On October 14, 2022, at 14:23 GMT, Airbus SE publishes a revised delivery forecast for the A320neo family: 720 aircraft for 2023 instead of the 700 announced in July. Twenty aircraft, on an order book of several thousand, a correction of 2.8%. The financial press covers it in one paragraph. Inside the industrial apparatus, the shock is of another order.
A commercial aircraft is not manufactured. It is assembled from subsystems produced by a hierarchy of suppliers whose production was already underway before the July forecast and whose schedule had been calibrated to that forecast. CFM International had committed to a LEAP engine production rate sized for 700. Safran had aligned its landing gear deliveries accordingly. Spirit AeroSystems had adjusted the fuselage shipment rate. Each of these commitments had in turn propagated down the chain, to tier-two suppliers, to titanium and aluminum mills, to electronic component manufacturers, to specialized forges. The July forecast was not a prediction. It was an operational document that had already organized, at each level of the chain, the present work of several thousand companies distributed across fifteen countries.
At 14:23, the number changes. Twenty additional aircraft mean, at plausible propagation coefficients, approximately 40 additional engines (two per aircraft), 80 landing gear sets, 1200 additional tons of aluminum in rolling programs, a modified cadence on avionics lines, adjusted treasury positions, renegotiated hedging contracts. None of this is decided by Airbus. All of this is decided, in the hours and days that follow, by hundreds of people in companies that Airbus does not own, based on a document that Airbus has just modified.
The question that interests us is not how the industrial chain reacted. It reacted, as it always reacts, through a mixture of acceleration where slack existed, conflict where it did not exist, and discrete renegotiation of downstream promises to absorb the upstream revision. The question is what sort of object the July 2022 forecast was during the three months that separated it from its revision, and what it became after 14:23.
The naive answer maintains that the July forecast was a prediction that proved false and was corrected. The October forecast, in this reading, is simply more precise. This description is operationally false. It confuses two distinct functions that the same document fulfilled simultaneously.
Doctrine
An industrial forecast is two objects at once. It is an epistemic assertion about a future state of the world, and it is a performative instruction addressed to a production system. The first function is what the finance department reads. The second is what the supply chain executes.
When these two functions are carried by the same number, a forecast revision is not a belief correction. It is the rewriting of a set of instructions that have already been partially executed.
What distinguishes this situation from cases the corpus has treated elsewhere is that the forecast is not a single contractual deadline like the Apollo date of July 1969, fixed in 1961 and held until its realization. The forecast is continuously revised while the work it organizes is already underway. Each revision propagates backward through a production system that had been calibrated to the previous version. The present is not contracted by a single future; it is contracted by a succession of futures that crash into each other, each leaving operational traces that subsequent revisions cannot erase.
Interlude
A system constrained by a future that does not change allows itself to be analyzed in terms the corpus has already established. The end organizes the work. The form of what will come determines the form of what comes. The Deadline Structures the Work covers this configuration. Apollo is its emblematic case precisely because the date was fixed and unilateral; Kennedy could not revise it downward without losing the discourse, and the engineering apparatus could therefore be designed against an immobile constraint.
A system constrained by a future that is revised while work is underway is another animal. The July forecast organized three months of production. These three months leave residues: ordered parts that are now surplus, signed contracts that are now inadequate, reserved capacity that is now misallocated. The October forecast, which replaces July's on the epistemic plane, does not replace it on the operational plane. It superimposes itself upon it. The production system at 14:24 is not organized by the October forecast alone. It is organized by the July forecast as it was being executed until 14:23, plus the October forecast from 14:23 onward, plus the transition cost between the two. The system lives, permanently, in a regime where its present is shaped by a stratified sequence of anterior futures, each of which was at one moment the sole future toward which the system worked.
Material hysteresis, discussed elsewhere in the corpus (Hysteresis, OBS-012), offers a partial analogy. A plastically deformed metal does not return to its initial state when the load is removed; it carries the trace of the path traveled. An industrial system contracted by a sequence of revised forecasts is in an analogous situation with regard to its obligation schedule. The current state of the system is insufficient to describe it. One needs the sequence of forecasts that brought it there. Two industrial systems having the same present order book, the same stock levels, the same workforce, can behave differently if their forecast histories differ, because each revision has left sedimented constraints: unabsorbed cost overruns, more or less damaged supplier relationships, hedging positions more or less aligned with the new number.
The analogy is partial because the mechanism is different. Hysteresis in a metal is a property of the material's microstructure. Hysteresis in an industrial schedule is a property of the set of contractual and operational commitments that revisions leave intact or damage. In the metal, the path is inscribed in dislocations. In the industrial system, the path is inscribed in counterparty relations, in trust, in suppliers' capacity to absorb additional revisions without failing. The inscription is legal and economic, not structural. But the behavior is analogous: state alone does not determine response to the next perturbation.
Ce que cela implique
A revised forecast is not simply replaced; it is overwritten while its previous version still produces effects. This observation has three consequences that the single-deadline framework does not capture.
The first concerns the status of the revised forecast. When October's number replaced July's, July's did not become false. It became historical. The work it had organized had been real work, the contracts it had caused to be signed were real contracts, the parts it had caused to be produced were real parts. A revised forecast does not exit the world. It takes up residence in the world it has already produced. The industrial present at 14:24 on October 14, 2022 contained, materially, everything the July forecast had brought into existence, including what the October forecast now considered surplus.
The second concerns the nature of the constraint. In the Apollo configuration, the deadline is a boundary condition: a single specified result at a single specified time. In the continuous revision configuration, the constraint is not a boundary but a field: at each instant, the system is constrained by the current version of several prospective projections (next year's deliveries, three-year capacity plans, ten-year strategic targets), each of which is itself revised according to its own schedule. The present is not shaped by a future. It is shaped by the current configuration of a set of futures that co-evolve. The classical language of retrocausality, which speaks of the future acting on the past, suggests a linear relationship between two points. The industrial case suggests something else: the present is the intersection of a set of prospective documents, each of which is alive.
The third concerns the system's exposure. A production apparatus organized against a single immobile future can be optimized against that future: resources can be allocated without reserve, slack can be minimized, the path from present to target can be compressed. An apparatus organized against a continuously revisable future cannot afford this optimization. It must maintain a margin for revision, and this margin is a cost paid in the present for volatility situated in the future. Industries exposed to frequent forecast revisions (aerospace, semiconductors, automotive supply) structurally carry more slack than their optimization alone would require. This slack is the visible industrial face of retrocausal volatility. It is the price the present pays for being shaped by a future that refuses to stabilize.
Vecteur ouvert
The corpus has treated retrocausality in two registers. In the physical register, via the two-state formalism and delayed-choice experiments, the constraint exercised by the result is introduced by a measurement that fixes the post-selection. In the industrial register, via Apollo, Carnegie, Weibull, the container, the constraint exercised by the result is introduced by a specification that is maintained stable long enough for the organization it shapes to reach it.
What has not been treated is the regime where the specification is itself an object in motion. In aeronautics, in semiconductors, in any industry operating on capacity commitments of three or ten years against demand signals that update monthly, the specification is continuously edited while its execution continues continuously. The industrial object is not organized by a future. It is organized by the living difference between successive futures.
This raises a question the corpus has not posed. If what shapes the present is not a fixed future but the trajectory along which the future is revised, then the quantity one would need to predict to predict the present is not the specification itself but its revision rate. The Weibull curve predicts when the machine fails. What predicts when the forecast will be revised? Does there exist, for industrial specification, an analogue of the failure distribution: a probability law on revisions, estimable from the historical behavior of the institution that specifies, that would characterize the system's retrocausal volatility as Weibull characterizes its mortality?
If such a law exists, it would be the first object of industrial knowledge of a new kind: a metrology that would bear neither on the object, nor on the process, but on the instruction that governs both.
