Major Validations

Seven results anchor the program's public record, and one of them is reported open rather than confirmed.

The Crossing Calendar. MindCast published the free-cash-flow crossing schedule as a register of dated validation appointments rather than a forecast of prices. Oracle arrived on schedule at FY2026 free cash flow of –$23.7B, and the June drawdown traced to the financing and lease structure the register had flagged rather than to the cash burn itself.

The Crossing Was Never the Corrective Event. The controlling claim held that markets price evidence of conversion rather than contracted demand. Two July reports settled it in opposite directions within nine days: Amazon reported trailing free cash flow of negative $7.6 billion, raised capital guidance to $220 billion, and rose nearly 14 percent, while Alphabet registered a $514 billion cloud backlog and fell roughly 7 percent. A cash-burn thesis predicts both moves backwards.

Registration Lag, Tested in Nine Days. The July doctrine held that disclosure depth rather than capital scale governs the market's reaction to a capex report. Four hyperscaler earnings reports across nine days isolated the variable. Microsoft and Amazon connected spending to adoption, margins, and efficiency, and both re-rated. Meta published $31.1 billion of quarterly capital expenditure against free cash flow down 91 percent to $784 million, supplied no comparable conversion evidence, and fell hardest at 9 to 10 percent.

Capability to Authorization. MindCast forecast frontier-AI value migrating from what a model can do to who is permitted to deploy it. The Commerce Department's June 2026 trusted-partner allowlist delivered exactly that regime. The Anthropic export-control episode then supplied the mechanism at firm level: MindCast predicted the restoration would arrive capability-intact but access-gated rather than capability-reduced, with US-person access returning on or before July 17 (70–78%). Access returned July 1, gated at the authorization layer with the model untouched.

The Docket Migration. The accountability series predicted the center of gravity in AI litigation shifting from capability defendants to capacity and governance defendants — from what a company said its AI could do, to what it disclosed about the cost of running it. Apple and Tesla opened the era on capability claims. Barrows v. Oracle arrived February 3, 2026 on roughly $248 billion of off-balance-sheet lease commitments, and the Microsoft shareholder suit followed in June on capacity rationing behind a demand narrative — both against operators in the cohort's top tier.

Credit Differentiation Ahead of Operating Results. The register held that financing structure would separate issuers before any earnings figure did. S&P cut Oracle to BBB− against $638 billion of remaining performance obligations, technology spreads reached 89 basis points against the broader investment-grade market, and AI hyperscaler debt issuance reached $220 billion through August 10. Dispersion across issuers now carries the signal, which is what the register said it would.

The Telecom Rhyme — Open, Not Confirmed. MindCast staked the 2000 telecom cycle as a structural rhyme at 60–70 percent confidence: infrastructure capex outrunning cash generation, debt markets financing the gap, litigation lagging drawdowns by roughly six months. The first two elements confirmed. The distinguishing condition also held, and it cuts against the rhyme — today's borrowers carry monopoly-grade operating cash flows, so the cycle has repriced multiples without breaking balance sheets. The entry stays open, and the litigation-lag element settles on the next drawdown rather than on any date MindCast controls.

Major Outstanding Predictions

Every entry carries a deadline, a confidence band, and a condition that proves it wrong. Identifiers stay fixed once timestamped; the registry never renumbers a published claim.

🔎 AIRC-II.4 — Microsoft validation expansion. 65–75%. July 29, 2026 earnings materials materially expand disclosure in at least two categories among capacity economics, Copilot adoption, AI revenue contribution, and capital efficiency. Falsifier: qualifying expansion in fewer than two categories.

⚖️ AIRC-II.5 — Oracle structure registration. 70–80%. Disclosures, financing materials, or the Barrows record materially distinguish who owns and finances Oracle's AI infrastructure by December 31, 2026. Falsifier: no qualifying disclosure by the checkpoint.

📊 AIRC-II.1 — Gap-lag correlation. 60–70%. June 2026 idiosyncratic drawdown magnitudes rank-correlate positively with ex-ante disclosure-gap scores across the eight-firm cohort. Falsifier: Spearman correlation at or below zero. Scored no later than December 31, 2026, methodology frozen at publication.

♟️ AIRC-II.2 — First-mover reframe. 55–65%. At least one hyperscaler reframes capital guidance as milestone- or demand-gated by July 31, 2027, presented as measurement discipline, with announcement-week abnormal return no lower than −5 points. Falsifier: no reframe by the checkpoint, or a first mover punished below that threshold.

🏛️ AIRC-II.3 — Capital oversight formalization. 55–65%. At least two cohort companies disclose a board- or executive-level process connecting AI capital commitments to utilization, demand, or return thresholds by December 31, 2027. Falsifier: fewer than two by the checkpoint.

Event-triggered tier. Cohort language imitation following a successful reframe (60–70%), credit confirmation of durable stress within five trading days of a persistent firm-specific drawdown (55–65%), and a negative abnormal return on the first unreconciled revision of a newly introduced AI utilization metric (60–70%).

The frozen protocol. The gap-lag test locks its method before any return series has been examined. Disclosure-gap scores freeze first, using only documents available by May 29, 2026. The three-factor model runs second. June returns, corrective events, and litigation status stay out of the rubric, because each could smuggle the outcome into the score meant to explain it. Apple guards the design as a negative control — a material Apple gap score fails the design before any return is analyzed.

Amazon's crossing is in progress, carrying the cohort's most consequential unopened gap: diversified cash generation delays forced adjustment while AWS demand-allocation sensitivity builds the strongest next-test structure.