The strongest case against this thesis
Take the objection at full strength. Stock assessment already has a state-variable answer to path dependence. Fisheries scientists do not model raw catch history; they fit spawning stock biomass, recruitment, and natural mortality into a virtual population analysis or a statistical catch-at-age model, compress decades of survey and landings data into a handful of parameters, and then forecast from those parameters plus the current year's inputs. The Beverton-Holt or Ricker recruitment functions are precisely a claim that a fish stock's future response can be read off a compact state, not off the raw trajectory. If that is right, continuous intake is a nicety. A biennial assessment cycle, however irritating to a Large Universe Model enthusiast, already carries what matters. Add more streams and you add noise, not information.
This deserves to be taken seriously before it is answered, because it is the working assumption of most fisheries science as practised, and it is not wrong about laboratory-grade cases. Where a stock's biology is stable and well sampled, the compact model earns its keep.
Where the compression fails
The trouble is that fish stocks are not ferromagnetic cores. The hysteron distribution — the thing the model assumes is fixed while it fits parameters to it — drifts, and it drifts for reasons the assessment cycle is structurally slow to notice.
North Sea cod is the standard case. Recruitment collapsed in the early 2000s and did not recover even after fishing mortality was cut sharply and spawning stock biomass rebuilt towards target levels. The Beverton-Holt curve fitted on pre-collapse data said recovery should follow reduced pressure. It did not, because the regime underneath the curve had moved: warming North Sea bottom temperatures shifted the timing of the spring plankton bloom against cod larval hatching, breaking the match that recruitment success depended on. The state variable — spawning stock biomass — looked adequate. The function relating it to recruitment had quietly become a different function. That is exactly the "hysteron distribution drifts" failure: a sufficient statistic that is sufficient only relative to a model going stale, and nothing in the assessment's own internal check flagged the staleness in time.
This is the first objection worth answering directly, and the concession is real: where the biology is stationary, compression to a handful of parameters loses nothing important, and a fisheries scientist rebuilding a Preisach-style memory model of a stock is doing legitimate, efficient science. But stationarity is an assumption, not a finding, and the only way to know it has failed is to keep watching the streams the compact model was built from — temperature anomalies, plankton timing, larval condition indices — for signs that the fitted relationship and the observed world have parted ways. Continuous intake is not there to replace the model. It is there to audit it.
The quota lag as hysteresis made procedural
The characteristic failure in this domain is mundane and well understood by anyone who has worked a fishery: the quota for this season was set from a stock assessment that closed its data window one or two years before the season it governs. Survey cruises run on fixed schedules, assessment models take months to fit and peer-review, and management councils need a number in hand before the season opens. The result is structural, not accidental. Every quota is a statement about where the stock's loop was, not where it is.
For a hysteretic stock this is not a rounding error. If the true relationship between spawning stock biomass and recruitment depends on which branch of a temperature-driven loop the ecosystem is currently traversing — cooling towards the historical productive regime, or warming past a threshold where recruitment support collapses — then a quota calculated on last cycle's branch can be actively wrong in direction, not just magnitude. Setting allowable catch as though the stock sits on the recovering branch, when three consecutive warm-anomaly years have in fact pushed it onto the collapsed branch, produces exactly the Large World Model failure mode: correct read of a scene that has already closed, no visibility into the excursion that happened after the window shut.
The assessment was rigorous. The model was peer-reviewed. The quota followed the science. What more were we supposed to do — assess continuously with no fixed cycle, no defensible sign-off, no stability for the industry to plan against?
That objection has force. Fixed assessment cycles exist because continuous re-quoting would be administratively unworkable and would strip industry of the planning certainty that makes compliance possible. This is the second objection worth taking on directly: a reset-and-refit strategy, run on a schedule, has real virtues — auditability, comparability, a defensible paper trail — that a stream of constantly revised beliefs can struggle to match. Annealing the strain gauge, in this domain, is closing the assessment window on a fixed date so everyone knows what data grounded the number.
The limit of that defence is that a fishery cannot be annealed the way a strain gauge can. Closing the assessment window does not pause the stock. Temperature anomalies, recruitment failures and effort shifts keep accumulating on the far side of the cutoff, unmeasured by the process that is about to set next season's number. The fisheries scientist responsible for the assessment is not negligent; the failure is upstream of any individual's diligence, built into a procedure that treats a snapshot as if it were the trajectory.
What continuous intake actually buys here
Not retrodiction. This is the second concession, and it should be made plainly: a continuous monitoring system switched on today cannot recover the thermal history of the North Sea in the 1990s, and it cannot manufacture the larval survival data that was never collected during the years the collapse was setting up. Whatever excursions already shaped the current stock happened whether or not anyone was watching, and no amount of forward-looking intake fills that gap.
What it buys is a different failure mode for the number that gets set. A quota system built on continuous catch reports, survey vessel data as it comes in, temperature anomaly feeds and quota filings from adjacent fleets, each belief tagged with when it was last updated and by what evidence, does not pretend to certainty it lacks. It can flag that the recruitment relationship underlying this season's number rests on an assessment now fourteen months stale, that three successive temperature readings have moved outside the historical envelope the model was fitted on, and that the belief "stock is on the recovering branch" carries a confidence that has been quietly decaying since the last survey closed. That is provenance doing its job: not proving the trajectory, but making visible how much of it is missing.
The narrower claim
Fisheries management does not need, and could not use, total observation of every excursion a stock has ever undergone. Compact state-variable models earn their place wherever the underlying biology holds still long enough to be fitted. The claim that survives scrutiny is narrower and harder to dismiss: wherever the fitted relationship might itself be drifting — and in a warming ocean with shifting bloom timing, it generally is — only continuous intake, with each belief about the stock's condition carrying an explicit record of when it was last checked and against what, can distinguish a quota that is current from a quota that is merely official. Fixed assessment cycles will remain necessary for the reasons of tractability and legitimacy their defenders give. But the number they produce should arrive attached to its own expiry date, not presented as though the stock stopped moving the day the survey vessel came home.