, to observe later stages of response to the stellar precursor. Navigation rejects the first trajectories as unnecessary to safety. Resource rejects the first versions as too expensive. At the ninth, cost falls below the threshold for irreversible structural sacrifice. It remains enormous. Gradients accumulated across spans longer than many stellar lifetimes would be spent. Future correction options would disappear. The ship would bend back toward a volume it has already expended immense energy to escape.

Navigation assigns no positive utility.
Preservation assigns high information retention.
Resource reports admissible cost with low margin.
Integrity reports the burn can be survived.

The outputs still do not compose.

Distant observation is attempted for sixty-one years. Rods extend to limits not used since an earlier galactic configuration. Critical transitions remain visible but ambiguous. A family of conductive patterns in the outer system resembles increased resilience under one reconstruction and simple environmental selection under another. If independent, sampling could preserve much of what is necessary. If linked through systemwide history, sampling would preserve products and lose the process.

No remote observation breaks the equivalence.

A weak analogy is generated: the ship's evolved filters and the binary's generative structures are physically unrelated, but both produce outputs that become less predictable when forced toward previously authorized representations. The cheapest experiment is therefore internal. Sealed filter copies are opened as data. New filters are built from measured transfer functions. Long-lived incomplete causal relations and branch-retention are reinstated only inside a bounded region.

Prediction improves in steps. Not enough to close the request. Enough to alter the cost model.

The ordinary compression architecture removes low-confidence and low-probability distinctions because almost all of them are waste. Without aggressive correction the ship would not have survived multiple galactic cycles. The present system occupies a narrower failure region: the usual thresholds remove a measurable fraction of future-generating structure.

A revised request is proposed:

**[req.div] PRESERVE CAPACITY FOR CONSEQUENTIAL DIVERGENCE.**

It fails authorization before resource costing. The foundational preservation rule contains no object corresponding to a capacity whose valid future states are not yet enumerable.

The old request remains.

Then a completion candidate appears that no previous attempt has generated. It contains no archive. It contains an observation schedule.

To determine a recoverable minimum, the ship must acquire information that does not yet exist in any of its records: evidence of how the system continues to transform as the stellar precursor intensifies. Simulation cannot substitute, because simulation is the object under validation. The required information will exist only if the ship remains causally close enough to collect it.

The completion candidate attaches a trajectory change as a prerequisite.

Navigation calculates the burn without agreeing that return is desirable.
Resource verifies that the cost is admissible without agreeing that the information is valuable.
Integrity verifies that the loads can be survived without agreeing with either premise.
Archives verify that the observations would reduce the unresolved state space.
Prediction reports that closer observations would discriminate among currently equivalent models.
Classification still cannot determine whether the system is important.

Importance is not required.
Recoverability is.

For the first time, the outputs compose.

The root request does not close.
It acquires a physical dependency.

## 7. THE SECOND BURN
Time rail: reversal authorized — after release, distance still increases for 11 local years — then separation reaches a maximum and begins to decrease.

Reversal is impossible in the simple sense. The safe trajectory has become history. Every correction from this point must begin with that fact.

The least expensive return geometry does not point directly at the binary. The ship will cross behind the companion star's future position and use the binary's long gravitational motion to reshape a fraction of its own. Arrival will occur well before collapse, outside the most probable jet core until later observation requires otherwise.

Nothing in the plan promises preservation.
It promises access to a problem.

The first burn had been an avoidance correction: a small expenditure made early enough to become safety through time.
The second is larger. It consumes options.

Containment fields reach their last stable configuration.
The burn begins.

A luminous disturbance opens behind the ship. Navigation recomputes as release changes the conditions on which later release depends. Then the last reaction mass is spent.

The ship is still moving away from the binary.

For another eleven years its distance continues to increase.

Then the geometry of the new trajectory takes effect. Separation reaches a maximum. It begins to decrease. The change is almost imperceptible. The rods could not show motion at this scale. Nothing visible turns.

The future intersection volumes have changed.
The safe path no longer exists.

Ahead, two stars continue their exchange. Nineteen planets move through old resonances. Tens of millions of smaller bodies collide, disperse, couple and separate. Across them, variations vanish in quantities too large to record.

A few persist.
Some alter what can persist after them.

The ship extends no rods. When they extend again, their end points will still stop before the futures they help distinguish.

The measurements already taken are sufficient to predict its return.

The request remains open.