This is the deliverable a screening engagement produces, shown in full. Five illustrative candidate coordinates went into the candidate-site screen (POST /v2/insights/siting/sites, methodology v1.29) on 9 August 2026; what came back is an exception ledger: each site gets one of four verdicts (proceed, investigate, reject or insufficient evidence), with the material findings, the evidence behind them, the coverage gaps stated per check, and a frozen, retrievable answer receipt. The coordinates are illustrative; the evidence, the verdicts and the receipt are real and pinned to the run below. There is no score: a site with no visible connection path fails a gate, and a site whose evidence is too thin is abstained on, never ranked low.
Until 9 August 2026 this page presented a run computed under methodology v1.7 (receipt 26aa555c). That screen fired decisive findings from evidence our own later rules do not permit to decide a target of this size. Since v1.27 every piece of evidence carries a classification: which direction it points (import demand or generation export), what voltage grain it lives at, and which connection tiers it may therefore decide rather than merely inform. A 300 MW load resolves to the transmission tier, and two of the v1.7 findings were decided at grains no such load could connect at:
Correction, 10 August 2026: this paragraph said the identical five coordinates and parameters were rerun under v1.29. That was not true of Candidate B. The v1.7 screen sized Leeds at 150 MW; the August rerun sized it at 300 MW, while the other four held at 300. An external appraisal caught it, and it mattered: Leeds’ connection-grade evidence is 202.6 MW available at Kirkstall B (132 kV GSP), which is short of 300 and clear of 150, so the changed target could have carried the verdict. The receipt could not have revealed it either, because until this date a receipt froze only the request-level target and never the per-site ones, so both screens recorded an identical 300 MW. That is fixed: the controlled rerun freezes site_load_mw per candidate, and the scenario is now declared once in ops/flagship_rerun.py, which refuses to report a comparison whose frozen inputs do not match the declaration. The controlled result: Leeds holds at investigate at its original 150 MW, and the only verdict that moves between v1.7 and v1.29 is Dublin. The conclusion below survived the correction; the description of the inputs did not.
The five coordinates were rerun under v1.29 at their original targets. Dublin clears to proceed. Teesside, Leeds, the Permian Basin and Tromsø hold their verdicts, now resting on evidence at the applicable grid-supply-point and 132 kV+ grain: Teesside’s constraint is restated at Saltholme (132 kV GSP) reading the same 60.4 MW, and Leeds’ export-direction evidence is replaced by 7,376 MW of contracted demand-direction connections, which genuinely does compete. Four checks that did not exist in v1.7 now report. The newer rules changed the reasons more than they changed the answers, which is the result we would want and not one we could have assumed.
The v1.7 receipt has not been deleted. It is retained, marked retracted with its reason, and linked to its replacement in both directions; its export carries the withdrawal above the title so a forwarded copy cannot read as current. Retrieve the withdrawn receipt and compare it against the current one below. Quarantine, never delete: the record of a correction is part of the evidence.
The underlying defect was not confined to this page. Receipts pinned the answer and a methodology version label but never the methodology text, so an export rendered later showed the current rulebook above findings computed under older rules. From v1.29 every receipt freezes its complete rulebook (methodology, confidence, limitations, the screening thresholds and the evidence-applicability registry rows) with a SHA-256 over it, so the rules verify exactly as the answer does. Receipts archived before that change say so on their face rather than borrowing rules they were never computed under.
| Candidate | Verdict | Material finding / basis | Next action |
|---|---|---|---|
| D · Tromsø, NO 300 MW | reject | Nearest ≥275 kV substation (Balsfjord, 420 kV) is 52.1 km away; beyond the 40 km connection gate. No visible EHV connection path at screening grain. | Stop, or supply connection evidence Kortex cannot see. |
| A · Teesside, GB 300 MW | investigate | 400 kV substation 3.1 km away and 18 EHV points within 25 km; but the operator’s own published demand availability at the nearest applicable point (Saltholme, 132 kV GSP, 3.2 km) is 60.4 MW against a 300 MW target, and every published candidate point at that grain is below target; 1,935 MW of demand-direction transmission connection agreements sit at stitched connection sites within 50 km (3 projects, same-direction competition), beside 7,375 MW accepted-to-connect in the generation direction (225 projects, ECR vintage 2026-02-17) and 4,615 MW in the local planning pipeline (116 REPD projects). | Operator feasibility inquiry, or evidence of a different connection point. |
| C · Permian Basin, US-TX 300 MW | investigate | 345 kV substation 6.8 km away, 59 queue projects within 50 km (30% completion); but basin baseline water stress is extremely high: a wet-cooled load needs a cooling-water story before the site can carry it. | Confirm cooling technology and water source; dry/hybrid cooling changes this materially. |
| B · Leeds fringe, GB 300 MW | investigate | 275 kV substation 6.1 km away; published demand availability at the nearest applicable point (Kirkstall B, 132 kV GSP, 6.2 km) is 202.6 MW against a 300 MW target, and it is the only published candidate at that grain; 7,376 MW of demand-direction transmission connection agreements within 50 km (6 projects) contest the same import capability, beside 11,165 MW accepted-to-connect in the generation direction (396 projects, ECR vintage 2026-02-17). | Operator feasibility inquiry, or evidence of a different connection point. |
| E · Dublin fringe, IE 300 MW | proceed | 400 kV converter station 13.6 km away and low-medium water stress; no visible screening-level disqualifier. Under v1.7 this site was investigated on ESB Networks’ published availability at a 38 kV station group (Finglas, 5.6 km) reading 0.0 MVA against a 300 MW target. A 38 kV group could never carry a transmission-tier load, so under v1.29 that figure is served as context and cannot decide: the finding does not fire and the site clears. The Dublin connection constraint remains real and is stated in the evidence; what changed is that it is no longer decided at a grain that could not bear it. | No screening-level disqualifier: merits a connection feasibility and engineering screen. |
The fourth verdict, insufficient_evidence, is returned when fewer than three checks produce evidence at a site: the decision cannot be defended, so the screen abstains instead of scoring. None of these five triggered it; the thinnest non-rejected dossiers are Candidates C and E at 5 of 8.
The same shortlist, re-screened twice. These exact five candidates were screened three times on 12 July 2026 as the graph grew, and each run is frozen by its own receipt: 56da6c4a (morning), 19e2acc7 (after the GB connection registers landed), and the run on this page. Candidate B flipped proceed → investigate when the GB Embedded Capacity Register exposed an 11.2 GW accepted pipeline nearby; Candidate E flipped when ESB Networks’ published capacity exposed 0.0 MVA demand availability around Dublin. Both facts existed all along; they were public, just not connected. Two more enrichments the same night added the Irish connection registers (EirGrid TSO contracted + ESB DSO lists) and the GB planning pipeline (REPD + the NSIP register); and a final enrichment added the EA regulatory water layers (CAMS abstraction classes + Flood Zones 2/3); the screen ran eight checks that night, and the GB candidates carried 7 of 8 with evidence; the screen has since grown further (the path-to-power ledger, land fabric). Comparing any two receipts shows exactly what changed and why. That is what re-screening a saved shortlist against a growing graph looks like.
This run was archived at computation time. The receipt returns the canonical answer exactly as computed (not a recomputation) with the request parameters, the per-source register records and each source’s dataset_version fingerprint as they stood. graph_snapshot_id pins the schema/source register; data_snapshot_id pins the observed data state of every contributing source.
receipt_id 4ea95cff-c6e5-49e3-be2f-919cf5eb71c7 computed_at 2026-08-09T13:39:00Z methodology 1.29 methodology_sha 389720d30d85b26e7be2e2ebd2a0f9817a8745d30882f6d0dcaacfdc0b7d2ea3 supersedes 26aa555c-754c-4810-b69d-aee868119569 curl https://kortex.orkora.com/v2/receipts/4ea95cff-c6e5-49e3-be2f-919cf5eb71c7
Receipt retrieval is public: anyone you forward this to can pull the frozen answer without a key and verify the response hash. Two hashes ride on every receipt from v1.29: response_sha256 pins what was concluded, and methodology_sha256 pins the rules it was concluded under, so an auditor holding only the artifact can verify both without trusting that neither has moved. Re-running the screen returns the current graph’s answer, which may differ; that difference is the point of the receipt.
This receipt records that it supersedes the withdrawn v1.7 run, and that run records that it is superseded_by this one. The chain reads in both directions on purpose: a receipt that could only point backwards would leave anyone holding the older artifact unable to discover that it had been overtaken.
Take the deliverable itself: the same frozen answer as the client-ready artifact; download XLSX (exception ledger · findings · evidence · provenance sheets) or download PDF (the screening memo). Both are rendered from the receipt, never recomputed, and carry the snapshot ids, the response hash and the pinned rulebook for verification. For the contrast, pull the withdrawn v1.7 memo: it opens with its own retraction.
The second screen template shares the first one’s discipline. Six illustrative counterparties; identified by LEI, Companies House number, or bare legal name; went into the diligence screen (POST /v2/insights/diligence/entities, methodology v1.0) on 14 July 2026. Same four verdicts, same exception-ledger shape, same receipt. There is no risk score: an exact primary sanctions-list match fails a gate; an entity the registers cannot identify is abstained on, never guessed at.
| Counterparty | Verdict | Material finding / basis | Next action |
|---|---|---|---|
| E · submitted as a bare name “BANCO NACIONAL DE CUBA” | reject | Exact primary sanctions-list match: BANCO NACIONAL DE CUBA (OFAC SDN, programs: CUBA). The gate fires on the submitted name even though the registries cannot resolve the identity; a designation match must not hide behind a resolution failure. | Refer to sanctions counsel before ANY further step. |
| F · submitted as a bare name “Meridian Consolidated Energy Partners” | insufficient_evidence | No register (GLEIF, 3.36M entities, or Companies House, 5.73M) carries this name, exactly or punctuation-normalised. The screen abstains and lists the closest register candidates; it never guesses an identity into a verdict. | Supply an LEI or Companies House number. |
| B · BlackRock, Inc. LEI 529900VBK4… | investigate | One high-confidence ICIJ offshore-leaks candidate: “Blackrock Fund Inc.” (St Kitts & Nevis, match score 0.737); a namesake to adjudicate, not a finding of wrongdoing. Ownership-network position on-row: 100th-percentile centrality, 115 reported direct children, a control family of 73,617 entities. | Review the ICIJ records and confirm identity before drawing conclusions. |
| D · Eagle New Energy Ltd CH 06367329 | investigate | Companies House status: Liquidation; and 6 outstanding registered charges (2 satisfied): assets are pledged as security, scope unknown at count grain. Dossier in full below. | Obtain the insolvency practitioner’s filings before any reliance. |
| A · Iberdrola SA submitted “IBERDROLA, S.A.” | proceed | Resolved by punctuation-normalised name to LEI 5QK37QC7…; ACTIVE in GLEIF; sanctions searched (zero exact matches, vintage on-row); ICIJ searched (zero); observed footprint in the Kortex graph: 173 stitched generators, 47.0 GW across six fuels, modelled attributed revenue $2.4B. | No visible screening-level disqualifier; merits confirmatory KYC. |
| C · Savills IM LLP CH OC306423 | proceed | CH number joins GLEIF through the registration-authority map (LEI 529900ZIX1…); Active + ACTIVE on both registers; PSC names a corporate controller (Savills IM Holdings Limited); zero outstanding charges; 8 of 9 checks with evidence. | No visible screening-level disqualifier; merits confirmatory KYC. |
Verdict order is the contract: reject = a gate failed (sanctions, or a legally dead entity) · insufficient_evidence = identity unresolved or the dossier is too thin to defend · investigate = material findings need resolving · proceed = no visible screening-level disqualifier, never a clean bill.
receipt_id c93dc919-2a2b-4328-8213-0ed809875f5c computed_at 2026-07-14T02:13:40Z graph_snapshot c77503da11cd844e data_snapshot 707ed0728e2de441 curl https://kortex.orkora.com/v2/receipts/c93dc919-2a2b-4328-8213-0ed809875f5c
Take the deliverable itself: download XLSX (exception ledger · findings · evidence · provenance sheets) or download PDF (the screening memo). Rendered from the frozen receipt, never recomputed.