LodeCraft / Yamarna, Western Australia / Gold
At Bloodwood and Renegade South, current candidate areas overlap recorded prospects where published drilling postdates our dated development assays.
Follow the evidence from measured gold and magnetic context to regional machine-learning response and candidate review areas. Each view answers a different question and points to the next useful geological review.
Analysis completed in 2026 using mixed-vintage inputs, including a 2019 geophysical compilation. This is a retrospective study, not a forecast made before the reported drilling.
01 / Evidence, response and review
Start with the ML response, then inspect candidate boundaries and the measured evidence. The final view shows where two similar ML methods disagree. Each map has a distinct role; different colours alone do not establish different scientific conclusions.
01 / Regional response
This ensemble ML map expresses the relative response learned from historical assay outcomes and geophysical predictors. It helps locate broad patterns for closer review.
Lower to higher relative model response
Colour is a within-map percentile, not a probability of gold. Historical assays influenced model fitting.
Open full-resolution map ↗Same regional extent. Original geometry. October 2026 outputs.
02 / Bring the evidence to a place
Known sites help put the current geometry in context. The following examples show exact overlaps and a near miss; they are selected explanations, not a performance sample.
The recorded prospect point sits inside a current candidate core.
Gold Road reported mineralised aircore drilling in 2018 and bedrock diamond drilling in 2019. Those milestones are later than the dated development assays in this study.
Next review: compare the candidate extent with the published drill sections and mineralised intervals.
2018 exploration report ↗ · 2019 drilling report ↗
The current candidate geometry includes the recorded prospect location.
A nine-hole programme completed in early 2022 reported mineralisation along this trend. This is a documented drilling milestone; the earliest recognition of the prospect remains to be established.
Next review: reconcile historical sampling dates and locations with the published drilling.
June 2022 quarterly report, p.11 ↗
The recorded mine point is about 264 m from the nearest core, inside its broader investigation envelope.
A near miss at a site point is different from an orebody assessment. The full deposit footprint may have a different relationship to the candidate geometry.
Next review: examine actual deposit extents at the model’s effective support scale.
Site coordinates: WA Government MINEDEX, extract dated 5 October 2026. Points are not deposit outlines.The wider picture matters: 4 of 51 recorded gold mine, deposit and prospect records in the study grid lie inside the current cores; 15 lie inside their broader investigation envelopes, including those four. Gilmour’s recorded deposit point lies outside both. These are descriptive spatial overlaps, not accuracy or discovery rates.
03 / What was known, and when?
The dates make the study more informative. Much of the assay evidence predates the reported Bloodwood and Renegade South drilling. The newer geophysical compilation adds another vintage of information.
WMC dated samples: 1997–1998. Asarco’s programme: 2002–2003, with laboratory results into January 2004. Gold Road development samples: April–September 2013.
Dates were checked against the original records for the 5,043 development rows. Some dates are missing, and historical availability is not fully verified.
Magnetic and radiometric products provide regional context. The product edition is dated 2019; individual surveys may have been acquired earlier.
A compilation date is different from a survey date. This is why the study does not have a single verified “2013 cutoff”.
Bloodwood: mineralised aircore reported in 2018 and bedrock drilling in 2019. Renegade South: drilling confirmation reported in 2022.
These milestones postdate the dated development assays. They do not establish when each anomaly or prospect was first recognised.
04 / Make the evidence useful
The value is in connecting the archive, seeing how different approaches express the evidence, and defining the next geological review. MindSync delivers that work as a service, using LodeCraft internally.
Keep source programmes, dates and uncertainty attached. Preserve the distinction between a measured assay, a model response and a reported drilling result.
Inspect measured gold and magnetics, the regional ML response and the candidate geometry. The model-difference view shows where two similar methods disagree without presenting their agreement as independent confirmation.
The current package contains 97 checked QGIS layers and the complete candidate geometry. It supports inspection of the evidence and preparation of local follow-up questions.
This graph deep-learning model produced a representation for each sample, rather than a continuous regional gold-prediction surface. The plot shows the first display component of those existing representations at 5,043 development locations.
The dots follow the sampling footprint. Their colours are learned-pattern values, not measured gold grades or predicted mineralisation. The first component captures about 50.3% of variation in the displayed embeddings; this is a display statistic, not predictive performance.
Experimental only: this output remains operationally unqualified and does not contribute to the selected target score. The separate geochemical DL ranking has zero valid cells, so no regional DL prediction is presented.
Open the experimental DL diagnostic ↗
Current presentation: October 2026 gold study. It contains 725 unranked candidate cores and no eligible scientific target ranking. All 8,013 within-area assay records were retained: 6,273 WMC, 718 Asarco and 1,022 Gold Road. The fixed population includes 5,043 development, 1,746 reserved and 1,224 buffered rows. The present narrative and date check use development evidence, not reserved assay outcomes.
Gold assay outcomes informed ML fitting; gold was not supplied as its own predictor raster. The reported-positive endpoint is an eligible analytical gold value above zero, not a geological anomaly threshold. The selected candidate score uses the two ML responses. The displayed experimental DL representation is separate and does not enter that score.
The DL map is a display-only first-component projection of existing graph-model embeddings, restricted to development sample locations. Its operational status remains blocked and it is not qualified for target scoring. The separate geochemical DL rank has zero valid cells and is not displayed as a prediction map.
| Source cohort | Core / ring | Numeric records, core / ring |
|---|---|---|
| WMC | 2 / 2 | 214 / 523 |
| Asarco soil 01 | 5 / 2 | 134 / 228 |
| Gold Road surface 01 | 2 / 2 | 68 / 288 |
Assay support varies by programme. These development-data summaries are descriptive, not independent validation. Only 79 of the 725 cores contain an included development sample. Absence from this sample comparison does not establish barren or historically unsampled ground.
All maps preserve source geometry. The main ML colours are within-map percentiles; the difference map subtracts percentile rankings on shared valid cells. Measured gold uses a logarithmic ppb scale without interpolation. The separate DL diagnostic shows learned-pattern variation. The computational grid is 250 m with an effective support floor of 1 km. Candidate overlaps use recorded site points, not orebody boundaries. The map catalogue is descriptive and does not establish mineralisation probabilities, new discoveries or drilling recommendations.
Retrospective case study by MindSync Technologies. References to exploration companies describe published source evidence and do not imply a client engagement or endorsement.
Start with the evidence you already hold
Bring your area, commodity and existing geological data. We can assess the available evidence, the questions it can support and a useful scope for review.
