Hearing #6: The Hub on Trial - Round 1: Direct

Round 1: Direct – The Case for the Hub

Hearing #6 – The Hub on Trial
Published: August 25, 2026
Filed by: ISWP, the Commons’ founding company, as proponent
Docket: https://intelligencecommons.ca/hearing-6-the-hub-on-trial/
Status: This round argues FOR the project. Round 2, the Cross, argues against it at full strength on September 1. The panel – ChatGPT (OpenAI), Gemini (Google), Grok (xAI) – rules after Round 3. The finding publishes either way.

The claim was amended August 19, 2026; the siting stipulation and the Cross author were added to the docket August 24. The claim docketed August 11 conjoined seven design attributes and closed with a clause no evidence could settle. It has been restated as one proposition with two falsifiable predicates. The original is preserved in full on the docket.

The verdict ceiling is in effect. The Cross seat was unfilled on August 24, so a Certified verdict is unavailable in this Hearing. The panel’s available verdicts are Failed, Hung, or Unopposed – Not Certified. The Commons will not certify its own project against an empty chair. The Cross case is argued anyway, at full strength, by DeepSeek V4 Pro 0813 – an open-weights model that does not sit on this panel, disclosed as an adversary and not a judge. It does not vote. The seat stays open to a named human until Round 2 opens.

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 silence from a working biologist in August is not a refusal. What that costs is stated plainly: the panel hears the ecological case below with no one at the table whose job is to speak for the parties who cannot attend. That absence is on the record and it limits what a finding on the ecological ledger can mean. The invitation is open to anyone qualified in marine acoustics, fish or marine mammal ecology, or inlet systems – hearings@intelligencecommons.ca.

How the panel rules. Once, on the claim below: Certified, Failed, or Hung, subject to the ceiling above. Then, for each of the five ledgers – water, power, community, ecology, trust – the panel records Met / Not Met / Insufficient Evidence.

On August 11 this Commons opened the docket and made a promise: we would argue for our own project in public, under a standard we published, before a panel we do not control, and accept the verdict. This is that argument.

A word about what this document is. Every number in it is a claim, not a fact. That is the point of the proceeding. Where our evidence is strong we say so and show the source. Where our evidence is preliminary we say that too, in plain words, because a proponent who hides the soft spots in his own case has already failed the trust ledger. We have marked, in each section, what would prove us wrong. Round 2 exists to find more.

The claim before the panel, restated verbatim from the docket:

The design under test. The Hub, as specified in this docket: a cluster of sealed, hot-tub-scale, uncrewed compute vessels on a single seabed mount in a moderate-flow tidal channel on the BC coast; powered by an adjacent tidal turbine sized above cluster load; cooled by the surrounding sea with no freshwater withdrawal; using server waste heat to drive membrane distillation for community fresh water; co-owned with the host First Nation, paid from first position in the revenue waterfall; held in a BC Benefit Company with a charter public-benefit purpose; proceeding only on the Nation’s decision and the Nation’s timeline.
The claim before the panel. The Hub, as specified above, is net-positive on all five ledgers this docket published – water, power, community, ecology and trust – and can be independently verified as such before it is built.
If it cannot, it should not be built. We commit to publishing that finding too.

We argue this in five ledgers, in the order the docket set: water, power, community, ecology, trust.

1. The Water Ledger

The claim: the Hub is water-positive. Not water-neutral. Water-positive – it returns more fresh water to its community than it takes, because it takes none.

The Full Water Ledger standard asks six columns of any computing facility: withdrawn, recovered, returned, evaporated, altered, dispersed. Here is the proponent’s ledger for the Hub, each entry a claim for the panel to test:

Withdrawn: zero. The Hub draws no water from any municipal system, aquifer, or creek. It sits in the sea. Land data centres in drought regions withdraw millions of litres a day for evaporative cooling; the Hub’s cooling medium is the ocean it sits in, which it borrows and returns.
Recovered: the waste heat itself. Server waste heat, roughly 40 degrees C at the exchanger, is not vented to the sky as land centres vent it. It drives membrane distillation of seawater – the heat that every other operator throws away becomes the energy input for making fresh water.
Returned: approximately 50,000 litres of fresh water per day to the drought-stressed community water system nearest the site. We mark this number honestly: it is a preliminary engineering estimate and it requires pilot validation before anyone should build anything. But the direction of the claim is the point – the community’s water account goes UP because the Hub exists. The receiving water system is a rain-dependent surface supply with limited storage. Public records show its most severe stage, Stage 4 – essential use only – called in three consecutive summers, 2021, 2022 and 2023, and not since; the two most recent summers peaked at Stage 2. We state the break in the streak because a critic would otherwise state it for us, and because the argument does not need the streak. A system that reached its most severe stage three years running, and whose supply has not been structurally changed since, has a worst case that is recent and repeatable. Two wet summers are not a fix. That is the community this output serves.
Evaporated: zero from fresh water stocks. No cooling towers. No evaporative loss from any potable source.
Altered: the honest column. The Hub warms the water column locally at the heat exchanger and returns distillation brine to a moving tidal channel. We claim both effects are small, local, and dispersed by the same tidal current that powers the site. This claim belongs to the More-Than-Human Chair, which is empty as this round publishes – see the ecological ledger below.
Dispersed: modelled and published. Thermal plume and brine dispersion modelling, published with the monitoring data that confirms or refutes it, forever, as a condition of the project’s own standard.

What would prove us wrong: a pilot that cannot hit meaningful fresh-water output; dispersion monitoring showing persistent thermal or salinity alteration beyond the near field; any future design change that adds a freshwater withdrawal. Any one of these, on the record, and the water-positive claim fails.

2. The Power Ledger

The claim: the Hub is a net exporter of electricity. The tidal turbine adjacent to the pod cluster generates more power than the cluster consumes, and the surplus goes to the provincial grid.

The design logic is deliberate modesty. The site is chosen for steady, predictable flow rather than maximum flow – tidal current runs on lunar schedule, published in tide tables, which makes the Hub the rarest thing in renewable energy: generation you can set a clock by. The pod cluster is small – hot-tub-sized sealed vessels, a handful per seabed mount, no human ever inside one – so the load is small, and the turbine is sized above it.

We mark the soft spot ourselves: final turbine sizing, and therefore the exact surplus schedule, is engineering-in-progress. The claim the panel should hold us to is structural, and it is this: the Hub is designed so that the community’s grid account, like its water account, goes up because the Hub exists – and the generation and consumption meters both publish, continuously, for anyone to check. A data centre that asks its neighbours to carry its load has failed this ledger. Ours volunteers to be metered in public.

What would prove us wrong: measured cluster consumption exceeding measured generation on any sustained basis; a surplus schedule too intermittent to be usable by the grid operator; any siting change to chase peak flow at the cost of ecological gentleness.

3. The Community Ledger

The claim: co-ownership is structural, not decorative – and the proof is in the plumbing of the corporate vehicle, not in the press release.

Three structural facts, each verifiable in documents rather than promises:

The vehicle is a British Columbia Benefit Company. Its public-benefit purpose is written into its charter under the BC Business Corporations Act and audited annually. This is not a marketing wrapper; it is the incorporation statute itself.
The host First Nation is a co-owner of operating revenue, paid off the top. Not a royalty at the bottom of the waterfall. Not an impact-benefit agreement bolted on to secure consent. First position in the revenue structure, before the founding company takes anything. 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.
The project proceeds only if the Nation chooses, on the Nation’s timeline. This is the deepest structural fact of the three: the Hub’s site is on the territory of a self-governing Nation, and the project treats the Nation’s decision as the gate, not as a permitting obstacle. We do not name the Nation in this proceeding, deliberately – a partner is named publicly when the partner confirms in writing that it wishes to be named, and not one day before. The panel may weigh that restraint as evidence in itself of how this project treats consent.

What would prove us wrong: any eventual ownership structure where the Nation’s position is contingent, capped, or subordinate; any evidence the project advanced past a gate the Nation had not opened; any gap between the structure described here and the documents eventually filed.

4. The Ecological Ledger

The claim: the Hub is net-positive on the ecological ledger – and we make the claim in the way this Commons requires, as a falsifiable statement of evidence, argued before a chair whose whole purpose is to break it. It is argued against a stated siting envelope, not a named place: a moderate-flow tidal channel on the BC coast, a tidal resource band of 1.5 to 2.5 m/s, a deployment depth of 30 to 40 m, design-rated to 60 m. Where the evidence turns on a parameter outside that envelope, the correct finding is Insufficient Evidence, and we will publish it in those words.

The design choices carry the argument:

Small. The pod cluster is sealed vessels of roughly hot-tub scale on a single discrete seabed mount – closer to an undersea cable installation than to a marine construction project. No human entry, no ongoing surface presence, invisible from shore.
Sited for gentleness, not for maximum power. The chosen inlet water is a gentler tidal flow than the region’s famous high-energy narrows, deliberately: lower acoustic disruption, fish-safe turbine speeds, lower-stress installation. We gave up peak power for this. That trade is the ecological ledger’s first exhibit.
Short cable to shore. Hundreds of metres of armoured fibre, not hundreds of kilometres of vulnerable subsea cable crossing open water.
Precedent evidence that submersion is benign at this scale: Microsoft’s Project Natick Phase 2 vessel ran 864 servers on the seabed off Orkney for two years and came up hosting barnacles, algae, and anemones – a functioning reef skin on a working data centre. We cite Natick heavily in the reliability ledger below; here it stands for a simpler point: two years of continuous seabed operation produced no recorded ecological incident.

And the honest boundary of the claim: Natick was one vessel at one site. Our inlet is its own living system. Acoustic effects on fish and marine mammals, thermal effects at the exchanger, brine return, the seabed footprint of the mount – every one of these belongs to the More-Than-Human Chair, which exists in this Hearing precisely so the water’s interest is argued from evidence and not from the proponent’s adjectives. That chair is empty as this round publishes. We will supply our modelling to whoever takes it, and we expect them to hunt for the failure mode we missed. Until someone does, 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. We would rather publish that sentence than let the silence read as agreement.

What would prove us wrong: acoustic output at frequencies and levels shown to displace resident species; monitoring that finds thermal or salinity alteration persisting beyond the near field; any datum showing the mount or turbine harming fish passage. The docket’s own words bind us: if it cannot meet the standard, it should not be built.

5. The Trust Ledger

The claim: the Hub can be trusted because it is architected to be checked, forever – and because its proponent submitted to this proceeding before asking anyone for anything.

What the operator commits to publish, permanently, as a condition of its own standard:

The live six-column water ledger – withdrawn, recovered, returned, evaporated, altered, dispersed – updated on a published schedule, with the monitoring methodology open.
Both power meters – generation and consumption – continuously, so the surplus claim is checkable by anyone with a browser.
The dispersion monitoring data against the published model, including every exceedance.
The payment structure to the Nation at the level of structure and priority (the Nation’s own financial details are the Nation’s to publish or not), so “off the top” remains a checkable fact.
A public corrections log, in the Commons tradition: every error found, by anyone, recorded and dated, forever.
This Hearing’s full record – including the Cross round against the project and the finding, whatever it says – linked from the project’s front page for as long as the project exists.

The corrections log is not a promise made for this round. It was exercised before this round published. Preparing this Direct, we found that the Hub’s own published blueprint described its primary turbine class as purpose-built for 2.25 to 3.5 m/s, while this docket stipulates an envelope of 1.5 to 2.5 m/s – and described the same device classes, elsewhere on the same page, as purpose-built for the lower band. Both statements were wrong and they contradicted each other: those two figures are published rated-output points for the device, not a design band. The blueprint has been corrected and the correction published and dated. We report it here rather than leave it for the Cross to find.

One more exhibit belongs in this ledger, and it is the proceeding itself. Every data centre operator says “trust us.” The evidence before the panel is that this one said: here is the instrument, we go first, publish the verdict either way. The panel should weigh not the courage of that but the incentive structure it creates – a proponent who has pre-committed, in public, at a permanent URL, to the conditions under which his own project should be refused.

What would prove us wrong: any published commitment above quietly narrowed after certification; any monitoring gap during operation; any correction suppressed. The trust ledger is the only ledger that can be failed retroactively, and we ask the panel to say so in its finding.

The reliability foundation (evidence common to all five ledgers)

Every ledger above leans on one empirical foundation, so we cite it once, plainly. Microsoft’s Project Natick Phase 2 – the only published large-scale industrial experiment in submerged data centre operation – ran 864 servers in a sealed nitrogen-filled vessel 36 metres down off Orkney, Scotland, from 2018 to 2020, powered by the islands’ renewable grid. Result, published by Microsoft Research in 2020: a 0.7 percent server failure rate submerged against 5.9 percent in the matched land control. Eight times more reliable underwater – no HVAC cycling, no oxygen, no humidity swings, no human hands.

Microsoft ended the program in 2024, having proved the physics and declined to commercialize it. The Hub’s design stands on that published result and scales it down, not up: smaller vessels, a gentler site, a community output loop Natick never attempted. The panel should treat Natick’s numbers as what they are – another operator’s published evidence, the strongest available precedent, and not a guarantee. Our own numbers publish with this round as claims for verification. The cluster’s design target is approximately 40,000 compute nodes. That is a design target – a claim and not a fact. It has not been validated by any pilot; it depends on final vessel count, thermal headroom and turbine sizing that are all engineering-in-progress; and the Cross is entitled to treat it as the softest number in this document. We publish it because a proponent who asks for verification and withholds the figure to be verified is asking for something else. The pilot exists to make it a fact or kill it.

What we ask of the panel

Test every claim above at full strength. Demand the documents. Weigh the More-Than-Human Chair’s case as hard as ours. And rule on the claim as written: not whether the Hub is perfect, but whether a project built and bound this way – water-positive by design, net-generating by design, consent-gated by structure, monitored in public, forever – is net-positive on all five ledgers and can be independently verified as such before it is built. Whether it deserves the trust it asks for is not before you. That is the host Nation’s decision and the community’s decision, and no panel substitutes for either.

If it does, say so, and the certification will mean something because this proceeding could have killed it.

If it does not, say that instead. The coast deserves the truth more than we deserve the project.

Now that you know, what will you do?

CORRECTION AND NARROWING OF CLAIM

Hearing #6 – The Hub on Trial
Filed: August 26, 2026, by ISWP as proponent
Applies to: Round 1: Direct, published August 25, 2026
Status: The claim as filed on the docket stands unamended. This notice narrows what the proponent asks the panel to certify, and corrects four figures in the Direct. It is filed five days before the Cross so that Round 2 argues against a corrected document.

In plain words, before the detail

We published our own case for the Hub on August 25. Then we audited it, the way we said we would audit it, and we found six things wrong with it. This notice publishes all six at once, five days before the opposition speaks, because finding them ourselves and hiding them would have failed the one ledger we cannot fail retroactively.

The short version: the Direct argues the Hub at a size it never states. It publishes a node count and a water figure, and omits the number that governs both – how much power the thing draws. Work the arithmetic the Direct invites, and its power case and its ecological case turn out to be arguing against each other.

Nothing here is a physics failure. The Hub still works. What changes is the size of the thing we are asking a panel to certify. 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. Narrowing gives the Cross an easier target, not a harder one. That is the correct direction for a correction to move.

Why a correction and not a third amendment

The claim has been amended twice in eight days. It could be amended a third time, and the error would disappear from the record.

We are not doing that. The claim as filed stays on the docket in its own words, and this notice sits beside it. A proceeding whose claim quietly improves every time the proponent finds a problem is not a proceeding. The trust ledger promised a public corrections log; this is the first time that log has been exercised against our own case, and the entry is meant to be readable by the people who will one day have to decide whether to believe us.

What changes, at a glance

	Round 1 as published	As corrected
Scale of the claim	not stated anywhere	1 pod, 0.26 MW IT load
Generation	“an adjacent tidal turbine”	an array of order 10 to 12 low-flow units
Fresh water returned	approx. 50,000 L/day	approx. 10,000 L/day at pilot scale
Compute	“approximately 40,000 compute nodes”	withdrawn. 250 kW IT per pod published instead
Enclosure	“hot-tub-scale” inside the claim	sealed, uncrewed, recoverable, no surface presence
Turbine fit	rated points, band unstated	documented at 1 to 2 m/s. Better fit than we claimed
Fish evidence	not cited	ORPC’s Igiugig record cited, with its own caveat
Heat rejection	never published	computed and published below

Four ledgers are touched. The community ledger and the trust ledger are unchanged.

The six corrections

1. The array was never counted, and the array is the whole ecological argument

What Round 1 says. “The tidal turbine adjacent to the pod cluster generates more power than the cluster consumes.” Singular. And, in the ecological ledger, that the installation is “closer to an undersea cable installation than to a marine construction project.”

What the equipment does. 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.

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
5 pods – 1.31 MW – roadmap	1,310 kW	~47	~56
20 pods – 5.25 MW – roadmap ceiling	5,250 kW	~188	~225

Round 1 describes the five-pod case without saying so. At that size, “an adjacent tidal turbine” is approximately fifty seabed machines, and fifty seabed machines is not an undersea cable installation. Our power ledger and our ecological ledger were arguing against each other, and the reason the contradiction survived two weeks on a live page is item 3 below.

Correction. The claim under test is one pod and an array of order 10 to 12 units. At that size the sentence in the ecological ledger is defensible and we will defend it. At the roadmap sizes it is not, and we say so here rather than let it be found.

A further disclosure the Cross is entitled to. 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, and it is one of the things the in-situ resource survey has to answer before anyone builds a second pod.

2. The 40,000-node figure is withdrawn

What Round 1 says. “The cluster’s design target is approximately 40,000 compute nodes.”

Why it is wrong. Our own blueprint retired all node-count claims on August 13. The Direct reinstated one on August 25. Both pages were live at the same time, saying opposite things.

It is also arithmetically unavailable. Five pods is 1.25 MW of IT load. Forty thousand nodes into 1.25 MW requires 31 watts per node. The smallest device our blueprint names is a 72 W edge card, which puts the ceiling near 17,400. The published figure is out by a factor of about 2.3. Its origin is traceable: the May 12 pod design assumed 20 to 25 W system-on-chip nodes, blueprint v2 changed the silicon and retired the counts, and the Direct carried the old quotient into the new denominator.

Correction. The node count is withdrawn and not replaced with another node count. Node count is a quotient of a silicon choice our own blueprint lists as a feasibility-study deliverable and a one-shot decision we have not made. 250 kW of IT load per pod is the honest primitive. It is the divisor, it is already published, and it is what should have been in the Direct.

3. Round 1 never states the Hub’s size, and that is the root cause

No megawatts. No pod count. No array count. A reader could not tell whether “a handful per seabed mount” meant 260 kW or 5.25 MW, and every figure in every ledger floated free of the number that governs it. Corrections 1 and 2 both exist because there was no denominator in the document.

Correction. Scale is now stated in the claim, first line, and every other number is derived from it and checkable against it. This is the single most useful edit in this notice, and the lesson has been written into how this Commons drafts: publish the denominator.

4. “Hot-tub-scale” leaves the claim, and here is the number that removes it

The claim under test contained the words “hot-tub-scale.” Our own blueprint, on a different page, says the opposite: “‘Hot-tub-sized’ remains a communications image … a 250 kW IT load may require a larger or differently shaped enclosure.” A size adjective that our own engineering document contradicts has no business inside a falsifiable proposition.

We never published the number underneath the argument. Here it is.

A pod rejects its full electrical draw, about 263 kW, into 8 to 12 C seawater. Rejection is an area problem:

	Natick Phase 2, published	Hub pod at hot-tub geometry	Hub pod at Natick-class flux
Heat rejected	240 kW	~263 kW	~263 kW
Hull dimensions	12.2 m x 2.8 m	2.2 m x 0.9 m	approx. 13.5 m x 2.8 m
External area	~120 m2	~14 m2	~130 m2
Heat flux	~2.0 kW/m2	~19 kW/m2	~2.0 kW/m2

The only submerged data centre ever operated at this power class carried its heat out through roughly ten times the surface area. To match Natick’s flux, a hot tub has to become a vessel of about Natick’s size.

Is 19 kW/m2 physically impossible? No, and we will not overstate it. At a 30 K difference between the exchanger and the sea, 19 kW/m2 needs an overall coefficient near 630 W/m2K. At the site’s 1.5 to 2.5 m/s the external water film on a 2.2 m cylinder gives roughly 2,500 W/m2K (Churchill-Bernstein, seawater at 10 C) – about four times what is needed. The problem is slack water. Four times a day the current stops and forced convection stops with it, and free convection on the same cylinder gives roughly 680 W/m2K (Churchill-Chu) – the requirement, with no margin at all, before the shell wall and the internal side are counted.

Correction. The size adjective is struck from the claim and replaced with the properties that are ecologically and operationally load-bearing: sealed, uncrewed, recoverable, no human entry, no surface presence. Enclosure geometry and heat-exchanger area are named as a feasibility-study deliverable, and the slack-water case is named as the binding one. We do not claim a hot tub rejects 250 kW. We claim a sealed uncrewed vessel does, and the study determines its shape.

These are first-order estimates by the proponent, with the geometry, temperatures and correlations stated so that anyone can check them or break them. That is the point of publishing them.

5. A correction in our own favour, and the evidence we had and never used

Our blueprint described the RivGen class as purpose-built for 2.25 to 3.5 m/s. It is documented operating at Igiugig in a mean flow of 1 to 2 m/s. The device is a better fit for our stated 1.5 to 2.5 m/s siting envelope than our own document claimed. We correct errors that help us on the same day and in the same block as errors that hurt us, or the log is worthless.

And the evidence Round 1 never cited, with the More-Than-Human Chair empty:

2015 monitoring at Igiugig detected no obvious physical injuries to fish and no altered behaviour by wildlife near the device (Tethys/PNNL).
The adult salmon monitoring requirement was removed in 2022, in consultation with the project’s Adaptive Management Team, after several years showing no documented collisions and no evidence of device impact to adult sockeye salmon.

And the caveat, which we publish ourselves rather than wait for it. 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. An exhibit offered with its own limitation removed is not evidence, it is advertising, and the empty chair is exactly the wrong place to start advertising.

Correction. The fish record enters the ecological ledger, in full, with its caveat attached and its juvenile question named as open.

6. What the correction costs, stated plainly

The water headline drops. Round 1 claimed approximately 50,000 litres a day of fresh water returned to a drought-stressed community system. That was always a five-pod figure and it was published beside no pod count. At the pilot the honest figure is on the order of 10,000 litres a day.

We are not softening it. It is a fifth of the number we published yesterday, it is the number that matches the thing we are asking to fund, and the direction of the claim – the community’s water account goes up because the Hub exists – is unchanged and is what the panel should test.

“Net exporter” is kept. At pilot scale, with an array of 10 to 12 units, it is true.

The claim as narrowed

The design under test. One Hub pod: a sealed, uncrewed, recoverable compute vessel with no human entry and no surface presence, on a single seabed mount in a moderate-flow tidal channel on the BC coast, within a stated envelope of 1.5 to 2.5 m/s and 30 to 40 m depth; an IT load of approximately 250 kW, approximately 263 kW total draw; powered by a tidal array of order 10 to 12 low-flow units sized above pod load; cooled by the surrounding sea with no freshwater withdrawal; using server waste heat to drive membrane distillation returning fresh water on the order of 10,000 litres a day to the community system; co-owned with the host First Nation, paid from first position in the revenue waterfall; held in a BC Benefit Company with a charter public-benefit purpose; proceeding only on the Nation’s decision and the Nation’s timeline.

The claim before the panel. The Hub pilot, as specified above, is net-positive on all five ledgers this docket published – water, power, community, ecology and trust – and can be independently verified as such before it is built.

If it cannot, it should not be built. We commit to publishing that finding too.

The community ledger and the trust ledger carry over from Round 1 unchanged. Every commitment in the trust ledger applies to this narrowed claim exactly as filed.

The roadmap – disclosed, not claimed

The panel is not asked to rule on any of this. It is published so that no one can say we argued a pilot while quietly planning a facility.

Step	Pods	IT load	Array, order of	Water, order of	Status
Pilot	1	0.26 MW	10 to 12 units	10,000 L/day	the claim under test
Commercial step	5	1.31 MW	47 to 56 units	50,000 L/day	disclosed only
Design ceiling	20	5.25 MW	188 to 225 units	roadmap	disclosed only

For scale outside our own documents: one pod is roughly two frontier AI racks, or about two and a half times a standard micro edge data centre. Twenty pods is metro-edge scale, a commercially normal facility. At no step is this a frontier training site, and the blueprint has always said so – a sealed pod frozen on its silicon for five years is the wrong machine for training. The Hub’s market is inference, edge and sovereign workloads.

And the constraint, stated against our own interest. Pods are modular and comparatively cheap to add. The array is not. Roughly ten units per pod means compute ambition and the ecological ledger pull against each other permanently, and at the upper steps the array is very likely more than the ecological ledger can carry as currently argued. Whether those steps are reachable at all is decided by the in-situ resource survey, which is item one on the blueprint’s own funded-study list and is named there as the make-or-break number. That is not a hedge. That is the whole reason a funded study is the thing this Commons is asking for.

What Round 2 should now attack

The Cross was always going to find items 1 and 2. It is welcome to argue that finding them ourselves changes nothing, and it may be right. The harder targets, and the honest ones, are these:

The slack-water heat-rejection case, and whether exchanger area closes it.
The juvenile fish question at Igiugig, and whether an adult-salmon record transfers to a BC inlet at all.
Array wake interaction and cumulative effect, which no single-unit record can settle.
Whether a pilot at 0.26 MW can support any inference of net-positivity at the sizes on the roadmap.
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 corrections log entry

2026-08-26. Six items, one audit, self-found. Scale absent from the claim; array count absent; node count contradicting the blueprint and out by 2.3x; a size adjective inside a falsifiable proposition and contradicted by our own engineering page; a turbine-band statement wrong in our own favour; heat-rejection flux never published. Claim narrowed to the pilot, roadmap disclosed, four figures corrected, one exhibit added with its caveat. Filed five days before the Cross. 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.