Adds the evidence: indicator values, thresholds, sources, full grids.
Reading as
Depth and role are independent. Role sets a starting depth and pulls the modules you use first to the top; you can always go deeper.
Awaiting this week’s cycle.Every figure below is from the 2026-08-18 cycle and is
7 days old. The next scheduled run is Sunday 30 August, 08:00 UTC.
Posture map
The interpretation, separated from the observation
The monitor reports what the indicators read. This page reports what we make of them. Keeping the two on
separate surfaces is the point: a thesis that lives next to its own restated numbers can drift from them
without anyone noticing, and nothing on the page contradicts it.
Regime AMBERMethod asymmetric-investor-brief:2026-08Issue 7 · 18 Aug 2026
Every number on this page is rendered by reference from the monitor at page load. None of it is authored
here. If the monitor moves and this page does not, that is a bug in the render, not a difference of opinion.
Conviction is not one number
Structural · cyclical · tactical
A single conviction label averages disagreement away. Structural anchors, cyclical confirmation and tactical
stress are read on different horizons from different sources and they do not have to agree — and right now they
do not. Each class carries its own read, and where they conflict the conflict is the finding.
StretchedValuation and concentration anchors near record levels, none yet through warn.
Cycle Coincident (CC)
SofteningServices still expanding, underemployment drifting up.
Tactical Signal (TS)
BenignCredit spreads tight and tightening. This is the disagreement.
Why this is close to BIFURCATED
Anchors stretched against benign tactical pricing is the textbook shape of a market that has not yet repriced a structural problem. Under the K3 model this is close to BIFURCATED and is published as such rather than averaged away.
HIGH
Anchors, cycle and tactical readings point the same way.
MODERATE
Majority agreement with one dissenting class.
BIFURCATED
Classes disagree materially. Not a weak signal — a specific and informative one: it says the market and the structure are telling different stories, and it is often the most valuable state to publish.
LOW
No coherent reading. Breadth too thin to score.
How far forward the data reaches
Forward-looking · as of 18 August 2026 · no dated path publishedNo dated path, and the reason why
A quarter-by-quarter path out to 2033 is the most persuasive thing this site could draw and the least
supportable. The pipeline produces no dated forward path at all. Its forward-looking objects are a distribution over four named regime states, which carries no horizon of its own, and a 6–18 month safe-haven horizon. The
scenario frameworks above are that object, published with their horizons stated. Beyond them the honest
surface is this statement of absence, until a versioned projection object exists to replace it.
Horizon: 6–18 monthsNo quarterly path beyond it is published
The pipeline produces no dated forward path at all. Its forward-looking objects are a distribution over four named regime states, which carries no horizon of its own, and a 6–18 month safe-haven horizon. A quarter-by-quarter posture path to 2033 would be an authored shape, not a derived one, so it is not published. What would make one publishable is specified in ops/reboot/forward-posture-map: a versioned projection object carrying, per quarter, its basis, its inputs and a confidence interval.
Not in any pipeline object today
Per-quarter posture values beyond the current cycle
Confidence intervals on any forward value
A dated projection object with a method version
Posture in time
Backward-looking · record 24 June – 18 August 2026The cumulated record, nothing more
This is every reading the pipeline has ever published, drawn from the cumulated state rather than
authored for the chart. It is short, it is not on a regular cadence, and four of its ten entries repeat the
week before rather than reporting anything new. Those four are hatched. Bar colour is the regime published that
cycle, so the risk-on and risk-off read comes from the record rather than from a caption. An eight-week record with four
carried values is what the engine actually knows about its own history, and a longer, smoother line would be
a drawing rather than a record.
Scroll the chart sideways to see the full record →
Green — Risk-on
Amber — Caution
Carried forward, marked =
No score
Record length
8 weeks
Published cycles
10
Distinct values
5
Carried forward
4 of 10
Why this is the whole chart
Ten entries across eight weeks on an irregular cadence. There is no weekly guarantee in the record yet. Four entries repeat the previous score. They are drawn hatched and marked with an equals sign. A carried value is not a new observation and the chart says so rather than implying nine weeks of independent readings.
System average. Negative is deteriorating macro health, positive is improving. It is not the 0-100 posture composite and the two are not interchangeable.
Two scenario frameworks, named apart
Different questions, different horizons
Two probability sets over what looked like the same future used to sit on one page without a stated
relationship, which reads as an error. They answer different questions over different horizons. Both are shown,
and the sentence that connects them is below them rather than left to the reader.
Macro regime, 4 quarters
4 quarters
GMM regime_shift_probabilities
Stagflation persists55%
Deflationary bust30%
Inflationary boom10%
Goldilocks5%
Geopolitical escalation, 1–2 quarters ◇
1–2 quarters
Brief prose, pending a register per R24
Base case — constrained, no closure50%
Fast cascade — Hormuz disruption30%
De-escalation20%
How they relate
Two distributions are published because they answer different questions, and until this edition they were presented as if they answered the same one. The regime distribution asks which macro regime holds over the next four quarters. The escalation distribution asks how the current geopolitical episode resolves over the next one to two quarters. They are not alternatives and their probabilities do not add across frameworks. Any thesis card citing a probability names which framework it came from.
Standing theses
Each one carries what would break it
A thesis is admitted here only if it names the indicators that support it, the indicators that would falsify it,
the condition under which it stops working, and the case against it. The evidence rails read live from the
monitor — click any Evidence chip for the source, its tier and its revision policy.
Each card names what the thesis asserts about the macro state, what would falsify it, and the basis any model posture rests on. The one thing this site will not do is name an individual security to buy or sell. Asset-class and sector-level allocation, risk-on and risk-off posture, and the reasoning behind both are what it is for.
Thesis
Safe haven
Conviction MODERATE
In this regime, monetary-credibility hedges and reserve-diversification assets are absorbing flows that historically went to reserve-currency duration.
Asserted in Stagflation persists · Deflationary bust
Precious metals roughly +42% to record highs on joint safe-haven and AI-industrial demand (World Bank, this edition). The v2.0 metals formula scores monetary credibility and reserve diversification with VIX deliberately excluded, so the reading is not a fear proxy.
A disorderly widening in high-yield with metals also falling would break the thesis: it would say the flow is deleveraging, not reallocation.
Where this goes blind
In severe system-wide stress, reserve-currency sovereign bonds may rally as a flight-to-quality beneficiary regardless of the rates readings. In that scenario this thesis and the engine behind it both stop being informative, and the card says so rather than being quietly withdrawn.
The case against
The BIS finding that global equities fell about 9% during the Hormuz outbreak, against 17–19% in prior comparable energy shocks, cuts against a naive safe-haven story: the stress was absorbed without the rotation the thesis predicts. Published because omitting it would be the more expensive choice.
Blind-spot rules on these inputs
Nominal M2 mirage. Nominal money aggregates grow with inflation, so a liquidity reading taken in nominal terms looks supportive during exactly the regime that is destroying real balances. Liquidity is read in real terms or not at all.
Thesis
Tariff legal fragility
Conviction HIGH
The current tariff structure rests on a statutory basis that has already been narrowed once, and the market is pricing the level rather than the durability of the level.
Asserted in Stagflation persists · Inflationary boom
IEEPA tariffs struck down; Section 122 surcharge expired; USTR re-enacted under Section 301 against 60 economies. Three distinct statutory bases used inside one year.
A durable services expansion through a further legal reversal would say the structure is priced as noise, correctly.
Where this goes blind
The effective tariff rate has no automated read. This thesis is currently asserted on legal and narrative evidence with the quantitative anchor missing, which is disclosed rather than implied.
The case against
The same litigation risk that could remove the tariffs could also entrench them: a Section 301 basis is more durable than the one it replaced, so the fragility may be decreasing, not increasing.
Thesis
AI capex financing
Conviction MODERATE
Hyperscaler capital expenditure has shifted from cash-funded to debt-funded, which changes the transmission from an equity-multiple question to a credit question.
Asserted in Stagflation persists · Deflationary bust
Over $690bn committed for FY26, increasingly debt-financed; BIS has flagged the leverage build. Index top-10 weight at 40% against roughly 27% at the dot-com peak.
Capex growth holding positive while spreads stay tight through a full quarter of heavy issuance would falsify the credit framing.
Where this goes blind
Capex growth has no automated read either. Two of this thesis's four anchors are unwired, which is why its conviction is not HIGH.
The case against
Debt-funded capex at these credit spreads is rational, not fragile. The thesis depends on spreads widening before the assets depreciate, and that sequence is an assumption rather than an observation.
Blind-spot rules on these inputs
Earnings suppression. A valuation ratio falls when earnings are inflated by one-off or accounting effects, printing a false green on the anchors. Where the denominator is suspect the CAPE reading is annotated, not adjusted.
Spread compression under scarcity. High-yield spreads can tighten because issuance has stopped, not because credit has improved. A tight spread on collapsing issuance is a false green and is flagged when issuance data is available.
Asset class against jurisdiction
Directional scores, published as scores
The matrix is a scored read of macro conditions by asset class and jurisdiction — the same scores an allocation view is built from. A score of −0.55 says conditions in that cell are poor on our scale. Scores are published here as scores, with their basis; no individual security is named anywhere on this site.
Directional stress, −1 (most stressed) to +1 (most supported)
Asset class
US
Euro area
UK
Japan
China
Gulf / MENA
EM
Equities
-0.28
-0.41
-0.35
-0.12
-0.44
-0.18
-0.52
Credit
-0.36
-0.44
-0.38
-0.19
-0.55
-0.24
-0.61
Rates
-0.42
-0.33
-0.47
-0.55
-0.21
-0.16
-0.48
FX
+0.11
-0.29
-0.34
-0.46
-0.31
-0.08
-0.57
Commodities
+0.18
+0.09
+0.06
+0.12
-0.14
+0.34
-0.11
Policy stance, the input every thesis shares
Central banks
Three of the four indicators in the tactical-stress class price off policy expectations, so stance sits on the
interpretation surface rather than the monitor. Rows marked unreported are not neutral readings — they are
absences in this issue.
Federal Reserve
3.50–3.75%
Hold, 9–3 dissent
Easing bias
ECB
Post-June hike
Hold, energy bias
Tightening bias
Bank of Japan
~1.00%
Normalising
Tightening
Bank of England
—
Not in this issue
Unreported
PBOC
—
Not in this issue
Unreported
Declared gap · cross-monitor connections
The pipeline counts four cross-monitor flags for this issue — findings from the payments, AI-governance and
financial-integrity monitors that bear on a macro thesis. The count reaches this surface; the flag contents do
not. Until they do, the AI-capex thesis below is sourced only from macro indicators, and says so on its card.
Where this framework goes blind
Rules, not caveats
These are standing rules about how our own indicators mislead, not a disclaimer. Each one names the indicators
it applies to, and each one is attached to every thesis that leans on them.
Earnings suppression
A valuation ratio falls when earnings are inflated by one-off or accounting effects, printing a false green on the anchors. Where the denominator is suspect the CAPE reading is annotated, not adjusted.
Applies to Shiller CAPE, S&P 500 top-10 weight
Nominal M2 mirage
Nominal money aggregates grow with inflation, so a liquidity reading taken in nominal terms looks supportive during exactly the regime that is destroying real balances. Liquidity is read in real terms or not at all.
Applies to 30-year Treasury yield, Auction indirect bid
Spread compression under scarcity
High-yield spreads can tighten because issuance has stopped, not because credit has improved. A tight spread on collapsing issuance is a false green and is flagged when issuance data is available.
Applies to High-yield OAS
Private-credit opacity
Marks are model-derived and lag. Amendments suppress the default rate. The reported number is a floor, not an estimate.
Applies to Private credit default rate, BDC discount to NAV
Registered gaps
Known missing, not yet built
Investor positioning / sentiment overlay. BofA Fund Manager Survey not licensed. The sentiment_overlay field is null on every edition and will stay null until it is. Named so the empty field is not read as neutral.
Inflation momentum and policy-path expectations. No admitted indicator. The largest hole in the framework: inflation momentum, real rates and the priced policy path are all absent, in a regime the model itself calls stagflation.
Cross-asset correlation model. No correlation structure. The matrix reports directional stress per cell and cannot say whether cells move together, which is the question that matters in a cascade.
Real-time flow data. Weekly cadence and no flow feed. A gate or redemption event inside the week is invisible until the next edition.
Registered gaps are known unknowns: named, dated, and reducible by buying or building a feed. Epistemic blind spots (above) are structural — they cannot be closed by adding data, only annotated. They are held separately on purpose, because collapsing them makes the first look permanent and the second look fixable.
Limits of the model
Published in full on the method page
US-centric by construction
Every admitted tripwire is a US series. Non-US assessments are inferred from US conditions plus central-bank divergence, not measured locally.
Weekly cadence
The engine runs once a week. Intra-week discontinuities are invisible until the next edition, and the fastest scenarios are precisely the ones that resolve inside a week.
No cross-asset correlation model
Cells are scored independently. The model cannot tell you what happens when three of them move at once, which is what a cascade is.
Private-credit opacity
Marks are model-derived, amendments suppress defaults, and the disclosure cadence is quarterly at best. The private-credit readings are the least reliable on the board.
Intervention override
A central bank or treasury can invalidate a threshold overnight. The board measures conditions, not the reaction function.
No sovereign directional signal in systemic stress
In severe system-wide stress, reserve-currency sovereign bonds may rally as a flight-to-quality beneficiary regardless of what the rates readings say. The engine goes blind on sovereigns in exactly the scenario a reader most wants a sovereign view.
Has any of this worked before
Every accuracy claim lives on one dated page, labelled by basis.
No track-record claim appears on a live surface. The validation page states, for each episode, whether the
result comes from live publication or retrospective simulation — the two are never merged — and carries a
disconfirming-evidence field that cannot be left empty.