Paths

The Stage Two gauge says how far the shift to heavy debt and AI has gone, and which hedges work. A path says what happens next. Here are the five paths for the next decade, the one we are on, and the signs that would show us moving.

Where we are now
AMuddle-through

AI keeps improving but spreads through the economy slowly. Inflation runs 3 to 4%, the Fed keeps rates high, and the debt grinds higher without a crisis. Cash earns more than inflation. It can last for years.

3 of 4 signs

↗ Muddle-through drifts the gauge up slowly, now 75 (separation). At this level bonds are not a reliable hedge, so every “holds up” line below assumes that.

AI spending by the four big cloud companies is up 87% on a year earlier (Q2 2026). The model book: at base weights, with 64% in lines that tend to hold up here.

Five ways the decade could go

we are here
AMuddle-through
AI spreads slowly, inflation runs 3 to 4%, the debt grinds higher.
3 of 4 signs
Holds up
T-bills, gold, energy
with the gauge under 50: stocks, long bonds, gold
Stage Two gauge
↗ drifts up slowly
Book
at base weights
See the signs ↓
BAI boom
Profits catch up with the spending and growth outruns the debt.
0 of 5 signs
Holds up
stocks, compute, power
Stage Two gauge
↓ pulls back down
Book would tilt
toward AI chips and power
See the signs ↓
CAI bust
AI stocks fall 40 to 70%, then rates get held below inflation.
0 of 6 signs
Holds up
cash, TIPS, gold
with the gauge under 50: long Treasuries, cash, gold
Stage Two gauge
↑ pushes up
Book would tilt
toward gold and trend funds
See the signs ↓
DFast takeoff
AI does most office work by around 2030.
0 of 5 signs
Holds up
compute, land, gold
Stage Two gauge
↑ pushes up
Book would tilt
toward AI chips, power and gold
See the signs ↓
EWar shock
Taiwan or Hormuz, oil above $150. Can come at any time.
0 of 3 signs
Holds up
energy, defense, cash
with the gauge under 50: energy, defense, Treasuries, cash
Stage Two gauge
↑ pushes up
Book would tilt
toward energy, defense stocks and gold
See the signs ↓

How we decide: a scenario counts once half its signs are met. A break wins first (war, then bust), then AI paying off (fast takeoff, then boom). Otherwise we are muddling through. Who holds up is a judgment from history and research, not a forecast.

The signs

›AMuddle-throughwe are here3 of 4 signs

AI keeps improving but spreads through the economy slowly. Inflation runs 3 to 4%, the Fed keeps rates high, and the debt grinds higher without a crisis. Cash earns more than inflation. It can last for years.

  • Hyperscaler capex growth slows to 10 to 25% a year
    capex +87.0% y/y
    AI spreading slowly. Today's spending is still racing. As of Q2 2026.
  • Unemployment between 4 and 5%
    4.20% unemployed
    As of Sep 2026.
  • Fed funds above 3.5%
    Fed funds 4.00%
    As of Oct 24, 2026.
  • CPI inflation between 3 and 4%
    CPI 3.72%
    As of Sep 2026.
›BAI boom0 of 5 signs

Productivity jumps, profits catch up with the spending, and growth outruns the debt. Rates stay high because the boom soaks up capital. If this happens the debt half of our thesis is wrong.

  • Big-four free cash flow up on a year earlier while capex still grows over 15%
    free cash flow -$34B y/y · capex +87.0% y/y
    Spending that pays for itself is a boom, not a bubble. As of Q2 2026.
  • Productivity growth of 2.5% or more for four quarters running
    lowest of 4 qtrs 2.24%
    As of Q2 2026.
  • The S&P 500 within 2% of a record
    S&P 2.45% below record
    Markets price a boom before the productivity numbers show it. As of Oct 23, 2026.
  • AI lab revenue above $200B a year
    $115B a year
    OpenAI about $50B at end-September (net accounting) plus Anthropic $65B at end-July (Bloomberg 2026-08-17). From the scenarios research of 2026-10-23, not re-fetched. As of Sep 30, 2026. Source
  • 10-year above 5% with term premium falling and stocks in an uptrend
    10-yr 4.96% · term premium +0.27 pts in 3 mo · S&P +4.53% vs 200-day
    High rates from strong growth, not from fear of the debt. As of Oct 14, 2026.
›CAI bust0 of 6 signs

Spending runs ahead of what AI earns. A cloud company cuts its budget, the debt behind data centers cracks, and AI stocks fall hard, like telecom in 2000. Recession widens the deficit while energy keeps inflation sticky, which pushes toward repression. A normal 10% dip trips one of these signs; a bust trips several.

  • A big-four company cuts its capex guide
    0 cuts
    No big-four cut as of the July guides; Q3 reports are due the week of 2026-10-26. Update then. As of Oct 23, 2026. Source
  • Hyperscaler capex growth below 15% a year
    capex +87.0% y/y
    As of Q2 2026.
  • High-yield spreads above 5 points
    spread +3.40 pts
    As of Oct 22, 2026.
  • The S&P 500 below its 200-day average
    S&P +4.53% vs 200-day
    As of Oct 23, 2026.
  • AI chip stocks 25% or more below their one-year high
    AI chips 21.2% below 1-yr high
    The market usually calls a bust before the economy shows one. As of Oct 23, 2026.
  • The big four spending more than their businesses bring in
    free cash flow $6.7B
    Free cash flow below zero means the build-out runs on borrowing. As of Q2 2026.
What usually follows: repression1 of 10 signs, not counted toward the bust

Governments shrink their debt by keeping interest rates below inflation for years, the way the US did after 1946. Savers in cash and bonds pay for it.

  • Federal interest above 25% of receipts
    19.5% of receipts
    As of 12 months to Sep 2026. Source
  • The 10-year real yield below 1% with CPI above 3%
    real 10-yr 2.64% · CPI 3.72%
    As of Sep 2026.
  • The Fed cutting with CPI above 3.5%
    Fed funds +0.25 pts in 3 mo · CPI 3.72%
    As of Sep 2026.
  • Long rates rising while the Fed cuts
    10s-30s -0.13 pts in 3 mo · Fed funds +0.25 pts in 3 mo
    The bond market pricing the debt. As of Oct 22, 2026.
  • The Fed's balance sheet growing over 3% in 13 weeks with CPI above 3%
    balance sheet +0.09% in 13 wks · CPI 3.72%
    As of Sep 2026.
  • 10-year breakeven inflation at 3% or more
    2.34% breakeven
    As of Oct 23, 2026.
  • Stocks and long Treasuries both down over three months
    S&P +3.10% in 3 mo · long Treasuries -1.13% in 3 mo
    Stage Two's own definition. The hedge stops working. As of Oct 23, 2026.
  • Gold at a new record
    26.1% below record
    As of Oct 23, 2026.
  • Central banks buying 800 tonnes a year or more
    720 t a year
    Metals Focus full-year 2026 forecast, about 15% below 2025. A forecast, not a count. As of Oct 22, 2026. Source
  • Bitcoin moving more with gold than with the Nasdaq
    +0.36 gold less Nasdaq
    Six months of weekly returns. Bitcoin trading as money, not as tech. As of Oct 23, 2026.
›DFast takeoff0 of 5 signs

AI does most remote office work by around 2030. Profits soar while wages fall behind; most goods get cheaper while land, power and anything scarce gets dearer. Politics turns hard on the winners: windfall taxes, maybe public stakes in the AI labs.

How far AI has come: the longest task the best model finishes on its own, four times in five (METR, May 31, 2026).
3.1 hours
a workday
a work-week
  • An AI model completes a full workday's task 80% of the time
    3.1 hours
    The early warning. AI capability moves before jobs and productivity do. Claude Mythos Preview, 3h06m at 80% (May 2026). METR says its suite is near saturation above 16 hours. As of May 31, 2026. Source
  • An AI model completes 40-hour tasks 80% of the time
    3.1 hours
    A full work-week of independent work. Claude Mythos Preview, 3h06m at 80% (May 2026). METR says its suite is near saturation above 16 hours. As of May 31, 2026. Source
  • Productivity growth of 4% or more
    productivity 2.24%
    As of Q2 2026.
  • Recent-graduate unemployment at 8% or more while productivity grows 2.5% or more
    5.60% unemployed · productivity 2.24%
    Output rising while young graduates can't find work is the AI signature. Graduate unemployment alone isn't, because the NY Fed puts most of the rise so far on remote work. Q2 2026. The NY Fed attributes about 64% of the recent rise to remote work, not AI. As of Q2 2026. Source
  • A windfall tax or lab nationalisation advancing in Congress
    no
    No windfall-tax or lab-nationalisation bill advancing; proposals are talk only before the 2026-11-03 midterms. As of Oct 23, 2026.
›EWar shock0 of 3 signs

A blockade of Taiwan or a long closure of the Strait of Hormuz. Oil spikes, chip supply breaks, and emergency spending and price controls follow. It can hit during any of the other four.

  • Brent above $130 on a 30-day average
    Brent $122
    As of Oct 16, 2026.
  • Hormuz crude flows below half of prewar
    70.0% of prewar
    About 9.4M barrels a day in mid-October, roughly 70% of prewar (Kpler, as reported in the scenarios research). As of Oct 16, 2026. Source
  • A PLA quarantine or customs-inspection drill on Taiwan
    no
    No large encirclement drill in 2026; ADIZ sorties about 160 a month against about 300 in 2024 and 2025. As of Oct 16, 2026. Source

met   not met   no read. Signs are counted, never weighted, and no odds are printed.

Moving to another scenario does not change the model book on its own: each scenario names the tilt the book would take, and it applies only after review. Research and opinion, not investment advice.