Prediction #001 — Linka Tungsten Project

SEALED — INDEPENDENT OUTCOME PENDING

Case ID

GS-PRED-0001-LINKA

Asset

Linka Tungsten Project, Nevada, USA (Linka / Conquest / Hillside workings, Lander County; Linka Mine workings approx. 39.3286 N, -116.8366 W)

Operator

Viking Mines Limited (ASX: VKA; OTCID: VKALF)

Sealed

2026-09-10T21:06:51+02:00

Cutoff

2026-09-10T23:59:59Z

Priority

LOW — as an overlooked-element target

Confidence

HIGH

Case classification

1. What is being predicted

The ELEMENT CONTENT AND REPORTED VALUES of the next batch of ALS Reno assays from the Linka maiden RC programme (holes other than LKRC0002), as those assays are disclosed to the ASX in mid-to-late September 2026. Assay numbers in drill samples. Nothing else.

Explicitly not predicted:

• Not ore, not a resource, not economics, not a JORC estimate, not metallurgical recovery, not share price.
• Not whether the 900 t bulk sample proceeds, not permitting, not the remaining 47 holes of the originally scoped 63-hole programme.
• Not the exact hole-by-hole depths (collar table for the 16 holes is a raster image in ASX 2026-09-10 Appendix 1 and is not in the corpus in machine-readable form).
• Not the announcement date. 'Mid to late September' is the operator's own wording (ASX 2026-09-10); if it slips, that is not a failure of this prediction.

2. The call

THE PENDING BATCH WILL BE REPORTED AS TUNGSTEN ONLY. I predict that the mid-to-late-September batch announcement will contain no grade, in ppm, %, or g/t, for any element other than tungsten (reported as W% and converted to WO3% by the operator's stated W ppm / 10,000 x 1.2611). Probability I assign: 0.80. This is the single most confident statement in this file and it is the one that should be scored first.

Probability assigned to that scenario: 0.8 — 0.80 that the pending batch is reported as tungsten only. Stated in advance so that a near-miss cannot later be described as a hit.

3. Quantitative predictions

• P1. At least TWO holes in the next reported batch return an intercept of >= 5 m at >= 0.30% WO3.
• P2. LKRC0003 returns an aggregate reported intercept of >= 8 m at >= 0.30% WO3, beginning at less than 20 m downhole.
• P3. LKRC0004 returns an aggregate reported intercept of >= 6 m at >= 0.30% WO3, beginning at less than 15 m downhole.
• P4. The best single reported intercept in the batch (>= 5 m length) has a weighted average grade BETWEEN 0.30% and 1.20% WO3. I am predicting a band, not a floor: I am explicitly predicting that the batch does NOT deliver a >= 5 m intercept above 1.20% WO3.
• P5. LKRC0001 returns no significant intercept, or is not reported as an assayed hole at all.
• P6. Individual 1 m maxima somewhere in the batch fall between 1.0% and 4.0% WO3.

Note the fourth item: it predicts a band with a ceiling, not a floor. A result far stronger than expected falsifies this prediction just as a weak one does. That is deliberate — a prediction that only fails downwards is not much of a prediction.

4. Element-by-element call

Gradeability rule. An element the laboratory does not report is scored UNGRADEABLE, never as a correct prediction. In an earlier case a widely repeated germanium figure turned out to have no data column behind it at all. Scoring against a number that does not exist would have manufactured a hit.

5. Why LOW priority, with HIGH confidence

LOW as a SECOND-CHANCE / OVERLOOKED-ELEMENT target, which is what this benchmark screens for. The tungsten itself is not overlooked - it is the subject of 47 ASX announcements in nine months and an active drill rig, so it fails the 'overlooked' test by definition. And no second element in this corpus reaches a credit-bearing threshold: across 54 site-specific documents the only non-tungsten numbers ever attached to Linka are 3.0 g/t Ag and 0.15% Bi from one ramp channel sample (ASX 2026-02-04), and the one historic multi-element episode explicitly found nothing material except molybdenum (ASX 2026-02-17). A screen looking for a missed critical metal should spend its next hour somewhere else.

6. Why the confidence is HIGH

This earns HIGH rather than MEDIUM on four independent legs. (1) Repetition under a Competent Person's signature: 'Scheelite, with traces of chalcopyrite and molybdenite are the only ore minerals recognized' appears in 24 separate Viking ASX announcements from 2025-12-16 to 2026-09-10 - it is the operator's own JORC Table 1 geology statement for Linka Mine, not a press line. (2) Revealed preference under a full multi-element dataset: ME-ME61, which the operator itself says 'reports 61 elements', was run at least twice (due-diligence surface sampling, ASX 2026-02-04; stockpile sampling, ASX 2026-06-09). In February the operator mentioned Ag/Mo/Bi qualitatively; in June, holding a second 61-element dataset, it published tungsten only. (3) Direct method precedent nine days before cutoff: the regional programme reported 2026-09-01 used the SAME pending-batch method (ME-MS89L / W-MS89L) and reported no element other than WO3. (4) A contemporaneous negative from the only multi-element work ever done on this ground: Union Carbide 1975, 28 surface samples for WO3 'and sometime other elements including Molybdenum, Copper, Zinc, Gold and Silver', with the operator stating 'No material results have been identified for the remaining other elements and are not reported' (ASX 2026-02-17). Four legs, three of them modern, all pointing the same way.

7. What would prove this wrong

Twelve falsification criteria, fixed before the result existed:

• F1 (kills the headline call): the mid-to-late-September batch announcement reports a numeric grade for ANY element other than tungsten in its body, bullets, or assay table.
• F2 (kills the Mo co-product call, P7): any reported interval of >= 2 m averaging >= 500 ppm Mo, or molybdenum presented anywhere in the release as a co-product, credit, or second critical mineral.
• F3 (kills the Cu call): any reported interval of >= 2 m averaging >= 0.10% Cu.
• F4 (kills the Zn call): any reported interval of >= 2 m averaging >= 1000 ppm Zn.
• F5 (kills the Ag call): any reported interval of >= 2 m averaging >= 30 g/t Ag.
• F6 (kills the Au call): any reported interval of >= 2 m averaging >= 0.2 g/t Au.
• F7 (kills the Bi co-product call, P11): any reported interval of >= 2 m averaging >= 0.15% Bi in drill samples.
• F8 (kills the B4 rejection, P13): any of Sn, Te, Ga, In, Ge or Co reported at a value the operator describes as elevated, anomalous, or of significance.
• F9 (kills the tungsten width call, P1-P3): NEITHER LKRC0003 NOR LKRC0004 returns an intercept of >= 5 m at >= 0.30% WO3.
• F10 (kills the grade-band call, P4): the best >= 5 m intercept in the batch grades above 1.20% WO3, or no hole in the batch reaches 0.30% WO3 over >= 5 m.
• F11 (kills the Mo-first ordering, P9): a non-tungsten element other than molybdenum is the first to be given a number at Linka.
• F12 (retrospectively invalidates the gradeability reasoning): the released ME-MS89L suite turns out to be a single-element tungsten determination with no other element measured at all - in which case F2 through F8 can never be tested and only F1, F9, F10 remain scoreable. I am recording this now so that an unscoreable outcome is not later read as a correct one.

8. Counter-evidence

Arguments against our own call, recorded at the time of sealing:

• Against My tungsten-only headline call. — ASX 2026-09-10 Appendix 1 is titled 'DRILLHOLE INFORMATION & ASSAY RESULTS' and is a raster image that pypdf cannot read. I have not seen its columns. If that table already carries Mo, Bi or Ag columns for LKRC0002, then the pending batch will carry them too and my rank-1 call is wrong from the moment I write it. I cannot exclude this and I am not pretending to.
• Against My tungsten-only headline call. — The pending method is written TWO different ways in the same document. ASX 2026-09-10 says 'ME-MS89L' in the sampling-techniques row and 'W-MS89L' in the quality-of-assay row; ASX 2026-09-01 does the same. 'ME-' is a multi-element prefix and 'W-' is a single-element prefix in the operator's own usage elsewhere (ME-ME61 multi-element vs W-MS85h for tungsten). If the ME- form is the correct one and the W- form is the typographical slip rather than the reverse, a full multi-element dataset exists for every pending hole and only editorial choice keeps it out of the release.
• Against My BARREN calls on Cu, Zn, Ag, Au. — The Union Carbide 1975 negative rests on 28 SURFACE samples by a method the operator states is unknown, analysed at two laboratories, with sample size and preparation unknown (ASX 2026-02-17). That is a weak instrument. A 1975 surface reconnaissance failing to find copper is not the same as modern ICP-MS on fresh sulphide-bearing tactite at 40-54 m depth failing to find copper.
• Against My BARREN call on Bi as a co-product. — 0.15% Bi is a genuinely large rock value - 1500 ppm - and it is not a stray grab sample, it is tied to a channel that returned 5 m @ 0.7% WO3 in the Linka Pit itself. If that tenor persists down-dip into the tactite the RC holes are cutting, bismuth would exceed my P11 bar.
• Against My HIGH confidence generally. — The QAQC in the corpus is thin and internally inconsistent. There is no umpire analysis, no twinned holes, and no independent verification of intersections. Worse, ASX 2026-09-10 states the standard/blank rate TWICE and disagrees with itself: 'Standards (1:40 samples) and blanks (1:40 samples)' in one row, and 'one certified reference material every 20 samples, one blank every 25 samples and selects one duplicate every 10 samples' in another. On top of that, LKRC0002's own splitter was judged 'not suitable for the wet sample which was being produced'. The single calibration point my tungsten prediction leans on is the hole with the acknowledged sampling problem.
• Against My BARREN calls on the six critical metals. — Nobody has ever assayed Linka drill core for Te, Ga, In, Ge, Co or Sn and published it. ME-ME61 may or may not include all six. Absence of a flag by the operator is weaker evidence than a measured non-detection, and I am treating it as stronger than it strictly is.
• Against My LOW priority. — If the sole criterion were 'is there metal here', the answer is plainly yes - the tungsten is real, wide, shallow, and confirmed below the historic stope. LOW is a judgement about this benchmark's overlooked-element question only, and a reader using it for anything else will be misled.

9. Missing evidence

• 1. THE ME-MS89L ELEMENT LIST. This is the whole game. Named as manifest gap #4: alsglobal.com is a JavaScript SPA and every probed path returns the same 80,831-byte shell. Until the suite is known, every non-tungsten call in this file is conditional and most may be unscoreable.
• 2. OCR OF ASX 2026-09-10 APPENDIX 1. Titled 'DRILLHOLE INFORMATION & ASSAY RESULTS', raster only. Its column headers alone would settle the gradeability question for the whole exercise BEFORE the result lands, and would also supply the collar table for the 16 holes.
• 3. OCR OF ASX 2026-02-17 APPENDIX 1. It contains the Union Carbide 1975 molybdenum values - the only molybdenum numbers ever measured on this ground and the only direct calibration available for P7/P8.
• 4. NBMG BULLETIN 105, Stager and Tingley (1988), 'Tungsten deposits in Nevada'. Not obtained - a $25 in-print product, 404 on every free path, absent from archive.org and HathiTrust. This is the source the operator itself cites for the Linka workings and its 3D model.
• 5. The 2021 tungsten skarn mineral resource assessment of the Great Basin - ScienceDirect 403. The most on-point modern regional treatment of exactly this deposit type.

10. Next best action

Re-run the EXISTING pulps. No new drilling, no new field work, no new sample.

PREP-31BY pulps for the mineralised intervals already exist at ALS Reno for every assayed hole - dried, crushed to 70% <2 mm, pulverised to 85% passing 75 microns. Take the ~50-80 pulps that fall inside the reported WO3 intercepts and put them through ME-ME61, the four-acid ICP-AES/ICP-MS package the operator has ALREADY used twice at Linka and which it states in its own JORC Table 1 'reports 61 elements' (ASX 2026-02-04, ASX 2026-06-09). Add a 30 g fire assay for gold on the same pulps if the gold question is to be closed rather than assumed.

11. Comparison against simple baselines

A prediction is only worth something if it beats an obvious rule. Each baseline was computed and sealed alongside the prediction, so none can be invented after the fact.

B1 — expect whatever the mine historically produced: TUNGSTEN ONLY, as WO3, at roughly 0.5 to 1.0% WO3.

B2 — the standard associated-element list for this deposit type: W skarn (MRDS model 42 / USGS 14a): tungsten with molybdenum, copper and bismuth, plus silver, gold, zinc, tin and antimony in some members of the class.

B3 — rank by size: NOT APPLICABLE. Size ranking requires a population to rank within. This exercise scores a single asset, and no ranking against peers is being predicted. Stated explicitly so that its absence is not later mistaken for an oversight. For t

12. What this does not claim

• This registry does not claim GeaSpirit discovered or independently verified Linka. The drilling, sampling and assaying are the operator's work.
• One sealed case is an anecdote, not a validation.
• This predicts assay numbers in drill samples. Not ore, not a resource, not economics.

13. Verification

The complete original prediction is published now, encrypted, as LINKA_PREDICTION_V1.enc. Anyone can hash that file today. When the independent assays are released we publish the decryption key, and anyone can then decrypt it and confirm the plaintext hashes to the value above. That means the file cannot have been written after the result — the ciphertext was already public and its hash fixed.

The page you are reading is derived from a fixed public schema applied identically to every case, so no case can quietly omit the fields that went badly. The evidence corpus stays private; its hash is published so it cannot be swapped for a different one later.

14. Independent timestamp

Everything above still relies on us: our page, our release, our server. So on the day it was sealed the hash of the encrypted prediction was also submitted to an independent third-party timestamping service, which records that a given hash existed before a given moment. We do not operate that service and cannot move the date it recorded.

What that fixes, precisely. The prediction was sealed at 19:06:51 UTC on 10 September 2026 and the independent timestamp was recorded the same evening. So the encrypted prediction demonstrably existed, in exactly this form, before that moment — and no party, ourselves included, can move that date afterwards. The pending assays it predicts were still at the laboratory.

What you can check today without trusting us at all. The encrypted original is published above and its SHA-256 is fixed. When the key is released, decrypting it must reproduce the prediction hash exactly. A file published before the outcome existed cannot have been written after it.

Being improved. The public verification step is being replaced by a timestamp token from an independent Time Stamping Authority under RFC 3161, so that the receipt can be downloaded and verified here rather than requested. Until that is in place this page claims no more than the line above.

That gives four layers, each answering a different objection: the site makes it readable, the SHA-256 makes it tamper-evident, the pre-committed ciphertext makes it impossible to have been written afterwards, and the independent timestamp takes us out of the chain of trust for the date.

← Back to the Prediction Registry