The Arithmetic and the Horizon

In 60 seconds

The post-2008 US-led financial regime is arithmetically unsustainable across ten concurrent forces — sovereign math, equity valuations, demographics, Mag 7 concentration, AI capex, energy supply, crypto-dollar plumbing, periphery credit, political amplifier, and credit fragility. The watch window is 2026–2027. The diagnostic block (Essays 1–6) describes the configuration; the world-beyond block (7–9) shows what it looks like from outside the US balance sheet; the political-amplifier and mechanism blocks (10–13) name how it transmits and operationalise the dashboard; the navigation block (14–15) prescribes posture; the optimism essay (16) holds the bear regime and the long-arc capability expansion as two separate horizons. Read the Master Table first; everything else maps to it.

The fiscal math of the United States does not work.

It is not a question of will or politics. It is a question of arithmetic. And once you accept that the math does not work, every other big question — about asset prices, about the dollar, about jobs, about what happens to a country that has promised more than it can deliver — becomes a question of how the arithmetic forces itself to balance, and over what timeframe.

That sentence is the spine of the sixteen essays that follow.

---

Two disclosures

I write this from Cape Town: a South African chartered accountant looking at the United States from the outside. That vantage matters more than it might appear — when the global benchmark for fiscal sustainability has become "developed-world unsustainability," the emerging-market periphery becomes the comparative case, and the question reframes from "how do I survive a US fiscal reset?" to "what does it look like when the country that issues the unit of account becomes the worst credit in the developed world, and the country I live in becomes a comparatively better one?" I do not have a complete answer. The honest position from here is that the question is real, and the periphery is, for the first time in my career, a structurally favourable comparison.

I also run a business that automates professional-services work. I am, in the small, on the supply side of the labour displacement I will describe in the AI sections of this manuscript. Neither fact makes me less interested in getting the analysis right. Both should make me more honest about where I am sitting when I write it.

---

What this book is, and what it is not

This is a book that commits to a window: 2026 through 2027. The arithmetic in the diagnostic block converges there. The historical analogues run on a comparable cadence. The triggers in the dashboard either fire inside that window or they do not, and the manuscript is structured so a reader can hold me to it.

I am calling a prediction. I am not calling a date inside the window. The distinction matters: the configuration described across the next sixteen essays is mathematically forced to resolve through some combination of inflation, statutory cut, real-asset depreciation, or dollar devaluation — but the precise calendar inside the watch window is determined by which trigger fires first, and that is genuinely unknowable in advance. The dashboard in Essay 13 makes the falsification conditions explicit. If the indicators move against the bear case inside the window, the bear case is wrong on those indicators, and I am wrong on the prediction. That is the contract this manuscript honours.

The diagnostic block (Essays 1–6) sets out the arithmetic: sovereign math, aggregate equity valuation, the demographic configuration that underwrote the post-1980 era and has now reversed, the Mag 7 concentration and AI capex cycle binding the consumer base, the energy bottleneck binding the capex, and the historical base rate against which the configuration should be priced.

The world-beyond block (Essays 7–9) covers what the system looks like from outside the US balance sheet: the crypto-dollar plumbing migrating in plain sight, the Pettis-architected dollar system that absorbs the surpluses it produces, and South Africa as the comparative-institutional case where the diagnostic block's safety rails have already been removed.

The political-amplifier block (Essays 10–11) names the mechanism by which the system cannot reform in advance: an information environment that has lost the cost curve for persuasive falsehood, an electorate whose median balance sheet has no buffer left to absorb short-term reform pain, and a set of historical preconditions for successful pre-emptive reform that, on inspection, have all eroded.

The mechanism block (Essays 12–13) makes the structural fragility legible at the level of where it actually transmits — private credit, commercial real estate, the Japanese government bond market, the consumer, AI capex — and operationalises the dashboard against which the prediction is tested in real time.

The navigation block (Essays 14–15) is operational. Defensive posture for the watch window; opportunity posture for the rotations the diagnostic block has identified. I commit to specifics where the diagnosis earns them, and I have tried to refuse the flinch that says "individualised circumstances" when an honest answer is available.

The book closes with one essay (16) on the case for optimism — calibrated, evidence-based, and operating at a different time horizon than the diagnostic block. The bear regime is the watch-window prediction. The long-arc capability expansion is real, structural, and not in conflict with the diagnosis. The two horizons are the discipline this manuscript holds throughout, and the closing essay is where they meet.

---

The two-horizon discipline

The single biggest analytical mistake available in this kind of writing is to collapse two genuinely separate time-horizon claims into one.

Watch-window fragility is a 2026–2027 question. The diagnostic block treats it as one. The configuration described across the next eleven essays — diagnostic, world-beyond, and political-amplifier blocks — is, by any cross-check available in the historical data, the most concentrated form of macro risk in modern peacetime — high aggregate equity valuation, weak underlying cash-flow base, sovereign fiscal posture exhausted, household balance sheet thin, labour-supply tailwind reversed, dollar plumbing under migration, political-institutional capacity to pre-empt the arithmetic at multi-decade lows. The watch window is the window in which this configuration either crystallises or does not, and the manuscript treats the crystallisation as the base case rather than the tail.

Long-arc capability expansion is a multi-decade question. The closing essay treats it as one. Global poverty has fallen from 44% to under 10% in two generations; child mortality from 18% to under 4%; literacy from 12% to 87%; life expectancy from 32 to 73. None of these is a forecast. They are the long arc, and they are real, and they are not in conflict with the watch window's bear case.

The discipline is to hold both. Position for the regime in front of you. Do not let the regime poison your view of the arc behind and ahead of it. The two horizons require different evidence, different postures, and different reader-stances — and the failure mode of macro writing in either direction is to forget which horizon you are writing on.

---

What I refuse to claim

I am not going to predict a specific market level on a specific calendar day. The watch window is two years wide for a reason: the trigger that fires first determines the path through the resolution, and the trigger sequence is not knowable in advance even when the configuration is. What I am willing to commit to is the window itself, the dashboard that tests it, and the falsification conditions stated in advance. If the window passes and the configuration has not resolved, I am wrong on the prediction. The dashboard will say so before I do.

In the meantime, the world does not look like it is collapsing. The S&P is at all-time highs. Gold is at all-time highs. The job market is wobbly but functional. The Fed is dysfunctional but still operating. Life goes on, and goes on, and goes on, until it doesn't — and even then it usually goes on more than the doomers think and less than the bulls assert.

A useful frame for what follows: this is a book about the regime that you are already inside, written for readers who suspect the regime has changed and want to see the arithmetic of it before they reposition. It is also a book for readers who think nothing has changed, and who deserve to see the strongest version of the case that has. It does not require you to agree with the conclusion to find the data useful. The footnotes are real and checkable.

---

Nine months in

Added September 2026. Nine of the window's twenty-four months have passed, and the contract above says I owe a reading.

None of the conditions Essay 13 names as falsifying the regime claim has fired. The Fed has not pivoted: on 16 September it raised rates, twelve votes to none, with consumer prices up 3.4% on the year. No hyperscaler is near $100 billion of AI revenue. US fertility set a new record low, 1.5995 for 2024.¹

The stress that is visible is showing up where the Master Table's first row said it would transmit: in the long end of the curve and in the government's interest bill. The 30-year bond auctioned at 5.308% on 10 September, the highest since 2001. In the first eleven months of fiscal 2026, net interest passed $1 trillion, 1.16 times defence spending, and took 21.0% of federal receipts, against 19.9% a year earlier.² Office delinquency is back within 34 basis points of its record, and the largest private-credit funds are paying out redemption requests at their 5% caps.

The transmission the rest of the table leads with is not visible. S&P 500 earnings rose 52% in the second quarter, and the 493 companies outside the Mag 7 grew 31.8%, their best since 2021.³ The Fed chair's own reading in August was that credit spreads "are near the low ends of their historical ranges", and that real consumer spending had been "increasing more than 2 percent over the past four quarters".⁴ And the liquidity clock that Essay 13 treats as its master variable has met its own falsification condition.

So the timing claim is exposed on the equity and credit channels, and I am not going to rescue it by widening the window or relabelling the channel. What I will say is narrower. The rates channel now operating is not a mechanism found after the fact. It is row one as written in May: the discount rate on equity cash flows rises. So far it has compressed the forward multiple through rising earnings rather than falling prices.

The window closes at the end of 2027, and "quiet" needs a definition written now rather than then. I am using the two Essay 13 already carries, with their loose ends pinned there in September 2026. The equity channel has fired if Indicator 3's bear conjunction is met: Mag 7 earnings revisions down 10% on the year, and the S&P 500 forward multiple more than 25% below its 30 June 2026 reading of 20.4, both for two consecutive quarters. The credit channel has fired if Indicator 5's composite is met: two of its three legs past their lines for a quarter. If neither has fired by 31 December 2027, the prediction was wrong on the channels this Introduction put first, and the manuscript will say so in those words.

---

Reader contract

What follows is dense. Where I have leaned on a single authority — Hussman, Goodhart-Pradhan, Marks, Brynjolfsson, Howell, Pettis, Rosling — I have named them. Where my reading diverges from theirs I have named that too. The manuscript is more useful read in sequence than at random; Essay 12 (the empirical inventory of where the cracks already sit) depends on Essays 1, 2, 4 and 5 having already been done.

But every essay also stands alone. The cross-references are explicit. If you came in for the AI argument, the EM-credit argument, the energy argument, or the optimism argument, you can read those essays in isolation and the rest of the book will not have moved underneath you.

I have tried to write this with the discipline a reader is entitled to expect from a chartered accountant: every load-bearing number is sourced, every contested figure is flagged as contested, every counter-argument I am aware of has been engaged on its merits or noted explicitly where I have declined to engage it. The places where I am reasoning from priors rather than from data are marked. The places where my own position should make you read more sceptically are disclosed.

I will revise these essays. I will date the revisions honestly. The regime is what the manuscript describes; the timing is genuinely unknown; the discipline is to hold the analysis at the right resolution and refuse the confidence trap in either direction.

---

The Master Table

The table below is the spine of the manuscript. Every subsequent essay references one or more of the ten forces; the dashboard in Essay 13 is built against the trigger column; the hostile-reviewer appendix (Appendix A) tests the stressed column against the strongest available bull-side rebuttals.

Read it across, not down: each row is a complete diagnostic claim — force / baseline / what stressed arithmetic looks like / how it transmits to asset prices / where in history a comparable configuration resolved / what to watch in the watch window. The point is not that any single trigger fires on any single date. It is that the ten forces sit, today, in a configuration whose individual baselines are documented, whose stressed scenarios are arithmetically derivable, and whose transmission channels are legible.

Updated September 2026. The Mag 7, AI-capex, demographic, energy and political rows now carry September figures, and five of the triggers in the last column have resolved. The 2026 Trustees Report put OASI depletion in the fourth quarter of 2032. The hyperscalers' Q2 2026 earnings produced no capex-guidance cut; three raised. Anthropic's round closed on 28 May 2026 at a $965 billion post-money valuation, above the $900 billion that The Math Doesn't Work named as the signal. PJM's 2028/29 capacity auction cleared at the cap, 6,831 MW short of its reliability requirement. And the Fed Board test ran against the capture reading, with a 12–0 hike on 16 September.¹⁰

October 2026. The year's largest macro shock sits outside this table. The war between the United States and Iran that began on 28 February caused what the IEA called "the largest supply disruption in the history of the global oil market". Brent averaged $66.60 a barrel in January and reached $138.21 on 7 April. Headline consumer-price inflation went from 2.4% in February to 3.7% in September while core held at 2.6%, and on 16 September the Fed raised rates.¹¹ Row one's transmission, a higher discount rate on equity cash flows, has operated this year, but at least half of the 10-year yield's rise since April was the expected path of policy on both standard term-premium models, and the shock moved that path (The Survival Framework, §IV). The shock is a driver of row one, not an eleventh force. The condition it adds needs no date: if an exogenous energy shock lifts headline inflation, the Fed's response raises the discount rate whatever the fiscal path is doing, and that part of the transmission can reverse when the shock does.

A reader sceptical of the bear case has, in this single table, the entire load-bearing claim of the manuscript in seven columns. The strongest bull-side rebuttals to each row are engaged in Appendix A. The point of the table is not to assert that every row breaks in the watch window; it is to make the configuration visible at one resolution, so that the rest of the manuscript can be read against it rather than recapitulating it.

The dashboard in Essay 13 takes these ten forces, attaches a falsifiable threshold to each (compressed to five composite indicators), and reads the composite. As first published in May 2026, it counted three of the five at or above their thresholds. On the thresholds Essay 13 itself defines, the count was none, and as of September 2026 it is still none; the liquidity clock has met its own falsification condition. Essay 13 sets out the correction. The point is not to vote-count — it is to make the regime claim falsifiable on the axes the manuscript itself has named.

The math is the math. The timing is the question. The regime is what we should be looking at — and the regime, on the arithmetic, has already changed.

---

Cape Town, May 2026. Revised September 2026.

Footnotes

1. Sources for each figure are in the essays that carry them: the FOMC decision in The Survival Framework (Essay 14); AI revenue and the falsification conditions in Reading the Clocks (Essay 13); fertility from NCHS, Births: Final Data for 2024, NVSR 75-2. ↩

2. US Treasury, Monthly Treasury Statement, August 2026, by budget function: net interest $1,017.0bn and national defence $876.2bn, October 2025 to August 2026. Receipts are outlays less the deficit. Auction yield from TreasuryDirect. ↩

3. FactSet, Earnings Insight, 28 August 2026. Blended growth rates. ↩

4. Federal Reserve, "In Our Time", speech at Jackson Hole, 28 August 2026. ↩

5. Added October 2026. Bid-to-cover and indirect bidding measure rationing, and the 10-year auctions strengthened as yields rose: the 7 October 2026 auction drew a bid-to-cover of 2.77, the highest of the 22 ten-year auctions since January 2025 (FiscalData — Treasury securities auctions data). The 30-year auction's high yield reads the price at which demand clears. ↩

6. FactSet, Earnings Insight, 8 May 2026 and 15 May 2026: S&P 500 forward 12-month P/E 21.0 and 21.4, against a five-year average of 19.9 and a ten-year average of 18.9. Corrected October 2026: this cell first read ~23×. That was the reading Howard Marks gave for the end of 2024 in The Calculus of Value (August 2025), not a May 2026 one. ↩

7. Long-term mean from the Federal Reserve's Z.1 accounts: nonfinancial corporate equities (FRED NCBEILQ027S) divided by nominal GDP (FRED GDP), quarterly 1950 to 2026-Q2: mean 87% (95% since 1970). The same series reads 256% in 2026-Q2 and 163% in 2000-Q1, close to the Tudor Jones comparison of 252% and 170%. ↩

8. EIA, Today in Energy, 20 February 2026; PJM new supply from the 2028/2029 Base Residual Auction report. This cell previously read "~2% of new AI demand covered by net-new generation". The 2% is The Energy Bottleneck's coverage ratio for hyperscaler nuclear PPAs alone, not for all new generation, and the cell now says so. ↩

9. Marist Poll, October 2025: 30% agree that Americans "may need to resort to violence to get the country back on course". PRRI, Political Violence in America: 20% agree that "true American patriots may have to resort to violence to save our country" (the measure the signals dashboard tracks). ↩

10. SSA, 2026 Trustees Report summary; Anthropic, 28 May 2026; PJM 2028/2029 Base Residual Auction report; FOMC statement, 16 September 2026. Hyperscaler guidance and Amazon's free cash flow: The AI Reckoning. ↩

11. Start of hostilities: US Department of State, Office of the Legal Adviser, 21 April 2026: "On February 28, the United States Armed Forces launched Operation Epic Fury". IEA news release, 19 March 2026: "The war in the Middle East is creating the largest supply disruption in the history of the global oil market." FRED — Brent crude spot (DCOILBRENTEU): monthly average of daily prices $66.60 in January 2026; high of $138.21 on 7 April. BLS — Consumer Price Index, September 2026: all items up 3.7% and all items less food and energy up 2.6% over twelve months; February from FRED CPIAUCNS, 2.4%. FOMC statement, 16 September 2026. ↩