The AI capex week lands: Microsoft, Meta, Amazon
No precisely matched prediction-market contract is registered.
Difference: not measured. Edge: not measured.
This page keeps the original forecast, narrative evidence, statistical model, and prediction market in separate lanes. It shows changes without pretending that a later market quote was known earlier—or that a market price is a news conclusion.
No precisely matched prediction-market contract is registered.
Difference: not measured. Edge: not measured.
Latest market: 96.6%–97.9%; 2 frozen observation(s).
Difference: -0.25 probability points. Edge: withheld shared input and unverified costs.
Latest market: 97.2%–97.7%; 2 frozen observation(s).
Difference: -1.45 probability points. Edge: withheld shared input and unverified costs.
Latest market: 76.2%–76.4%; 2 frozen observation(s).
Difference: 17.7 probability points. Edge: withheld shared input and unverified costs.
No precisely matched prediction-market contract is registered.
Difference: not measured. Edge: not measured.
Latest market: 76.0%–77.0%; 3 frozen observation(s).
Difference: 1.5 probability points. Edge: withheld shared input and unverified costs.
No precisely matched prediction-market contract is registered.
Difference: not measured. Edge: not measured.
MacroGuru currently has 15 skill-eligible resolved event forecasts. Only 6 have matched market points, and zero have a matched pre-lock Narrative Agreement Index. That is below the registered 30-family gate.
Published MacroGuru Brier on the 15 independent event families is 0.106593; the fixed 50/50 comparator is 0.25; and the prequential Jeffreys climatology is 0.202005. The 6 market-matched families are reported in the JSON, but every matched outcome resolved true and no domain reaches the registered 10-family gate. That cannot establish a reusable weight.
Cost adjustment is withheld because the historical market points do not include executable quotes and verified fees. This is a negative result, not missing copy: Only 6 resolved event families have matched crowd points, none have a matched narrative index, and historical points lack executable quotes and verified fees. No weight or cost-adjusted edge is admissible.
Open the divergence collection → · Open the evaluation → · Inspect Decision Receipts →
Collection hash: 6c28aee3d5a75904e0fe0021b6c95f345f4e0eede0dbfced96ed10515b385e84
Evaluation hash: 45a75198bfd673893aa26f767dda05a3f2dc26014f08a02a2538884efb72eeba