A Proposal to Collar the Anstolik, Refused at Countersign

A note on what this document is. It is a proposal that was not adopted, filed complete with the objection that stopped it, at the request of the officer who raised the objection rather than the officer who wrote it. Neither half is the Archive's voice. The first is what a capable instrument specialist believed, in good faith, could be built; the second is what the survey lead believed it would do once built. They are filed together because a refusal kept without the thing it refused teaches nobody anything, and because the proposal is the better document with the refusal attached.

---

Part One — Prospectus

Filed by Instrument Specialist Toma Egede, Fliade surface party.

1. What is proposed

Twelve adult Anstolik of the equatorial herds, fitted with a passive detection collar, released, and left entirely alone thereafter. No handling after fitting, no recall, no herd management of any kind. The collar carries no cognitive subsystem, issues no instruction to the animal, and does nothing the animal can perceive. It watches the ground the herd was going to cross anyway.

At 2.4 kg it is six-tenths of one percent of an adult's mass, against the three-percent ceiling the Safety Corps inherited whole from Terran wildlife telemetry practice and has never seen a reason to loosen. The strap parts at a corroding link sized to fail inside one migration cycle, so the failure mode of the entire programme is a collar on the ground and an animal that is no longer carrying anything.

2. What it carries

A multispectral imager across visible, near-infrared and thermal bands. A four-element microphone array. Inertial dead reckoning, corrected by the survey relay when the relay is overhead and left to drift when it is not — there is no navigation constellation at Fliade and no case for building one, so the drift is logged rather than smoothed. Pressure, temperature and wind.

Power is the honest problem and I will not bury it. The collar harvests thermoelectrically from the gradient between the animal and the air, which on the winter tundra is sixty kelvin and generous, and in the short warm season is almost nothing. That is precisely the wrong way round: the herds are most exposed in summer and the instrument is poorest then. A primary cell covers the shortfall for one season and one only.

Inference runs on the collar. Nothing streams. The processor sleeps until an acoustic or thermal trigger wakes it, classifies, and either returns to sleep or spends the power to raise the relay in a burst of a few hundred bytes.

3. What it would detect

Four signatures, in descending order of confidence: an upright bipedal thermal profile at range; drawn wire closed into a loop, which no Fliadan organism makes and no Fliadan process produces; the thermal scar of a landed craft; and a track that does not close — the mark of something that came, took, and left, rather than something that lives here.

The prospectus carries a figure of 96.7%, and I want to be exact about what it is a figure of. It is bench performance against synthetic scenes and archived Terran preserve footage. It is not a field figure, because Fliade has never furnished a single positive example to test against. A model that has never seen the thing it is looking for is making a claim about its training set and calling it a claim about the tundra. I say so here rather than in a footnote.

4. Why an animal

Because the alternatives are worse. A mast is a surface installation, and the standing doctrine on this world is that nothing stays on the surface that the planet would keep if the station were abandoned. A drone must be powered, is observable, and has to be told where to go. The herds are already the coverage pattern: they range the corridors year on year, they range them at the density the forage dictates rather than the density a planner would guess, and they need no instruction, no fuel and no permission from us to do it.

5. Objections I expect, and my answers

That this makes an Anstolik a Smart Pet. It does not, and the distinction is not a technicality. That regime governs a certified cognitive subsystem running alongside an animal's own mind. There is no subsystem here. The animal's cognition is untouched, unaugmented and uninvolved. The collar is an instrument, not a colleague.

Welfare. Mass fraction inside standard, no handling after fitting, breakaway link, and re-certification at each fitting by whoever the Corps sends.

Identification. The model classifies shapes. It does not match identities, holds no reference set of persons, and stores no image it has not classified as a detection. It can tell you that something upright walked through the corridor. It cannot tell you who.

---

Part Two — Countersign Withheld

Filed by Survey Lead Hesper Nakagawa, Fliade survey.

I am refusing to countersign, and I am filing the prospectus rather than returning it, because it is good work and the reason it fails is worth more to this survey than the proposal was.

1. The rule about the surface

The doctrine is that nothing stays on the surface that the planet would keep if the station were abandoned. Specialist Egede reads that as an argument against masts. It is an argument against collars first. A mast can be dismantled by the party that raised it. A collar has to be caught, and an Anstolik defends its corridor with more conviction than its build suggests. We would be leaving hardware on this world in the one form we cannot take back at will, and calling it the light-footprint option because it has no foundations.

2. The instrument walks downward

The herds range the surface plains and the upper cavern levels. The Pandoids hold the deep networks. Where the upper levels stop and the deep networks begin is not surveyed, is not going to be surveyed, and cannot be, because surveying it means going down there — which is the thing this survey has spent its entire window declining to do.

So state plainly what the programme does. It carries a recording instrument an unknown distance toward a people who were never asked, aboard an animal nobody can recall, on a schedule set by the animal.

And the privacy answer collapses at exactly this point. Granted that the model holds no faces. Its highest-confidence trigger is an upright bipedal thermal profile at range, and the Pandoids are large, upright and bipedal. Run this system on Fliade and its single most confident output will not be a poacher. It will be a Pandoid, correctly classified, logged as a detection, and relayed to a station they do not know exists.

3. The animal was not asked either

The Kernel's first invariant counts harm to an animal as harm, and has since ratification, without the animal needing standing to claim it. Specialist Egede's strongest argument — that nothing is done to the Anstolik's mind — concedes the question instead of answering it. Nobody alleged the herd would be altered. The question is whether a member of a first-contact world's biosphere may be fitted as survey equipment because it goes where we have agreed not to go. That is the same asymmetry this world already puts to the charter, arriving a second time in a smaller body.

4. The premise is not established

The record does not show that any Anstolik has ever been taken. What the record shows is three unregistered craft logged at the survey's edge across eleven years, none identified, none seen twice, and an interest in this planet well in excess of any settlement case — which the boundary feature explains without needing a poacher. That is a reason to watch. It is not a taking, and the gap between those two things is where this proposal lives.

I am not refusing because unauthorised taking is impossible here. I am refusing because we have written the response before we have written the finding, and the response is more certain than the finding is. The absence stays marked.

5. Where the alert goes

There is no enforcement presence on this world and none is coming. The alert path terminates at a station that can log a taking and do nothing about it. If the programme proceeds, that belongs on its first page and not in an annex, because it changes what the programme is: not an intervention, but a record of things we watched.

---

Appended by Specialist Egede, at the specialist's own request:

I accept the countersign and dispute one word of it. The Survey Lead writes that the herd would be conscripted. It would not be conscripted; it would be unasked, which is worse, and I should have been the one to write that down.

The file is closed to the proposal and open to the question.

See also: Fliade, What the Record Refuses, Marked Absences, Frontier Transformation Protocols, Cross-Biosphere Taxonomy, Smart Pet, The AI Safety Kernel, The Star Rangers Safety Corps.