Predictions, Validations

Major Validations

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 — authorization, not capability, became the binding constraint on deployed capital.

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 and the Microsoft shareholder suit arrived on financing structure and capacity disclosure, against the two largest operators in the cohort.

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, FY2026 free cash flow at –$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 misses, graded at full size. [forthcoming]

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.

Core Publications

AI Repricing Cycle 2026 — Microsoft, Nvidia, Oracle, Meta and the Validation Tests Every AI Layer Now Faces — The diagnosis that opened the series. Names the validation regime, publishes the crossing calendar, maps the propagation loop connecting six ecosystem layers, and runs the Nash analysis showing why no hyperscaler can unilaterally cut capex without signaling structural surrender. The telecom 2000 cycle rhymes — capex outrunning cash generation, debt financing the gap, litigation lagging drawdowns by roughly six months at 60–70 percent confidence — with one difference containing the damage so far: today's borrowers hold monopoly-grade operating cash flows, so the cycle has repriced multiples without breaking balance sheets.

Microsoft, Oracle, Amazon and the Escape from the AI Repricing Cycle — Doctrine follows diagnosis. No firm escapes sector repricing, but every firm controls registration lag: the time between an internal change in AI economics and its credible external registration. Preventable exposure runs proportional to registration lag times expectation sensitivity times representation surface, and six levers close the four sequential delays plus two cross-system vulnerabilities. The governing simulation finding — the most dangerous signal is not missing data, it is accurate data answering the wrong question.

AI Accountability: When AI Promises Meet the Courts — The companion series, tracking the same gap where it lands in a courtroom rather than a print. One sentence holds it together: an AI-related claim becomes a legal liability at the moment the gap between the claim and the substrate beneath it can no longer be hidden. Four institutional types, each identified by the event that exposes it, plus a separate point-of-use reliability branch. The population-level finding: the docket's weight is shifting from the capability types that opened the era toward the capacity and governance types that now reach the largest operators.

Oracle, OpenAI, and the Capacity Economy — Inside the AI Infrastructure-Financing Lawsuit — Capacity accountability, anchored. Roughly $50 billion of capex in a single fiscal year assured to convert into revenue almost immediately, against $248 billion in off-balance-sheet lease commitments and a single counterparty projected to supply more than a third of future revenue under a commitment the buyer may be unable to fund. The type answers not whether the AI works but whether markets were told the truth about what running it costs.

The Microsoft Shareholder Suit and the Arrival of AI's Third Phase — Why the Next Competitive Edge Is Forecasting the Institution, Not Building the Model — Governance accountability, and the construct the whole capital series runs on: Governance Debt, the liability accruing when continuous operating reality outpaces a periodic disclosure rhythm. Shareholders contend a demand narrative outran disclosed capacity, and the gap surfaced in a single corrective session.

Tesla's Self-Driving Revolt: Full Self-Driving, Hardware 3, and the Warranty Substrate Apple's AI Illusion Already Mapped and Apple's AI Illusion: Narrative Control and the Law's Search for Structural Truth — The two capability cases that opened the docket, and the hinge between them. Apple anchors narrative arbitrage: a capability presented as ready, then deferred, after roughly $900 billion in market value had ridden on the timeline. Tesla anchors capability-to-substrate conversion, where a promise hardens into a fixed liability in the hardware meant to deliver it. Read together, they show a capability dispute becoming an infrastructure dispute once the substrate binds.

The Duty to Foresee — AI Deployment Readiness as Prospective Governance, and the Arrival of Agentic Duty of Care — Governance escapes the compliance frame by asking one forward question: what future disclosure could reasonably surprise investors? The Hand-test arithmetic prices the burden of asking against roughly $725 billion in planned annual deployment, against realized losses that already include a 27 percent monthly drawdown and two live securities dockets.

The Dual Nash-Stigler Equilibrium Architecture — Names the two exits from the hyperscaler capex trap: a new Nash focal point created by a first mover, or a Stigler-side institutional reset imposed from outside. The first mover chooses which one the industry gets.