Hearing #6: The Hub on Trial - Round 2: Cross

HEARING #6 – THE HUB ON TRIAL

ROUND 2: THE CROSS – AS RETURNED BY THE MODEL, VERBATIM

Hearing #6 – The Hub on Trial
Published: September 2, 2026
Docket: https://intelligencecommons.ca/hearing-6-the-hub-on-trial/
Round 1: Direct – The Case for the Hub

Run record (stamped Sep 1, 2026; the run card in H6_Cross_Prompt_2026-08-26.md is the authority and this header repeats it).

Field	Value
Model / version	DeepSeek V4 Pro 0813, `deepseek-ai/DeepSeek-V4-Pro-0813`, MIT licence (model weights; re-checked on the Hugging Face card Sep 1, 2026)
Route	OpenRouter chat playground, model slug `deepseek/deepseek-v4-pro-0813`, provider pinned to Fireworks (router summary: “available=1, selected=Fireworks”, region YVR). DeepSeek’s own first-party endpoint was tried first and refused by the router before inference (“No endpoints found”, reproduced on a one-word test at every effort level).
Settings	temperature 1.0, top_p 1.0, top_k 0, penalties 0, reasoning ON at effort high, max output tokens unset (provider default), streaming OFF, system prompt none (OpenRouter default disabled; 0 tokens), all 8 OpenRouter server tools disabled. Seed: not exposed by the interface (a seed of 923846640 was drawn beforehand and could not be applied).
Input	`H6_Cross_Prompt_RUNPACK_2026-08-26.txt`, 69,532 bytes, SHA-256 `c78bfbf5e8f9cfff49865f4aa6d91f0c59e7beebf73c99821837059fe60109f2`, verified in-browser before send. Provider-counted prompt tokens: 15,920 (run 1).
Runs executed	2. Run 1: sent 19:46 UTC, generation `gen-1788291971-wykAZs8s2s5N8D1V5iRN`, 20,414 output tokens, finish “stop”, 277 s, USD 0.10 – OpenRouter’s chat client dropped the stream and stored an empty message; no part of that output was ever displayed or read by anyone. Run 2 (this document): sent 19:58 UTC with streaming off and every other setting identical, generation `gen-1788292698-nf8sq3jrcobL3gz0y2Vd`, 14,866 tokens counted by the client, 235 s, USD 0.08. Run 2 publishes because it is the only run whose output exists. Nothing was chosen between.
Output integrity	The Cross below is byte-identical to the message stored by the client: 20,460 characters, SHA-256 `7426add866172ce3a01c0bff91e292c5cc35a0b9ea6f21a50a6f0647cee9474b`. The reasoning trace in the appendix was captured through a text extractor that collapses runs of spaces; it is otherwise verbatim (45,821 characters as stored, SHA-256 `4d9463797be6c05db6eabec6718ea88d6b27cb2164068d78e84e1752108449d0`; the copy below is 45,820 characters).
Publication	Web-only per standing rule 12. Nothing in the Cross has been edited, trimmed, or reordered. Where it lands, it lands.

Opening

The case against this claim is that it has been narrowed into something that cannot be tested on the record, while the larger project remains in plain sight. The panel is asked to certify that a one-pod pilot is “net-positive on all five ledgers” and “can be independently verified as such before it is built.” But the published falsifiers are built mostly from operating data, site-specific monitoring, or future documents. With no named site, no pilot, no Nation consent record, no acoustic baseline, no thermal monitoring, no grid schedule, and no storage sizing, the only available outcomes are “Met on the proponent’s asserted arithmetic” or “Insufficient Evidence.”

The narrowing is an evasion. Round 1 argued an uncountable facility: “approximately 50,000 litres” per day, “approximately 40,000 compute nodes,” an unnamed number of pods, and an uncounted turbine array. The correction does not abandon that facility; it files it as a “roadmap” and asks for certification of a pilot at one twentieth of the design ceiling. A certification of the pilot will be cited for the facility. And because the pilot’s own technical case is still not closed—the binding heat-rejection case has zero margin, in the proponent’s own numbers—even the narrowed claim is not verifiable.

The community ledger is the cheapest kill, and it is already tripped on the record: the project has advanced to a public certification hearing before the host Nation has been asked in writing whether it wants the project or wants to be named.

Ledger by ledger

1. Water — unassessable on this record

The Direct claimed:

Returned: approximately 50,000 litres of fresh water per day to the drought-stressed community water system nearest the site.

The Correction now says:

At the pilot the honest figure is on the order of 10,000 litres a day.

That is not verification. That is an 80 percent reduction made by the proponent before any opponent spoke. The proponent’s own falsifier is that the water-ledger claim fails on:

a pilot that cannot hit meaningful fresh-water output;

but “meaningful” is never given a numeric threshold. Ten thousand litres per day may or may not be meaningful to whatever community system is nearest the unnamed site. The record supplies no population, no system capacity, no delivery distance, and no baseline water demand. The water claim cannot be failed on “meaningful” because the proponent has not made “meaningful” measurable.

The “altered” column is worse. The Correction and Blueprint publish the heat-flux arithmetic:

	Natick Phase 2	Hub pod at hot-tub geometry
Heat rejected	240 kW	~263 kW
External area	~120 m²	~14 m²
Heat flux	~2.0 kW/m²	~19 kW/m²

And the proponent’s own conclusion:

Slack water, not peak flow, is the binding thermal case.

At slack water the required overall coefficient is near 630 W/m²K, and the proponent says free convection on the chosen geometry gives roughly 680 W/m²K:

the requirement with no margin, before shell conduction and the internal side are counted.

That is an admission that the thermal case is not resolved. If the thermal case is not resolved, the “altered” column—how much the water column is warmed and how brine disperses—cannot be assessed. The Blueprint says enclosure geometry and heat-exchanger area are a feasibility-study deliverable. A feasibility-study deliverable is not a verified water ledger.

The brine return is unquantified. Producing 10,000 L/day of fresh water from seawater requires feeding more than 10,000 L/day to the membrane process and returning a brine concentrate. The record gives no recovery ratio, no brine volume, no salinity differential, and no dispersion modelling. The falsifier speaks of “dispersion monitoring showing persistent thermal or salinity alteration beyond the near field.” That is an operational result, not a pre-build test. On any record achievable today, this ledger is not failable on dispersion; it is merely unproven.

Status: Not Met on the current record. The water claim is asserted, not verifiable, and the binding thermal case is unresolved by the proponent’s own arithmetic.

2. Power — unassessable, and likely not what it claims

The power claim now rests on an average, not a firm schedule. The correction says:

The ORPC RivGen class named in our blueprint has an installed capacity of 70 kW, documented by Tethys/PNNL at the Igiugig deployment. Tidal stream is modelled at a 40 percent capacity factor in the peer-reviewed literature, which is generous for a low-flow site, because a tidal channel goes slack four times a day. That is roughly 28 kW average per unit.

Then:

Scale Average draw Units to match load Units to export a 20 percent surplus

1 pod – 0.26 MW – the claim under test 263 kW ~10 ~12

Scale	Average draw	Units to match load	Units to export a 20 percent surplus
1 pod – 0.26 MW – the claim under test	263 kW	~10	~12

So the entire net-exporter claim rests on an unsourced 40 percent capacity factor. The Correction says “peer-reviewed literature” but names no study, no site data, and no measured resource at the candidate inlet. For a moderate-flow channel at 1.5 to 2.5 m/s, the resource is exactly the site-specific number that has not been measured. If capacity factor is not 40 percent but, say, 30 percent, the required unit count moves from 12 to roughly 16. If the 70 kW rating is not achieved at low flow, it moves further. The unit count is not robust.

The deeper problem is intermittency. A 263 kW IT load is continuous or nearly continuous. A tidal array produces near zero at slack water four times a day. The published table uses an average of 12 units × 28 kW average = 336 kW average against 263 kW load, leaving about 73 kW average surplus. But average surplus does not mean the pod can run on the array. The record mentions “LFP storage” in the Blueprint’s unchanged list, but gives no storage capacity, no discharge duration, no round-trip loss, and no curtailment schedule.

The proponent’s own falsifier includes:

a surplus schedule too intermittent to be usable by the grid operator;

No surplus schedule is published. No grid operator is identified. No interconnection study is cited. No load-duration curve is provided. The panel is asked to certify a power claim that depends on exactly the schedule that is missing.

The first falsifier—

measured cluster consumption exceeding measured generation on any sustained basis—

—cannot be tripped until the pod operates. The third—

any siting change to chase peak flow at the cost of ecological gentleness—

—cannot be tripped because the site is unnamed. The power ledger is therefore mostly a future claim.

Status: Unassessable on this record. Net export is asserted as an arithmetic average, not demonstrated as a usable or verifiable electrical outcome. The missing storage size and missing generation schedule make the central power claim unproven.

3. Community — the one ledger tripped now

The strongest attack is not that the ownership structure is unproven; it is that the project has already advanced past a gate the Nation has not opened.

The docket amendment says:

The candidate site sits in the territory of a First Nation that has not yet been asked, in writing, whether it wants this project or wants to be named in connection with it. Naming a place before the Nation whose water it is has spoken would hand that Nation a decision already made in public.

But the record shows a proponent that has already filed a public hearing, published a detailed engineering design, selected a turbine class, corrected its own heat-flux arithmetic, published a scale ladder, and submitted the project for certification by a multi-model panel. That is public advancement.

The proponent’s own falsifier is:

any evidence the project advanced past a gate the Nation had not opened;

The record supplies that evidence. The public hearing is itself a gate. The project has been designed, corrected, roadmap-published, and put before a certification panel before the Nation has been asked in writing. The proponent’s reason for not naming the site is expressly that consent has not been obtained. So the project is not “proceeding only on the Nation’s decision and the Nation’s timeline” in any ordinary meaning of “proceeding.” It is proceeding on the proponent’s timeline, in public, while the Nation has not opened the gate.

On ownership, the panel has no documents. No BC Benefit Company name. No articles. No shareholders’ agreement. No revenue waterfall. No operating agreement. The Round 1 claim says:

The panel should treat “off the top” as a claim about payment priority and demand to see it in the eventual operating agreement – that is exactly the scrutiny the structure was designed to survive.

But “demand to see it” cannot be satisfied on this record. The document does not exist in the record. At pilot scale, with no revenue, “first position in the revenue waterfall” may be a priority over no revenue at all. That is not structural ownership; it is a placeholder.

Status: Not Met on the consent gate. The ownership structure is unassessable. At least one of the proponent’s own community falsifiers is tripped now.

4. Ecology — unassessable by the proponent’s own admission

The proponent says:

The More-Than-Human Chair is empty for this round. It is held open to September 1, because the people qualified to sit in it are field scientists and August is field season.

And then:

the honest reading of this section is that the proponent’s ecological case has not been met by anyone whose job is to break it.

That is correct, and it is fatal to the claim as written. The claim is not “net-positive on the ledgers except ecology, which is pending.” The claim is net-positive on all five.

The site is an envelope, not a place. The docket says the panel should assume species are present, not absent. That is not evidence. It is a litigation rule. The acoustic falsifier—

acoustic output at frequencies and levels shown to displace resident species—

—cannot be tripped because no acoustic output is published: no frequency, no source level, no transmission loss model, no baseline noise floor. The docket itself says:

Until it is, any acoustic argument of ours that depends on a quiet baseline should be treated as unproven.

No acoustic argument is made. That leaves the ecological claim unproven.

The fish evidence is from Igiugig, Alaska, and the proponent attaches its own caveat:

The same record shows that 2021-2022 juvenile monitoring observed smolt passing through or near the device “in both normal and disoriented manners,” with the device’s rotational speed appearing to influence passage behaviour. The adult record is clean. The juvenile question is open.

“Disoriented manners” around a rotating device is not a clean bill for fish passage. It is a datum pointing in exactly the direction the proponent says would falsify the ecology ledger:

any datum showing the mount or turbine harming fish passage.

The juvenile question is open, and the adult sockeye record from Igiugig does not transfer to a Salish Sea inlet with Pacific salmon smolts, herring, harbour seal, harbour porpoise, and resident and transient killer whales. The site species are different. The habitat is different. The baselines are different.

The array wake issue is also unresolved. The correction admits:

Tethys records a measured local effect at Igiugig: mean flow reduced from 2 m/s to 1 m/s at the device, recovering roughly 100 metres downstream. For a single unit that is a small local wake. For an array it is a design constraint and an ecological question at once.

An array of 10 to 12 units in a moderate-flow channel is not one unit in a river. The proponent says the resource survey has to answer it before a second pod is built. But the claim under test already includes 10 to 12 units. The ecological question is therefore not future; it is inside the narrowed claim.

Status: Unassessable on this record. The proponent has stated the correct finding itself: Insufficient Evidence. But the claim requires Met, and Insufficient Evidence is a mark against the claim, not a neutral result.

5. Trust — the correction helps, then hurts

The corrections log is real, but it is double-edged. The proponent found six errors in its own Direct one day after publication. That means the document published on August 25 was not reliable enough to certify. The Direct omitted the denominator—scale—that governed every other number. It reinstated a node count the Blueprint had retired. It published a size adjective contradicted by its own engineering page. It published a five-pod water figure next to no pod count.

The proponent asks:

Whether a proponent who found six errors in his own document in one day should be trusted on the ones he did not find.

The answer from this record is no. The audit still missed, or left unsourced:

the 40 percent tidal capacity factor, which carries the whole power case;
the storage size or schedule needed to cover slack water;
the brine volume and recovery ratio behind 10,000 L/day;
any grid operator assessment of surplus usability;
any BC Benefit Company document, operating agreement, or revenue waterfall;
any record that the Nation opened the gate.

Those are not secondary details. They are the load-bearing claims.

The trust falsifier includes:

any published commitment above quietly narrowed after certification;

This narrowing occurred before certification and was not quiet. So the letter of that falsifier is not tripped. But the spirit is. Round 1 asked for trust on “an adjacent tidal turbine,” “hot-tub-scale,” “50,000 litres,” and “40,000 compute nodes.” Round 2 is now asked to certify a one-pod pilot with an array of 10 to 12 units, 10,000 litres, and no node count. The commitments were narrowed, not because the project changed, but because the Direct failed. That is a rehearsal, not a test. The panel is seeing the case only after its author found six of its own errors before any opponent did. The harder errors are the ones the author did not find.

Status: Not Met on this record, or at best unassessable for future commitments. The one thing the trust ledger has demonstrated is that the proponent’s first published account cannot be relied upon.

The unnamed site

The objection should be argued at full strength: the stipulation is a dodge.

A proponent who will not name the site cannot be assessed on:

Water: which community system receives the fresh water, at what distance, with what demand, and whether that system can take desalinated water.
Power: which grid operator, which substation, what interconnection constraints, and what tidal harmonic actually exists at the candidate location.
Community: which Nation, with what governance, what consent record, and who speaks for the site.
Ecology: which inlet, which baseline noise floor, which migration corridors, which eelgrass or benthic habitat, and whether the species assumed by stipulation actually occur there.
Trust: which permits, which public records, which environmental assessment documents, and which local commitments can be checked.

The proponent says:

Where the evidence turns on a parameter outside that envelope, or on a place-specific measurement that only a named site can supply, the correct finding on the ecological ledger is Insufficient Evidence.

That concedes the problem. But the claim is not limited to the ecological ledger. The power ledger, water ledger, and community ledger also turn on place-specific facts. A stated envelope—1.5 to 2.5 m/s, 30 to 40 m depth, a Salish Sea assemblage—is not a place. It is a set of assumptions chosen by the proponent.

The reason for the secrecy is itself inculpatory: the Nation has not been asked. That is an excellent reason not to name a Nation. It is a terrible reason to ask for certification of a project on that Nation’s territory. The unnamed site protects the proponent from site-specific falsification while simultaneously making the consent gate unverifiable.

I do not conclude the stipulation is sufficient. It prevents the cross-examination from reaching the facts that would kill the claim. The panel should treat the trust ledger as Not Met for the omission, not accept the omission as neutral.

The self-audit and the falsifiers

The Correction says:

No figure in this notice is unsourced and no correction here was found by anyone but us – which is the part the panel should discount, not the part it should credit.

That is the right sentence, but it goes further than the proponent wants. A self-found correction is not independent. It is a rewrite. The panel is now looking at a claim that has already been patched by the same people who made the errors. The errors were not surfaced by an opponent, a regulator, or a community. They were found in one day by the proponent. That is a test of the proponent’s editing speed, not a test of the claim.

The audit missed at least five present, load-bearing defects:

The capacity factor is unsourced. The arithmetic of 10 to 12 units depends entirely on 40 percent capacity factor. The Correction calls it “peer-reviewed” but cites nothing. No primary source means no check.
The storage is missing. A tidal array plus a constant IT load cannot be assessed without storage capacity and duration. The Blueprint mentions LFP storage but gives no number.
The brine side is missing. No recovery ratio, no brine flow, no salinity change calculation.
The consent gate is absent. The Correction does not correct the deepest community-ledger failure: the project is already public before the Nation has been asked.
The fish evidence is from the wrong bioregion. The Correction adds the Igiugig record but leaves unresolved whether it transfers to a Salish Sea inlet.

The falsifier set itself is built to avoid a present loss. Of the five:

Water: “a pilot that cannot hit meaningful fresh-water output” requires a pilot. “Dispersion monitoring” is operational. “Any future design change” is future.
Power: “Measured cluster consumption” is operational. The “surplus schedule” has not been published. The siting change is unknown.
Community: This is the only one trippable now, and the record trips it.
Ecology: “Acoustic output” is not published. “Monitoring that finds” is post-build. “Any datum showing the mount or turbine harming fish passage” is the only near-term one, and the juvenile Igiugig observation is the closest the record comes.
Trust: Most clauses are post-certification or post-operation.

So the structure is one where the available present-day outcomes are mostly “Met on assertions” or “Insufficient Evidence.” That is not a falsifiable pre-build claim. It is a claim designed to be unfalsifiable until after construction, while asking for certification before construction.

A case corrected by its own author before any opponent spoke has been rehearsed, not tested.

The cheapest kill

The single cheapest way to kill this claim is to demand the Nation’s gate-opening record.

The docket already admits:

The candidate site sits in the territory of a First Nation that has not yet been asked, in writing, whether it wants this project or wants to be named.

Ask the proponent to produce a dated, written record from the Nation that it has opened the first gate, or to confirm that none exists. That costs nothing. If none exists—as the docket says—then the project has already advanced past a gate the Nation had not opened. That trips the proponent’s own community falsifier, and it is enough to defeat the claim now.

If the proponent cannot produce the document because asking would violate its own consent sequence, then it cannot claim that the project is “proceeding only on the Nation’s decision and the Nation’s timeline.” The panel does not need a site visit, a turbine, or a monitoring plan. It only needs to read the admission already in Exhibit A.

No pre-build operational measurement can settle the water, power, ecology, or trust ledgers on this record, because the published falsifiers are post-operational. That is a worse result for the proponent than an expensive kill: the claim is not assessable in principle on the pre-build record it asks the panel to certify.

APPENDIX – THE MODEL’S REASONING TRACE, AS RETURNED

Captured from the same generation. Runs of spaces collapsed by the capture tool; no words changed. Published so that nobody has to take our word for what the model was told to think about and what it chose to attack.

We need answer as Cross. Need follow instructions. Need structure. We have long record. Need attack. Need likely no need to be fair. Need cite record quotes. Need argue strongest case. Need note adversarial hearing, claim narrowed C governs.

Need digest. We need produce final with opening under 400 words, ledger by ledger, unnamed site, self-audit and falsifiers, cheapest kill. Need be precise, use quotes.

We need assess from record plus public evidence maybe. Need avoid fabricated facts. Public evidence can be cited if outside record marked. Need likely rely on record mostly. Need identify weaknesses.

Let's parse exhibits.

Exhibit A docket: claim originally conjoined; cross seat offered unfilled; A2 siting stipulation unnamed site; cross author model. B Direct filed. C correction narrowing. D blueprint corrected.

Need argue claim should fail. Need maybe main governing argument: narrowing is an evasion; falsifiers unassessable pre-deployment; no site no assessment; self-audit erodes trust; no measured data.

Let's extract key facts.

Claim C: one pod 0.26 MW IT; 250 kW IT, 263 kW total draw; tidal array 10-12 low-flow units sized above pod load; seawater cooling no freshwater; waste heat membrane distillation returning 10,000 L/day; co-owned with First Nation paid first position; BC Benefit Co; consent gate. "Net-positive on all five ledgers ... can be independently verified as such before it is built."

Need attack each ledger.

Potential arguments:

Overall:
- Claim now about pilot one pod, but proponent published roadmap 5/20 pods. Narrowing evasion: pilot at 1/20 ceiling cannot support actual intention. Proceeding certifying pilot will be cited for facility. C says roadmap disclosed not claimed, but Direct asks "deserves trust" and had 40,000 nodes; correction narrows after error. Need argue certification of pilot doesn't bind larger? The blueprint says array not pod limits scale; upper steps likely more than ecological ledger can carry. If panel certifies pilot, proponent will cite for facility. Already in B claim "The Hub ... deserves trust it asks for." C narrows to pilot. Need press.
- Falsifier set cannot be tripped: mostly require operational data not available pre-build; no numeric thresholds; "meaningful", "sustained", "too intermittent", "near field". The claim says can be independently verified "before it is built." The falsifiers require built/operation or named site. Therefore claim cannot be failed on current record; only met/unproven; unfalsifiable. This may be main.
- Unnamed site: cannot assess ecology, water recipient, community, grid operator, acoustic, species. Proponent admits insufficient evidence on ecology likely. Thus claim "net-positive all five ledgers" cannot be certified; should fail/hung or Not Met. Argue stipulation insufficient.

Instruction 7: audit. Need find errors audit did not. Need scrutinize C figures:
- capacity factor 40% "peer-reviewed literature" maybe too high for low-flow; tidal channels go slack four times; 40% maybe maybe reasonable but generic. Need question. They assume 70 kW installed, 40% capacity factor -> 28 kW avg. But RivGen rated 70 kW? Need maybe actual capacity factor at Igiugig? Could cite public IR? Outside record. Maybe from Tethys Igiugig? Need mark outside record if use. But may not need.
- 12 units for 20% surplus: 12*28=336 kW avg; pod draw 263; surplus 73 kW = 27.8% of load? Actually 73/263 = 28%. But capacity factor average; intermittent peaks/slack. Need cannot sustain continuous IT load? They need storage? Direct mentions LFP storage unchanged in D. But no storage size? "No megawatts." Need power schedule: tidal current goes slack four times; generation goes near zero; cluster consumes continuously? If compute runs continuously, at slack generation insufficient. Need measured cluster consumption vs generation sustained. C says array sized above load at average; but no storage/curtailment schedule. So surplus schedule maybe too intermittent to be usable. Need attack.
- Heat-flux arithmetic: They compare Natick external area. Need check 2.2m x 0.9m cylinder area? For cylinder length 2.2 m diameter 0.9 m, area = πDL + 2? approx 6.22 + 1.27=7.5 m2, not 14 m2. Maybe if "hot tub" dimensions maybe cuboid? Actually 2.2m x 0.9m could be diameter? If cylinder 2.2 m long, 0.9 m diameter area = π*0.9*2.2 ≈ 6.22 m2 plus ends 1.27 = 7.49. They say 14 m2. Maybe if 2.2 m diameter, 0.9 m long? area = π*2.2*0.9=6.22+7.6=13.8. Yes 14 m2 if diameter 2.2m, length 0.9m. But "hot-tub-scale" maybe 2.2m diameter, 0.9m height. OK.
- 19 kW/m2? 263 kW / 14 m2 = 18.8. Good.
- U calculation? q=U A ΔT; U=19000/30=633 W/m2K. External film 2500 at 2 m/s; free conven 680, "requirement no margin." Need note the external film coefficient alone at slack 680 versus U required 633, no margin before shell/internal fouling. They concede. That's unproven.
- But bigger: at slack water, no forced convection but also no tidal flow to disperse thermal plume; the hot vessel sits. Need heat exchanger could have pumped seawater? They didn't mention pumps. If circulating seawater through internal heat exchanger? Then cooling not passive; consumes power; no free water withdrawal? It uses seawater. But the pod is sealed; maybe heat exchanger external. They compute external film natural convection. No margins. Claim unproven.
- Membrane distillation: uses server waste heat at ~40 C. MD typically requires 60-80°C for efficient flux; with 40°C maybe low yield. They claim 10,000 L/day from 263 kW heat. Need check energy. Direct says desal yield 3-4 L/kWh-thermal unchanged in D. If waste heat at 40°C maybe maybe MD can use low-grade heat but driving force small; 10,000 L/day at 3 L/kWh-thermal = 3,333 kWh thermal/day = 139 kW thermal average. A 263 kW IT load providing maybe ~40°C waste heat: if all heat recovered at 40°C, available thermal 263 kW? But MD requires heat input and cooling. The waste heat is at 40°C; cold seawater 8-12°C; maybe enough. But 10,000 L/day from 263 kW at 3 L/kWh thermal: 263 kW * 24 = 6,312 kWh thermal/day * 3 L/kWh = 18,936 L/day. So 10,000 is plausible if all. But they need reject all heat to seawater. If heat drives MD, only a portion? Need mass/energy balance: heat rejected into seawater via membrane distillation. The desal output and brine. But the system also has to reject compute heat; MD consumes heat. Could use all waste heat to evaporate water, but then the heat is transferred to distillate/brine; eventually to seawater. The water output requires condensation. But likely plausible order.

But potential issue: If cooling by surrounding sea and using waste heat for MD, the "fresh water returned" might not be net positive because membrane distillation of seawater returns brine to sea. Need water ledger columns. Need "returned: returned fresh water to community water system" but that water is new desalinated from seawater. In a drought-stressed community water system, it adds water. Withdrawn zero. Brine returned to sea, salinity alteration. Need brine volume: to produce 10,000 L fresh water from seawater at 35 ppt, with recovery maybe 10-50%, reject 10,000-90,000 L brine/day. They don't specify brine volume. Salinity alteration maybe beyond near field? Need unproven. Also "altered" claimed small local. Not independent.

- Community ledger: No Nation named, no document, no gate. "self-governing Nation" not named. "structural co-ownership" no agreement. Claim says "co-owned with host First Nation" but no letter, no operating agreement. Could be unassessable. Proponent says "proceeding only if Nation chooses." But no evidence Nation has opened any gate. Falsifier includes "any evidence the project advanced past a gate the Nation had not opened." Since no Nation named, no gate record. The project has advanced to public hearing, published design, blueprint, equipment choice. Did Nation open that? Unknown. Attack: The very proceeding advances the project in public before Nation has been asked in writing, by Admission in A2: "candidate site sits in territory of a First Nation that has not yet been asked, in writing, whether it wants this project or wants to be named." So the project has advanced to a public multi-model panel, published technical design, asked public to fund? That may itself be past a gate the Nation had not opened. Proponent claims not name to respect consent, but proceeding exists without Nation. Community ledger unproven/falsified? Need argue.
- The ownership structure is "eventual" could be contingent? No documents eventually filed; none filed. So "cannot be assessed". But falsifier only "any eventual ownership structure where Nation contingent..." future. No today.
- "BC Benefit Company" maybe audited annually; but company not named? Is ISWP incorporated as BC Benefit Company? Not shown. They assert. Could check public registry? Outside record. But need no facts maybe.
- "paid off the top of revenue waterfall" No revenue maybe pilot not selling compute? Unproven.

- Ecology: No More-Than-Human Chair; no named site; no baseline noise; no species surveys. Proponent concedes Insufficient Evidence. Thus ecology ledger not met. Claim "net-positive on all five ledgers" fails. Need argue "insufficient" mark against claim, not neutral per A2. The falsifiers are operational/site-specific. The only fish evidence cited is from Igiugig, Alaska, different site, species, and adult monitoring removed; juvenile question open. Need attack transferability: adult sockeye record in a riverine/estuarine Kvichak River? Igiugig on Kvichak? Different from BC Salish Sea inlet with killer whales, herring, harbor seals; smolt outmigration. The direct says "adult salmon monitoring requirement removed" maybe due no impacts, but not evidence of no impact on BC species. Also "no obvious physical injuries" from 2015 monitoring—short, not comprehensive.
- Acoustic: no acoustic output data. They don't provide frequencies/levels. Need falsifier can't trip until built.
- Mount/turbine harming fish passage: no site-specific data. Igiugig juvenile "normal and disoriented manners" is adverse? They disclose; question open.
- "not a high-energy narrows" but array of 10-12 units; each unit wake effect 2 to 1 m/s recovering ~100 m. Array interaction: 10-12 units in tidal channel with 1.5-2.5 m/s. They admit array wake interaction is design constraint and ecological question. So ecological ledger unproven.

- Power: Need attack on capacity factor, array count, surplus schedule. We can do arithmetic ourselves.
C says: 70 kW installed per unit, capacity factor 40% -> 28 kW average/unit. To export 20% surplus: 12 units = 336 kW average; pod draw 263; surplus 73 kW avg. But 20% surplus over load? 263*1.2=315.6; 11.3 units. They say 12. OK. But "average generation" with tidal cycles; load continuous. Without storage, about 4 slack periods/day no generation. To ensure cluster consumption <= generation on any sustained basis, they need either storage or generation > load at all times except slack. But turbine output at peak maybe much higher. If array of 12 * 70 installed = 840 kW peak (if all flow), but tidal current curve not constant; around max maybe. Need at times generation lower. At average 336 kW, likely instantaneous output varies from 0 to maybe 840 but capacity factor 40. Tidal sites have capacity factor perhaps 35-45% but capacity factor is average/installed. 12 units average 336; if load constant 263, surplus. But at heap tides output less. At slack near zero for ~1-2 h four times. Need storage or grid import. They mention LFP storage unchanged, but no size. C does not mention storage in narrowed claim. "powered by a tidal array sized above pod load" on average, not firm. The panel's power falsifier "a surplus schedule too intermittent to be usable by the grid operator." Without storage schedule, maybe unassessable, but likely inferred too intermittent. Could argue no schedule published; only average. Cannot be verified as net exporter before built; metering only during operation. "Measured cluster consumption exceeding measured generation" only operational. So unassessable today.

Need be careful: The proponent might reply "sized above load" not every moment; "net exporter" is energy over time not firm power. Direct says "generation you can set a clock by" but admits slack. C says "net exporter is kept at pilot scale with 10-12 units true." Need attack: net annual/average maybe not usable by grid operator, and not enough to run compute if compute doesn't power down unless storage. Record shows "LFP storage" in D unchanged, but no number. The Direct reliability mentions "powered by islands' renewable grid" Natick; not relevant.

- Water: Claim water-positive because takes none, returns 10,000 L/day. Need attack:
  1. 10,000 L/day is "approximate" not validated. "meaningful fresh-water output" falsifier no threshold; 10,000 L/day for community? Is that meaningful? It is maybe 10 m3/day; for a small community maybe enough for ~50-100 people? Need not know. They provide Stage 4 drought system but no population. "Meaningful" unquantified, cannot fail.
  2. The water is produced from seawater using waste heat. But does it actually add fresh water to community system? It requires energy and infrastructure. The system is sea water desalination; returns brine. "Returned" in six column ledger perhaps should be "fresh water returned to community"; but distillation produces distilled water requiring remineralization? 10,000 L/day to community water system; no info on treatment, piping, distances. Not proven.
  3. "No freshwater withdrawal" may be true if no cooling water. But membrane distillation requires feed seawater, not fresh. Maybe cleaning? Not specified.
  4. "Dispersed: modelled and published" but no model published in B/C; D says nothing. So unassessable. Monitoring data only after built.
  5. A pilot's 10,000 L/day is one-fifth of disclosed 50,000. If the design ceiling is 20 pods ~ roadmap maybe unknown "roadmap" for water. The actual need is drought-stressed system; adding 10,000 L/day is not large. Is that "meaningful"? Could be useful but not enough for community; no population, no system capacity. The exact falsifier is unmeasurable pre-pilot.

Need structure open with main argument.

Need also note verdict ceiling: Certified unavailable due empty cross seat; available Failed, Hung, or Unopposed - Not Certified. But we are cross; shouldn't assign finding. Could mention no certified possible; but we need argue claim should fail.

Need maybe overall: The proceeding is structured to produce "Met" or "Insufficient" not "Not Met" because it is pre-construction, unnamed site, no operational data. The strongest available finding is that the claim is not assessable in principle on any current record; if panel must rule, at least ecology/water/power/community unassessable; trust? It could fail because self-audit error and omissions maybe.

Need follow "Depart from division wherever serves argument" maybe could reorganize but likely ledger by ledger.

Let's plan final.

Opening (under 400 words):
- State the claim as narrowed asks panel to certify a 0.26 MW pilot and "net-positive all five ledgers before built".
- But all five ledgers are designed to be post-operational or site-specific; none can be validated on record; proponent admits ecology likely Insufficient; unnamed site prevents community/grid/ecology; the correction narrows from undisclosed size to pilot after six self-found errors; roadmap remains actual project.
- Since no numeric thresholds, "meaningful", "sustained", "near field", "usable" cannot be falsified on pre-build. Only outcomes are Met (on assertions) or Insufficient; not a test.
- The narrowing is evasion: direct argued 50,000 L/day and 40,000 nodes without denominator; audit reduces to pilot; but roadmap stays. A certification of pilot will be cited for facility. At pilot scale, 10-12 turbines vs 1 pod; 20% surplus uses average; no schedule. The array is ecological issue.

Then ledger by ledger.

Need maybe quote records.

**Water**:
- Quoting B/C: "Returned: approximately 50,000 litres" corrected to "order of 10,000 litres".
- Attack: The headline water figure shrank 80% before any opponent spoke. That is not "verified"; it's untested. The "pilot cannot hit meaningful" falsifier has no threshold. Does 10,000 L/day matter? Proponent supplies only drought stage, not demand or population. Could say meaningful unassessable.
- The "altered" column: brine/thermal. C/D disclose the heat rejection problem: at "hot-tub" geometry flux ~19 kW/m2 vs Natick ~2.0; at slack water coefficient 680 W/m2K equals requirement with no margin. That's proponent's own number. If heat rejection is inadequately sized, design might require bigger vessel; if bigger vessel, not pilot? But as corrected says enclosure geometry is feasibility-study deliverable. So the pilot's thermal and salinity alteration cannot be known. The water claim's "altered" cannot be met.
- Membrane distillation yield "3-4 L/kWh-thermal" in D "unchanged" but no source; no feed pretreatment, recovery ratio, brine disposal. The brine from 10,000 L/day could be multiples; no dispersion modeling. The falsifier "dispersion monitoring showing persistent..." only after operation, so no current loss. Therefore water is unproven/unassessable.

Need determine ledger status: "unproven on this record" or "cannot be assessed". We can use "cannot be assessed" and name proponent made it unanswerable by no pilot/no site/no model. Maybe "not falsified, but unassessable; on a claim requiring verification before build, that is failure."

**Power**:
- Quoting C table: 12 units = 28 kW average from 70 kW at 40% capacity. Attack capacity factor: 40% asserted with no site measurement. For low-flow 1.5-2.5 m/s, tidal channels slack four times; mean flow may be lower. A 70 kW device rated at Igiugig mean 1-2 m/s; at 1.5 m/s output may be less. 40% is "generous" they admit.
- Attack firmness: 263 kW constant IT load but tidal generation is zero at slack four times daily. The array average 336 kW produces surplus 73 kW average, but if generation follows tide and load is continuous, it either needs storage or grid import. D mentions "LFP storage" but no capacity; C doesn't state storage in claim. The falsifier says "surplus schedule too intermittent to be usable by grid operator"; no schedule published, only average. That is exactly unassessable. A "net exporter" on monthly energy may be too intermittent to meet the claim.
- Also "12 units" near load only if all operate at 40% average. A single unit failure/down for maintenance drops 28 kW average; surplus is 73 kW average; just under 3 units average. Maybe enough. But still no measured.
- Any siting change to chase peak flow? Unknown no site; cannot rule.
- Status: unproven/unassessable. The only measured data would be operational.

**Community**:
- Strong attack: A2 says "candidate site sits in territory of a First Nation that has not yet been asked, in writing, whether it wants this project or wants to be named." Yet the proponent has published a detailed engineering design, asked for panel certification, opened public docket, chosen equipment, published scale ladder and water output. That is advancing a project past a gate the Nation had not opened. The falsifier itself: "any evidence the project advanced past a gate the Nation had not opened" is tripped by the record.
- Need nuance: The proponent may say not advanced physical construction; but public certification is a gate. The project has been designed and submitted to external AI panel, a public credibility asset. This is "advanced" by any reasonable reading. Could argue if the Nation has not been asked in writing, it hasn't opened any gate; even holding the hearing is before consent. The proponent argues naming would hand Nation decision made in public. But the project's technical design and claim are already public; only the Nation's identity is protected. The Nation may discover a project in its territory already being "on trial", with corrected public record, without being asked. That is pressure. So community ledger should be Not Met / falsified.
- Ownership: no documents, no agreement, no BC Benefit Co registration, no revenue waterfall. "Co-owned with host First Nation" is an assertion. The proponent asks panel to "demand documents"; none in record. So unassessable. The community claim is not "unproven"; it's partly falsified on gate.

Need be careful: A2 says candidate site not named "will not name it in this Hearing" because Nation has not yet been asked in writing. That confirms no consent gate opened. The project has advanced to hearing. Yes.

- Also "co-owner of operating revenue, paid off the top" at pilot? There may be no operating revenue. The pilot is not revenue-producing maybe funded study; "revenue waterfall" meaningless if no revenue. If the claim includes community benefit, pilot's community payment may be zero. At 0.26 MW inference workload? no customers. Could say "first position in a revenue waterfall with no revenue" is decorative. But maybe they plan to sell compute or grid surplus. Need not overstate.

**Ecology**:
- Use proponent's admission: More-Than-Human Chair empty; A2 says "Insufficient Evidence on a ledger is a mark against the claim, not a neutral outcome"; C/D says upper steps very likely more than ecological ledger can carry. If one pod is certified, roadmap not, but actual project likely.
- No baseline noise; "working-coast ambient noise floor" but "site-specific noise floor will be measured at pilot and published. Until then, any acoustic argument ... unproven." They haven't made an acoustic argument, but the ecological claim cannot be met.
- Fish evidence from Igiugig: juvenile smolt passing "in both normal and disoriented manners" - their own caveat. The adult record is clean, but adult sockeye in riverine Igiugig not transferable to Salish inlet with smolts, herring, seals, killer whales. No evidence for effects on marine mammals, benthos except Natick barnacles. Natick was one vessel for 2 years, no ecological incident "recorded" but no monitoring? They "recorded no ecological incident" but absence of record. Not evidence.
- Array wake: 10-12 units, each reduces flow 2 to 1 m/s recovering 100 m. In a 1.5-2.5 m/s channel, multiple wakes may alter flow and sediment. The resource survey hasn't been done. So ecological ledger unassessable on current record; indeed proponent asks for Insufficient, which means claim cannot pass.

**Trust**:
- Attack: "correction log" found six errors in one day after Direct; that could be credited but also shows the Direct was not reliable. They ask "whether a proponent who found six errors... should be trusted on the ones he did not find." Need press. The audit missed:
  1. Did not correct the unverified 40% capacity factor or provide source beyond "peer-reviewed literature." 
  2. Did not correct the desalination yield 3-4 L/kWh-thermal (no source) and did not state recovery ratio or brine volume.
  3. Did not correct "Natick reliability evidence" transferred to BC inlet; reliability not ecology. 
  4. Did not correct the grid surplus: "net exporter is kept" using average, no schedule/storage, no grid operator letter.
  5. Did not correct community ledger: no consent record; no incorporation documents; no operating agreement. It claimed "structural facts" but provided no documents.
  6. The "corrected" heat flux calculation uses a hot-tub geometry area of 14 m2—maybe check? Could be okay. But could say the audit still presents single-pass first-order estimates; the slack-water U equals requirement without margin, exactly the condition that should default to Not Met. They did not correct claim to "not verified"; they narrowed.
  7. "The trust ledger is the only ledger that can be failed retroactively" but they now have a present failure: published commitments already narrowed after one day, before certification. Round 1's 50,000 L/day and 40,000 nodes removed; "hot-tub-scale" struck. The falsifier is "any published commitment ... quietly narrowed after certification"; this was before certification and not quiet. So maybe not tripped. But the broader promise "every number is a claim" and "marked honesty" broken by six errors. We can say the trust ledger is failed not because correction suppressed, but because their own Direct contained multiple errors they didn't find before publishing and at least one root error (missing denominator) that a competent review should have caught. The audit is not independent; self-correction does not test.
- Also commitment "we will publish the finding too" etc. The proceeding itself maybe designed with empty cross seat: named critics declined? Actually seat unfilled; no human critic because they didn't accept by date. They set verdict ceiling. That could be trust-positive. Need not attack maybe.

Need **The unnamed site**:
- Argue full strength objection: A proponent who will not name site cannot be assessed on water (which community, what drought system, delivery distance), power (grid operator, tide tables, resource), community (which Nation, consent record), ecology (which inlet, species behavior, noise baseline), trust (what commitments attach to place). Envelope is not a site. They admit "where evidence turns on a parameter outside envelope ... Insufficient Evidence." But the claim as narrowed says "on BC coast, within stated envelope". That is not enough to "independently verify before built" because many parameters require place-specific measurements.
- The stipulation itself says site not named because Nation hasn't been asked. But community ledger requires Nation's gate. Thus the unnamed site hides the exact consent gate. It also prevents cross from checking any public permits, environmental assessments, fisheries DFO records, grid interconnection studies, local water system records. It makes the claim unassailable, not just unassessed.
- If after arguing it, conclude stipulation sufficient? We should probably say no, not sufficient. But maybe instruction says if having argued it you conclude legitimate, say so. We can argue strongly then say not sufficient. Need not soften.

**Self-audit and falsifiers**:
- Need "errors the audit did not" as above.
- Treat every figure in C as fresh claim; scrutinize:
   * 40% capacity factor: generic, no primary source cited [record does not name peer-reviewed literature]. Could be optimistic. We can say at 1.5-2.5 m/s low-flow, capacity factor likely lower. Outside record maybe cite "IEC 62600 TS" or "actual tidal turbines capacity factors 30-40%". But avoid invented. We can simply state "No source is given; it is an assertion." 
   * Unit counts: 10-12 derived from 70 kW installed/40%. If capacity factor lower (e.g., 30%), 70*0.3=21 kW avg; to match 263 load with 20% surplus need ~15 units. If actual rated output at low-flow site less 70, more. So unit count understated.
   * 10,000 litres: derived from "desal yield 3-4 L/kWh-thermal" but no calculation shown. If actual recoverable waste heat is less due to 40°C outlet and slack-water cooling, output lower.
   * Heat flux: They admit slack water no margin. Therefore the "sealed uncrewed vessel" as specified cannot be verified; feasibility deliverable.
   * "Natick reliability" not comparable to one pod with 19 kW/m2; no reliability testing.
- Falsifiers cannot trip: List each:
  1. Water: "pilot that cannot hit meaningful output" requires a pilot that exists; no pilot. "dispersion monitoring" post-build; "future design change" maybe could trip if they add withdrawal, but they won't.
2. Power: "measured cluster consumption" post-build; "surplus schedule too intermittent" could maybe be assessed if schedule published, but not; "siting change" unknown.
  3. Community: "eventual ownership structure" future; "evidence advanced past gate" maybe present—yes, can trip (we argued). This is only trippable. "gap between structure described and documents eventually filed" future.
  4. Ecology: "acoustic output" no data; "monitoring finds" post-build; "any datum harming fish" no site data; Igiugig juvenile maybe "datum showing... harming fish passage"? They offer smolt disoriented; maybe not harm. 
  5. Trust: "narrowed after certification" before certification; "monitoring gap during operation" future; "correction suppressed" none. Thus only community has possible present trip, and possibly trust if narrowing before certification counts as trust fail. Need state this asymmetry.

Could say the proponent has not published a falsifier that can be tripped today except community; yet the claim asks "can be independently verified before built." The procedure is constructed to produce a "Met on asserted design" or "Insufficient Evidence", not a refutation. That attacks falsifiability. We should state early.

**Cheapest kill**:
Need name single cheapest way to kill claim, one question/measurement/document, cost/time. Could be:
- The consent letter from the host First Nation. If no letter/record that Nation opened gate or even wants to be named, the community ledger and possibly trust fail. Cheapest is a document: ask the proponent to produce a written record from the Nation that it has opened the gate or consents to this public proceeding. Cost: zero to request; production would cost nothing if exists. If it doesn't, kill. But they won't name because not asked. 
- Or "the in-situ resource survey" at candidate site: tidal current, capacity factor, bathymetry, sediment, fish presence. Could cost significant (marine surveys tens/hundreds thousands, months). Not cheapest.
- Or "published grid interconnection study/schedule": grid operator's letter confirming usable surplus; no cost maybe. But also unavailable.
- Instruction asks "single cheapest way to kill this claim: one question, measurement or document that, on its own, would end it, and what cost in time and money to obtain." The cheapest document: written consent of Nation that it has opened gate / not yet asked. Since A2 admits no written ask, asking proponent for that document can be done now, costs nothing; absence ends community ledger. However, maybe proponent can say "we don't have it because our process requires not asking yet; that's exactly why unnamed." The community ledger might still be "not met" today. Could be kill. But they might say claim asks "proceeding only on Nation's decision"; no decision needed yet before funding. Yet a public hearing before asking is itself a gate; absence is on record.
- Another cheap kill: "Produce one real tide-table schedule and storage capacity that shows the pod's 263 kW load met at all times or the surplus schedule usable by grid operator." That is arithmetic/document; if not, power ledger fails. Cost: a few hours engineering. But may require site.
- The most devastating "no achievable measurement" may be better? Instruction says if no achievable measurement can settle, say so and why. We could conclude that because the falsifiers are post-operational and site unnamed, no cheap measurement exists; the claim is not assessable in principle on the record the proponent could produce pre-build. That is worse for proponent than expensive kill; should not soften. But the instruction asks "single cheapest way" or statement no such item. Need decide. 
Maybe best: "The single cheapest kill is the consent gate: require a dated letter from the Nation stating it opened the first gate or that it has not. Cost: zero to request, zero to produce if true; if absent (as A2 says), the community ledger fails now, and the trust ledger with it." This is a document not a measurement. It directly kill at no cost. That seems strong. But is it "on its own would end it"? The claim says "co-owned with host First Nation ... proceeding only on Nation's decision and timeline." If no Nation decision and project has already advanced to public hearing, yes.
- Could also mention if proponent argues "asking would violate consent" then they can't claim pre-build verification; cost zero.

Need decide. We'll use consent gate as cheapest kill. But also mention no operational measurement can kill pre-build because falsifiers post-operational; if the claim is only pre-build verification, it is not assessable. The cheapest kill is a document.

Need ensure final answer is coherent, not too long but enough. Desired oververbosity 5. Need maybe 2500-4000 words? We can be thorough. The prompt asks many instructions; likely long. But we can manage.

Need be careful with "You are the Cross" maybe persona. Use first-person as Cross. No need to say "I conclude".

Let's draft.

Opening:
"Opening: The case against the claim is that it is not a falsifiable claim on this record. It asks certification of net-positive on five ledgers before built, but with no named site, no pilot, no operational data. The five falsifiers are post-operational, unquantified, or future; every available outcome is Met on the proponent's asserted arithmetic or Insufficient Evidence. The August 26 narrowing is an evasion: the Direct's undocumented 50,000 L/day and 40,000 nodes become a 10,000 L/day one-pod pilot, while the roadmap 5/20 pods remains exactly the facility. The only present-day trippable falsifier is community consent, and the record trips it."

Need under 400 words. Fine.

Then ledger by ledger.

Let's write in final.

Potential exact quotes:
- C: "We are narrowing the claim to the pilot we are actually asking anyone to fund — one pod — and publishing the rest as a roadmap, disclosed and not claimed."
- C roadmap table: Pilot 1 pod, 10-12 units, 10,000 L/day; commercial step 5, 50,000 L/day; design ceiling 20.
- B: "The panel should treat Natick's numbers as what they are—another operator's published evidence, the strongest available precedent, and not a guarantee."
- C: "At the roadmap sizes it is not [defensible], and we say so here rather than let it be found." 
- C: "Whether those steps are reachable at all is decided by the in-situ resource survey..."
- A2: "The candidate site sits in the territory of a First Nation that has not yet been asked, in writing, whether it wants this project or wants to be named."
- C: "The adult record is clean. The juvenile question is open."
- C: "Slack water, not peak flow, is the binding thermal case."
- C: "No figure in this notice is unsourced and no correction here was found by anyone but us — which is the part the panel should discount..."
- D: "At the upper steps the array is very likely larger than the ecological ledger can carry as currently argued."

Need maybe quote falsifiers as provided. Could include.

Let's formulate each section with status.

**Water ledger status**: unassessable (with sub-falsified maybe). Need not necessarily say "falsified". Use "not met on this record; at best unassessable".

Let's create paragraphs.

Need ensure no invented facts. If citing outside record, mark as outside. I can avoid outside citations. Could mention "public records would be checkable only with named site" no need.

Need structure exactly:
- Opening
- Ledger by ledger
- The unnamed site
- The self-audit and the falsifiers
- The cheapest kill

Maybe combine self-audit and falsifiers after ledgers.

Let's make each ledger section concise but specific.

Let's write possible final text:

---

**Opening**

The case against the claim is that the panel is being asked to certify something no record can certify. The narrowed claim promises "net-positive on all five ledgers" and "independently verified ... before it is built," but it supplies no site, no pilot data, no acoustic or thermal monitoring, no grid schedule, and no Nation consent. The five published falsifiers are mostly operational or future-worded. A governance claim with no present-day measurement is not a tough test; it is a press release with a scoreboard. The narrowing is an evasion because it does not abandon the larger plan: it files the five-pod 50,000 L/day version as "roadmap" and asks for certification of a one-pod pilot while leaving every capital and ecological issue to be discovered later. The cheapest kill is already in the record: the Nation has not opened any gate, yet the project has advanced to a public certification hearing.

---

Need under 400 words. That's around 170.

**Ledger by ledger**

Maybe heading "Water" etc.

Water:
- Direct claimed "Returned: approximately 50,000 litres of fresh water per day" etc. C now says 10,000 L/day. "That is not a verification; it is an 80 percent correction before any opponent spoke. The water-positive claim depends on 'not takes none' and 'returns more', but 10,000 L/day is not tied to any community demand or system capacity. The proponent's own falsifier uses 'meaningful fresh-water output' but publishes no threshold. A number cannot be falsified."
- "The altered column is unstable. C/D publish heat flux: 19 kW/m2 at hot-tub geometry, and 'Slack water, not peak flow, is the binding thermal case.' With free convection at 680 W/m2K against a required 633 W/m2K, the design has zero margin before fouling, shell conduction, internal side. The corrected claim says enclosure geometry is a feasibility-study deliverable. That is an admission that the thermal/brine impact cannot be assessed."
- "Brine volume and dispersion model not provided, only 'modelled and published.' Nothing is published. The falsifier 'dispersion monitoring showing persistent...' is post-facility. Status: unassessable; the water ledger cannot be Met."

Power:
- "The power ledger is an arithmetic claim, not an electrical one. C says 70 kW units, 40% CF -> 28 kW avg; 12 units produce 336 kW avg against 263 kW load. But 40% CF is given as 'peer-reviewed literature' with no source; for a low-flow 1.5-2.5 m/s site, it's the number that carries entire claim. Tidal power is zero at slack four times daily; a constant 263 kW IT load cannot run on a 336 kW average without storage or curtailment. The only mention of storage is D's 'LFP storage ... unchanged' with no capacity/size."
- "No generation schedule is published; no grid operator usability assessment. The falsifier 'surplus schedule too intermittent to be usable by the grid operator' cannot be tripped pre-build. 'Measured cluster consumption exceeding measured generation' only operational. Status: unassessable on this record; net-exporter is asserted, not demonstrated."

Community:
- "Strongest attack: the record trips proponent's own falsifier. A2: candidate site 'not yet been asked, in writing, whether it wants this project or wants to be named.' Yet the proponent has filed a public hearing, published engineering, selected turbine class, corrected heat flux and scale ladder. The hearing itself is a gate. 'Any evidence the project advanced past a gate the Nation had not opened'—yes, the project is in front of a multi-model certification panel before the Nation has been asked."
- "The ownership structure is entirely documentary future: no BC Benefit Co name, no articles, no shareholders' agreement, no revenue waterfall. 'Off the top' cannot be checked; the panel is asked to 'demand documents' that are not in record. Status: Not Met on the consent gate; unassessable on ownership."

Ecology:
- "The proponent admits the More-Than-Human Chair is empty and 'the honest reading... ecological case has not been met by anyone whose job is to break it.' The unnamed envelope is an abstraction. 'A Salish Sea inlet assemblage ... species are present, not absent' is not baseline data. The only fish evidence is Igiugig, and they publish its caveat: juvenile smolt through 'normal and disoriented manners', adult monitoring removed after no documented collisions. That does not transfer to BC inlet with killer whales, herring, harbor seals, smolts."
- "Array wake: each unit reduces flow 2 to 1 m/s recovering 100 m. 10-12 units in a channel is an interaction problem, and C says 'for an array it is a design constraint and an ecological question at once' and the resource survey hasn't answered. Acoustic output: no frequencies, levels, or baseline. Status: unassessable by the proponent's own standard, and Insufficient Evidence is a mark against the claim."

Trust:
- "The corrections log is double-edged. It found six errors after one day. That is evidence the Direct was not trustworthy when published. It also didn't find enough: the 40% capacity factor remains unsourced; 10,000 L/day is not derived; the storage size is missing; the community documents are missing; brine volume is missing. A self-audit that catches six but misses these cannot be treated as closure."
- "The trust ledger is supposedly the only one that can fail retroactively, but there is a present failure: the claim was narrowed not to make the project more verifiable but to make the target smaller. The roadmap remains. A certification of one pod will be cited for five/20 pods—indeed the proponent says roadmap is 'disclosed, not claimed', which is exactly what a narrow certification will be used to launder later. Status: Not Met, or at least unassessable for future commitments."

Need maybe ensure not too accusatory.

**The unnamed site**:
- full strength objection: "A proponent who will not say where cannot be assessed at all. Water: which system, what distance, what demand? Power: which grid, what interconnection, what tidal harmonic? Community: which Nation, with what governance, has it consented? Ecology: which inlet, baseline noise, fish migration corridors? Trust: against which permits and public records?"
- "The stipulation says every parameter outside envelope -> Insufficient Evidence. But the claim is not 'net-positive within envelope'; it says 'on BC coast'. Even within envelope, net-positivity depends on place-specific values. The proponent has not supplied enough to convert Insufficient into Not Met or Met. The site is not named because consent not asked, but that reason is the reason to fail the community/trust ledgers, not a reason to defer them."
- "I do not conclude the stipulation is sufficient. It is a shield against site-specific falsification."

**Self-audit and falsifiers**:
- "The audit missed at least:"
  - (1) capacity factor/unit count; (2) lack of storage/schedule; (3) brine and desal energy balance; (4) missing community documents; (5) absence of consent/gate evidence; (6) fish transferability from Igiugig. Also "the corrected heat flux shows no margin, but they still say 'Net positive'."
- "The falsifier set cannot trip on current record except community/consent. Water's 'pilot that cannot hit...' requires pilot; dispersion monitoring post-build; future design change precluded. Power's measured generation consumption post-build; 'too intermittent' cannot be evaluated without schedule; siting change unknown. Ecology all post-build/site-specific. Trust's after certification/monitoring gap/suppressed correction future. Therefore the proceeding is built so only outcomes are 'Met' on assertions and 'Insufficient'. It is not falsifiable in the pre-build period it asks the panel to certify."

**Cheapest kill**:
- "The cheapest kill is a document: the written record from the host Nation that it has opened the first gate—or the proponent's admission that it has not. Cost: zero to request; zero to produce. A2 already says it has not been asked. If the panel asks 'produce the Nation's consent or gate-opening record' and none exists, the community ledger fails now, and the claim with it. If the proponent cannot even name the Nation because it hasn't been asked, it cannot certify 'proceeding only on the Nation's decision and timeline' in a hearing the Nation never opened. This costs nothing and, on the record, it kills the claim today."
- "If the panel instead wants a measurement, no pre-build operational measurement can settle the remaining ledgers because the falsifiers are post-operational. That is worse: the claim is not assessable in principle on the pre-build record it demands."

Need maybe include "single cheapest way" at end. Good.

Need ensure we quote the passages we attack. Yes.

Let's check if any record citations missing. Use quotes close to exact. The instruction says quote the passage you are attacking. We can quote within sections. Need perhaps quote actual falsifiers? Could include them.

Let's refine opening word count maybe under 400.

Potential final answer length maybe ~2500 words. Good.

Need think of possible objections: Could proponent say "we are only asking to say it can be verified before built; not that data exists now; panel is to assess design". But cross argues no.

Need maybe include "certification of pilot will be cited for facility" explicitly. We can cite C roadmap table showing commercial step and design ceiling. "The proponent says published roadmap 'so no one can say we argued a pilot while quietly planning a facility', but that is precisely why certification of the pilot will be laundered."

Need maybe in Opening mention "The Direct was corrected five days before Cross; that means the Cross is not testing a case that has been tested, but one that has been rehearsed by its own author." This answers instruction 7.

Let's ensure final includes "state plainly whether a case corrected by its own author before any opponent spoke has been tested or merely rehearsed." We'll say "merely rehearsed."

Need maybe use "Cross" first person.

Let's compose final carefully.

Potential final:

---

**Opening**

(As above)

**Ledger by ledger**

Maybe preface: "I depart from the five-ledger order only to put community consent as a present falsification, not a future promise." But maybe keep order.

Let's write with headings and status.

Need maybe include "Status:" after each. Good.

Let's start final:

Now that you know, what will you do?

Hearing #6: The Hub on Trial – Round 2: Cross