The Scenarios in Detail — how A and B would unfold

This ledger measures one question every day. It is not “when will AGI arrive?” It is a race: can the AGI-driven growth produced by Big Tech capital expenditure outrun the high-rate regime? AGI arriving and that growth defeating the burden of rates are two different events. Even if AGI arrives, if it never shows up in productivity and fiscal numbers, the race is lost anyway. So instead of predicting a single date, this ledger defines both sides of the race as scenarios and stacks evidence on each.

Scenario A — growth dissolves the high-rate problem

Scenario A holds when the following mechanisms engage in sequence.

• A productivity surge. AI moves from laboratory performance into real-economy productivity. The marker is nonfarm productivity growth sustained at 3% or more.
• Real growth above baseline. Real GDP growth runs persistently above the Congressional Budget Office (CBO) baseline — for example, 3.5% or more. Growth lifts tax revenue.
• A stabilising fiscal path. On revenue upside, CBO deficit projections are revised down, and the debt-to-GDP ratio stabilises or falls.
• Rates fall the “good” way. Real yields decline through growth absorption rather than forced easing — the market reads high rates as the price of growth, not as issuance risk.

Along this path, high rates stop being a weight on the whole economy. Under the principles of the desk that maintains this ledger, any response to accumulating A-side evidence is discussed only at quarterly gates.

Scenario B — the thesis breaks and high rates persist

• Slipping milestones. Frontier progress slows or benchmarks stagnate. If the capability curve bends, the entire growth narrative is pushed back.
• A rising term premium. Long yields rise even though the Fed is not hiking. The market begins charging a premium on long maturities for the fiscal path — the so-called bond vigilantes.
• Weak long-end auctions. Treasury auctions at long maturities repeatedly draw soft demand.
• Fiscal projections worsen. CBO deficit projections are revised up.
• A constrained Fed. It becomes explicit that, despite fiscal pain, the Fed cannot cut.
• A weaker dollar, faster gold buying. The dollar weakens while gold buying and central-bank gold accumulation accelerate — a signal of distrust in dollar credit itself.

In B, high rates persist and their burden passes into the real economy through the refinancing wall. The desk's record (9 October 2026) even notes the expected order of transmission: private credit first, then the Russell 2000 (mid-2027), investment-grade and high-yield refinancing (2028), commercial real estate (2028-2030), and the US consumer last — with the damage steepening from late 2027 and an inflection in 2028.

Why two scenarios and a gauge, not a single prediction

A single-date prediction is discarded when it misses, and when it hits, skill cannot be told from luck. Framed as a race, each day's material can be classified as evidence for one side or the other — and even when a classification is later judged wrong, it stays in the record and can be checked.

The gauge is not a probability. It is the ratio of weights of directional evidence accumulated in this ledger, nothing more. If the gauge reads A 46%, that does not mean “Scenario A has a 46% chance”; it means supporting evidence carries 46% of the total weight of directional evidence recorded so far. Neutral evidence is excluded from both sums.

The desk also writes down, in advance, what would move the gauge. For the 15-30 October 2026 window the agreed triggers were: CapEx guidance maintained and no FOMC hike on 28 October → 55:45 toward A; an FOMC hike, or a guidance cut together with CFO financing language → 45:55 toward B. Writing the conditions before the outcome is what prevents the interpretation from being rewritten after it.

Daily Watch, Quarterly Act

Evidence is recorded daily. But scenario judgements, and any decisions attached to them, are made only at quarterly gates. Revising a conclusion in reaction to a single day's figures is how noise gets mistaken for signal. The daily pages of this ledger are observation; judgement is quarterly work.

The daily tells — three gauges read every day

Between gates, three daily tells take the race's temperature.

(i) The 10-year TIPS real yield, read with breakeven inflation

The two series are read together. If the real yield rises while breakeven inflation falls or holds stable, that is an A-side reading — the shape of “good” high rates, absorbed by growth. If real yields and breakevens rise together, that is a B-side reading — the stagflationary shape of high rates.

(ii) Gold versus the real yield — the fastest tell

Gold normally falls as real yields rise. If real yields are above 2.5% and gold falls in textbook fashion, that is an A-side signal. If real yields are high or rising and gold rises anyway, the market is read as beginning to distrust dollar credit itself — a strong B-side signal. Because it reacts before the others, this ledger treats it as the fastest tell.

(iii) The high-yield spread (HY OAS) bands

If spreads stay ultra-stable below 300bp despite high rates, that is A-side: the corporate sector is absorbing the rate regime. A spike above 450-500bp is B-side: marginal firms and commercial real estate cracking — a precursor of credit seizure.

Read on the latest prints (ledger close, 26 October 2026): the TIPS real yield stood at 2.66% on the official Treasury figure (23 October) with breakevens around 2.34% — high real yields with stable breakevens, the A-side configuration of tell (i). Gold, after a -4.1% week, was testing the $4,000 level around $4,030: gold fell while the real yield stood at 2.66%, the textbook direction for tell (ii) — though with the support test unfinished it is not a settled reading. HY OAS was 340bp (+13bp on the day), in the middle band above 300bp and below 450bp. The three tells do not always agree, which is why the gauge is read as the sum of the whole ledger, not from any single indicator.

The financing watch — the upstream causal chain

A guidance cut is a signal that arrives after the fact. The financing sits upstream of it. The “financing bottleneck” flagged by Broadcom is registered in this ledger as an upstream cause that could produce a future CapEx guidance cut. The chain runs: financing costs rise → the language appears in CFO remarks on earnings calls → guidance is cut → the input reaches the desk's pre-committed circuit breaker (guidance cut for two consecutive quarters, or an explicit AI-capex cut). The watch is aimed at the front of that chain.

The Rule A yellow card is raised only when two or more of the following three legs hold at the same time. As of 26 October 2026, the count is 0 of 3.

• Leg (i) — CFO language. An explicit mention, on an earnings call, of external financing costs or payback constraints. The judging window is the earnings season of 27-29 October 2026.
• Leg (ii) — guidance. CapEx guidance from the Big Four (Microsoft, Google, Amazon, Meta) in aggregate — or from a key name — is actually cut. Judged in the same window.
• Leg (iii) — credit and rate tripwires. Any one of these establishes the leg: the 10-year yield touches 5.50%; the 10-year closes at 5.00% or above for 60 consecutive trading days, counted from 14 October 2026 (any single close below 5.00% resets the count to zero); HY OAS reaches 400bp or more; or the 5-year CDS of two or more of the Big Four widens by +50bp or more against its 14 October 2026 baseline, subject to an absolute floor of 80bp.

What does not count is also fixed in advance. The September 2026 issuance cliff (from $113B to $23B) is read as supply held back after June front-loading, not as demand refusing — it does not count. Single-name CDS records at Oracle, Broadcom, or Nvidia reflect vendor-financing costs rather than stress at the Big Four payers; they are watched, but they do not count toward the leg.

November carries one sub-checkpoint that is not a leg: the outcome of Anthropic's IPO (success and pricing, versus further delay or withdrawal), and whether vendor-financing CDS stress spreads.

Read through this frame, the state on 26 October 2026 is: Meta's large equity raise (a $35B concurrent offering plus a planned $60B ATM in 2027) is not leg (i). It mentioned no payback constraint; it was financing raised to press ahead, and CapEx guidance of $135-145B was reaffirmed the same day. The 60-day count stands at day 1 — the 23 October close was exactly 5.00% — and resets on any close below 5.00%.

The judgement frame — timing, and what the gauge is not

The desk's own consensus on timing, reached at the end of a three-round debate, is a median of 2030, range 2029-2031. The distribution of outside forecasts is far wider (lab leaders' remarks cluster in 2026-2030; a strict-definition prediction market's median sits in early 2033; a large researcher survey's median is 2047).

December 2027 is set down as the “judgement month”. That does not mean AGI is expected to arrive that month; it is the point fixed in advance at which the evidence accumulated by then will be used to settle the thesis's cruxes.

And once more, plainly: the gauges and ratios on this page are not probabilities. Within the sample that is this ledger, they are a tally of how directional evidence has stacked up on each side.