No change - 4 new data releases. S2 52% | S1 40% | S3 4% | S5 3% | S4 2% | S6 0%
| ID | Scenario | AI trajectory / Macro | Initial prior | Current | Week delta | Current prob. |
|---|---|---|---|---|---|---|
| S1 | Soft plateau | plateau / expansion | 15% | 39.5% | - | |
| S2 | Productivity wave | moderate / expansion | 30% | 52.1% | - | |
| S3 | AI abundance | fast / expansion | 10% | 3.8% | - | |
| S4 | Macro headwinds | plateau / contraction | 20% | 1.7% | - | |
| S5 | Structural disruption | moderate / contraction | 20% | 2.9% | - | |
| S6 | Hard landing | fast / contraction | 5% | 0.0% | - |
How sure of itself should the model be? It treats each signal as separate evidence and multiplies their effects together. Several signals overlap, such as the credit-card and mortgage delinquency rates, so one underlying development can be counted more than once and the result can look more certain than the evidence justifies. The evidence weight w softens every signal's effect: at 1 they count in full; at 0.5 each is halved in strength, so x4 becomes x2 and x0.25 becomes x0.5. Each row shows today's probabilities at that setting. If the full-strength row looks overconfident, for example if its contraction odds sit well below what markets or forecasters put on a recession, a lower weight is worth considering.
| Evidence weight | S1 | S2 | S3 | S4 | S5 | S6 | Contraction (S4+S5+S6) | Fast AI (S3+S6) |
|---|---|---|---|---|---|---|---|---|
| w = 1 (in use) | 39.5% | 52.1% | 3.8% | 1.7% | 2.9% | 0.0% | 4.6% | 3.8% |
| w = 0.5 | 29.1% | 47.2% | 7.4% | 6.9% | 9.1% | 0.4% | 16.4% | 7.8% |
| w = 0.33 | 24.5% | 42.7% | 8.6% | 10.5% | 12.5% | 1.1% | 24.1% | 9.7% |
No active signal changed state this week - the probability vector is unchanged.
Ratings reflect expected performance under each scenario. Robustness is probability-weighted using current scenario estimates. Strong Mixed Weak Speculative
| Asset class | S1 Soft plateau 40% | S2 Productivity wave 52% | S3 AI abundance 4% | S4 Macro headwinds 2% | S5 Structural disruption 3% | S6 Hard landing 0% | Robustness prob-weighted |
|---|---|---|---|---|---|---|---|
| Broad US index funds | Strong | Strong | Mixed | Weak | Weak | Weak | 94% |
| AI / tech (concentrated) | Mixed | Strong | Strong | Weak | Mixed | Mixed | 77% |
| International stocks | Strong | Mixed | Weak | Mixed | Mixed | Mixed | 68% |
| Real estate / REITs | Strong | Mixed | Mixed | Weak | Weak | Weak | 67% |
| Gold / hard assets | Mixed | Weak | Weak | Strong | Strong | Mixed | 24% |
| TIPS / short bonds | Mixed | Weak | Weak | Strong | Mixed | Mixed | 23% |
| Cash / T-bills | Weak | Weak | Weak | Strong | Strong | Mixed | 5% |
| Human capital / skills | Strong | Strong | Mixed | Mixed | Weak | Weak | 94% |
| Small AI co. exposure | Mixed | Strong | Speculative | Weak | Mixed | Mixed | 77% |
| Signal | Latest obs | Value | Change | 1y chg | State | Range |
|---|---|---|---|---|---|---|
| 10Y-2Y Treasury spread | 2026-10-16 | 0.35 | +0.10 4w | -0.23 | normalising | -0.05 1.05 ▲ Value 0.35 NEW |
| Credit card delinquency rate | Q2 2026 | 2.85 | -0.06 1q | -0.19 | stable | -0.30 0.30 ▲ 1y chg -0.19 |
| Mortgage delinquency rate (single-family) | Q2 2026 | 1.86 | -0.02 1q | +0.08 | stable | -0.30 0.30 1y chg +0.08 ▲ |
| Nonfarm labour productivity | Q2 2026 | 120.0 | +0.42 1q | +2.24% | historical range | 0.90 2.60 1y chg +2.24% ▲ |
| Unemployment, bachelor's degree+ | Sep 2026 | 2.50 | -0.20 1m | -0.30 | moderate | 2.40 3.60 ▲ Value 2.50 |
| Job openings (JOLTS total nonfarm) | Aug 2026 | 7,079 | -256.0 1m | +2.31% | stable | -11.00 11.00 1y chg +2.31% ▲ |
| Wage growth, bachelor's degree+ (Atlanta Fed) | Sep 2026 | 3.80 | -0.10 1m | -0.80 | moderate | 2.90 5.60 ▲ Value 3.80 |
| US investment grade OAS (ICE BofA) | 2026-10-08 | 0.82 | +0.02 4w | +0.07 | tight | 1.05 ▲ Value 0.82 |
Tracked for context but not used in the probabilities yet: each needs its thresholds and multipliers reviewed first (docs/DECISIONS.md). Set mode: active in config/signals.yaml to apply one.
| Signal | Latest obs | Value | Change | 1y chg | State | Range |
|---|---|---|---|---|---|---|
| Consumer vs mortgage DQ spread | Q2 2026 | 0.99 | -0.04 1q | -0.27 | stable | -0.30 0.30 ▲ Change -0.04 |
| Labour share, nonfarm business | Q2 2026 | 93.45 | -1.40 1q | -3.39% | capital share rising | -0.90 ▲ 1y chg -3.39% |
| College / high-school unemployment ratio | Sep 2026 | 0.62 | -0.02 1m | -0.02 | ratio stable | -0.06 0.06 ▲ 1y chg -0.02 |
| Unemployment, bachelor's degree, age 20-24 | Sep 2026 | 6.93 | -0.14 1m | +0.17 | stable | -0.60 0.60 1y chg +0.17 ▲ |
| Software development job postings (Indeed) | 2026-10-09 | 78.10 | +1.48 4w | +21.46% | growth | 4.00 1y chg +21.46% ▲ NEW |
| Sahm rule recession indicator (real-time) | Sep 2026 | 0.00 | +0.07 1m | -0.23 | normal | 0.32 ▲ Value 0.00 |
| 5y5y forward inflation expectation | 2026-10-16 | 2.33 | -0.02 4w | +0.06 | anchored | 2.43 ▲ Value 2.33 NEW |
| 10Y real yield (TIPS) | 2026-10-15 | 2.78 | +0.17 4w | +1.02 | high | 1.95 Value 2.78 ▲ NEW |
| METR 50% task time horizon, frontier | - | - | - | - | ||
| Big-4 hyperscaler capex guidance, current year | - | - | - | - | ||
| Job cuts citing AI (Challenger) | Aug 2026 | 3,462 | -7,508 1m | - | - | |
| Firms using AI (Census BTOS) | 2026-05-03 | 19.80 | - | - | - | STALE |
10Y-2Y Treasury spread [pct pts]
Credit card delinquency rate [percent]
Nonfarm labour productivity [index]
Unemployment, bachelor's degree+ [percent]
Job openings (JOLTS total nonfarm) [thousands]
Wage growth, bachelor's degree+ (Atlanta Fed) [pct (12-month median)]
US investment grade OAS (ICE BofA) [percent]
Labour share, nonfarm business [shadow] [index 2017=100]
Unemployment, bachelor's degree, age 20-24 [shadow] [percent (12-month avg)]
Software development job postings (Indeed) [shadow] [index Feb-2020=100]
Sahm rule recession indicator (real-time) [shadow] [pct pts]
5y5y forward inflation expectation [shadow] [percent]
10Y real yield (TIPS) [shadow] [percent]
Probability = initial prior x the current multiplier (LR) of each active signal, rescaled to 100% and recomputed every run. Active state highlighted yellow. (no change) = same state as the previous run.
Difference between 10-year and 2-year US Treasury yields. Negative (inverted) curve has preceded every US recession since 1970.
Switching buffer (hysteresis): 0.05 pct pts
Classified on: Value | Daily | Current value: +0.3500
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| inverted_and_widening | Value < 0 | x0.6 | x0.5 | x0.4 | x2.5 | x1.8 | x2.0 |
| normalising ◀ (no change) | 0 ≤ Value < 1.0 | x1.4 | x1.3 | x1.0 | x0.7 | x0.8 | x0.5 |
| expansion_signal | Value ≥ 1.0 | x1.8 | x2.0 | x1.5 | x0.2 | x0.3 | x0.2 |
Bright line: value < -0.5 → Yield curve deeply inverted - recession signal elevated
Share of credit card loans 30+ days past due at all commercial banks (seasonally adjusted). Credit cards only - the all-consumer-loans series is DRCLACBS. Rising rate signals consumer stress.
Switching buffer (hysteresis): 0.05 pct pts
Classified on: 1-year change | Quarterly | Current value: -0.1900
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| rising | 1-year change > +0.25 | x0.6 | x0.5 | x0.4 | x2.0 | x1.5 | x1.8 |
| stable ◀ (no change) | -0.25 < 1-year change ≤ +0.25 | x1.2 | x1.1 | x1.0 | x1.0 | x1.0 | x0.9 |
| falling | 1-year change ≤ -0.25 | x1.5 | x1.4 | x1.2 | x0.4 | x0.5 | x0.4 |
Share of single-family residential mortgages 30+ days past due at commercial banks. Pre-2008 crisis peak was ~2.5%; 2010 peak was ~11%.
Switching buffer (hysteresis): 0.05 pct pts
Classified on: 1-year change | Quarterly | Current value: +0.0800
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| rising_national | 1-year change > +0.25 | x0.5 | x0.4 | x0.3 | x2.5 | x2.0 | x2.5 |
| stable ◀ (no change) | -0.25 < 1-year change ≤ +0.25 | x1.4 | x1.3 | x1.2 | x0.6 | x0.7 | x0.4 |
| falling | 1-year change ≤ -0.25 | x1.6 | x1.5 | x1.3 | x0.2 | x0.3 | x0.2 |
Bright line: value > 3.5 → Mortgage DQ above pre-2008 level - immediate review
Output per hour worked in the nonfarm business sector. Sustained acceleration above 2.5% YoY is the primary positive confirmation signal for AI economic impact.
Switching buffer (hysteresis): 0.1 pct pts of YoY growth
Classified on: YoY % | Quarterly | Current value: +2.2400
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| accelerating_above_2pct | YoY % > 2.5 | x0.5 | x1.8 | x3.0 | x0.3 | x0.6 | x0.8 |
| historical_range ◀ (no change) | 1.0 < YoY % ≤ 2.5 | x1.3 | x1.2 | x0.8 | x1.0 | x1.0 | x0.8 |
| below_1pct | YoY % ≤ 1.0 | x1.5 | x0.7 | x0.3 | x1.5 | x1.3 | x1.2 |
Unemployment rate for workers with a bachelor's degree or higher (LNS14027662). Historically runs ~0.55x the overall rate (~2-2.5% in normal conditions). Rising relative to HS-graduate unemployment is the key S5 structural disruption signal.
Switching buffer (hysteresis): 0.1 pct pts
Classified on: Value | Monthly | Current value: +2.5000
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| elevated | Value > 3.5 | x0.3 | x0.5 | x1.5 | x0.7 | x3.0 | x2.5 |
| moderate ◀ (no change) | 2.5 < Value ≤ 3.5 | x1.3 | x1.3 | x0.8 | x1.2 | x0.7 | x0.5 |
| low | Value ≤ 2.5 | x1.8 | x1.5 | x1.0 | x0.8 | x0.3 | x0.2 |
Bright line: value > 4.0 → College-educated unemployment above 4% - monitor S5 signal
Total nonfarm job openings from the BLS Job Openings and Labor Turnover Survey. Used as a proxy for aggregate labor demand. Rising YoY = expanding hiring appetite; falling YoY = demand contraction. Monthly, released with ~5-week lag.
Switching buffer (hysteresis): 1 pct pts of YoY growth
Classified on: YoY % | Monthly | Current value: +2.3100
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| rising | YoY % > 10 | x0.8 | x1.5 | x1.1 | x0.5 | x0.6 | x0.3 |
| stable ◀ (no change) | -10 < YoY % ≤ 10 | x1.2 | x1.0 | x1.0 | x0.9 | x1.0 | x0.9 |
| falling | YoY % ≤ -10 | x0.8 | x0.7 | x1.2 | x2.0 | x1.8 | x2.2 |
Spread between credit card delinquency (DRCCLACBS) and single-family mortgage delinquency (DRSFRMACBS). Imperfect proxy - payment hierarchy means consumers default on mortgages last, so widening partly reflects prioritisation not quality. Retained as a noisy supplementary signal with reduced weighting (ratios closer to 1.0). Derived from existing FRED signals. Classified on the one-quarter change (delta_4w); before Sep 2026 that metric was always 0, so this signal never moved the vector.
Computed as the difference of: Credit card DQ (DRCCLACBS) | Mortgage DQ (single-family) (DRSFRMACBS) (aligned by observation date)
Switching buffer (hysteresis): 0.05 pct pts
Classified on: Short-horizon change | Quarterly | Current value: -0.0400
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| widening | Short-horizon change > +0.25 | x0.9 | x0.85 | x0.8 | x1.4 | x1.3 | x1.5 |
| stable ◀ (no change) | -0.25 < Short-horizon change ≤ +0.25 | x1.0 | x1.0 | x1.0 | x1.0 | x1.0 | x1.0 |
| tightening | Short-horizon change ≤ -0.25 | x1.2 | x1.2 | x1.1 | x0.75 | x0.8 | x0.7 |
Median 12-month wage growth for continuously employed workers with a bachelor's degree or higher, from the Atlanta Fed Wage Growth Tracker. Decelerating wage growth for educated workers is a leading indicator of structural disruption (S5). Monthly, ~6-week lag.
Switching buffer (hysteresis): 0.1 pct pts
Classified on: Value | Monthly | Current value: +3.8000
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| decelerating | Value < 3.0 | x1.2 | x0.7 | x0.4 | x1.8 | x2.0 | x1.5 |
| moderate ◀ (no change) | 3.0 ≤ Value < 5.5 | x1.2 | x1.0 | x0.9 | x1.0 | x1.0 | x0.9 |
| elevated | Value ≥ 5.5 | x0.7 | x1.5 | x1.8 | x0.4 | x0.3 | x0.2 |
Bright line: value < 2.5 → Educated-worker wage growth below 2.5% - structural disruption signal
Option-adjusted spread of the ICE BofA US Corporate (IG) index vs Treasuries. Market price of investment-grade credit risk. Directly addresses the existing bright line at 200bps. Pre-GFC normal 80-120bps; COVID peak ~373bps; GFC peak ~600bps. FRED reports it in percent (0.81 = 81bps), so thresholds here are in percent.
Switching buffer (hysteresis): 0.05 pct pts
Classified on: Value | Daily | Current value: +0.8200
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| tight ◀ (no change) | Value < 1.0 | x1.0 | x1.3 | x1.2 | x0.4 | x0.8 | x0.2 |
| normal | 1.0 ≤ Value < 2.0 | x1.1 | x1.0 | x1.0 | x1.1 | x1.1 | x0.9 |
| stress | Value ≥ 2.0 | x0.6 | x0.5 | x0.6 | x2.2 | x1.4 | x2.5 |
Bright line: value > 2.0 → IG spreads above 200bps - recession hedge review triggered (bright line Section 8)
Labour compensation as a share of nonfarm business output (BLS, index). Implements catalogue signal C2 (capital vs labour share) with a proxy: a falling labour share means a rising capital share. Q2 2026 was the lowest level on record.
Switching buffer (hysteresis): 0.1 pct pts of YoY growth
Classified on: YoY % | Quarterly | Current value: -3.3900
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| capital_share_rising ◀ (no change) | YoY % < -1.0 | x0.8 | x1.2 | x2.5 | x0.6 | x1.5 | x2.8 |
| stable | -1.0 ≤ YoY % < 1.0 | x1.3 | x1.2 | x0.8 | x1.1 | x0.9 | x0.6 |
| labour_share_rising | YoY % ≥ 1.0 | x1.5 | x1.0 | x0.4 | x1.3 | x0.8 | x0.4 |
Bachelor's+ unemployment rate divided by the HS-graduate rate, 3-month average. Implements catalogue signal B2 (historical ~0.55). Convergence toward 1 is the structural-disruption pattern. Carries bright line BL6.
Computed as the ratio of: Unemployment, bachelor's+ (LNS14027662) | Unemployment, HS no college (LNS14027660) (aligned by observation date)
transform ma3, Switching buffer (hysteresis): 0.01 ratio (3-month avg)
Classified on: 1-year change | Monthly | Current value: -0.0171
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| ratio_rising_toward_1 | 1-year change > +0.05 | x0.3 | x0.5 | x1.5 | x0.7 | x3.0 | x2.5 |
| ratio_stable ◀ (no change) | -0.05 < 1-year change ≤ +0.05 | x1.3 | x1.3 | x0.8 | x1.2 | x0.7 | x0.5 |
| ratio_falling | 1-year change ≤ -0.05 | x1.8 | x1.5 | x1.0 | x0.8 | x0.3 | x0.2 |
Bright line: value > 0.80 → BL6: College/HS unemployment ratio above 0.80 - escalate monitoring; structural disruption flag
Unemployment of young bachelor's graduates, 12-month average (the monthly series is not seasonally adjusted and very noisy). The entry-level channel where AI displacement shows first (Stanford/ADP 'canaries' research). Display only until LRs are set.
transform ma12, Switching buffer (hysteresis): 0.1 pct pts
Classified on: 1-year change | Monthly | Current value: +0.1667
| State | Boundary | Likelihood ratios |
|---|---|---|
| rising | 1-year change > +0.5 | not set (display only) |
| stable ◀ (no change) | -0.5 < 1-year change ≤ +0.5 | not set (display only) |
| falling | 1-year change ≤ -0.5 | not set (display only) |
Indeed job postings index for software development. Implements catalogue signal A3 (white-collar postings) for SOC 15 only; legal and business/financial postings are not yet covered.
Switching buffer (hysteresis): 1 pct pts of YoY growth
Classified on: YoY % | Daily | Current value: +21.4600
| State | Boundary | xLR per scenario | |||||
|---|---|---|---|---|---|---|---|
| S1 | S2 | S3 | S4 | S5 | S6 | ||
| accelerating_decline | YoY % < -15 | x0.3 | x0.6 | x1.8 | x0.7 | x2.5 | x2.0 |
| flat | -15 ≤ YoY % < 5 | x1.2 | x1.3 | x0.8 | x1.0 | x0.9 | x0.7 |
| growth ◀ (no change) | YoY % ≥ 5 | x1.5 | x1.8 | x1.2 | x0.5 | x0.3 | x0.2 |
3-month average unemployment rate minus its low over the prior 12 months. 0.50+ has marked the start of every recession since 1970. Macro-axis contraction detector. Display only until LRs are set.
Switching buffer (hysteresis): 0.02 pct pts
Classified on: Value | Monthly | Current value: +0.0000
| State | Boundary | Likelihood ratios |
|---|---|---|
| triggered | Value >= 0.5 | not set (display only) |
| rising | Value >= 0.3 | not set (display only) |
| normal ◀ (no change) | Value ≥ 0.3 | not set (display only) |
Market-implied average inflation for the five years starting five years ahead. Inflation-regime overlay: the six scenarios have no inflation axis, and the 2026 oil shock plus Fed hikes make it relevant to the asset matrix. Display only.
Switching buffer (hysteresis): 0.03 pct pts
Classified on: Value | Daily | Current value: +2.3300
| State | Boundary | Likelihood ratios |
|---|---|---|
| unanchored | Value > 2.75 | not set (display only) |
| elevated | 2.4 < Value ≤ 2.75 | not set (display only) |
| anchored ◀ (no change) | Value ≤ 2.4 | not set (display only) |
The inflation-adjusted yield on 10-year Treasuries (TIPS). Overlay with no LRs, because the direction of its effect depends on why it moved: expected policy rates (tighter Fed; later credit-stress risk, S4), term premium (fiscal and uncertainty; mostly asset prices), or a higher neutral rate (investment boom, faster growth; S2/S3). Read it with labour share: a falling labour share with rising real rates points to an investment-led boom; with falling real rates, to demand-deficient disruption (S5). The right-hand chart splits the nominal 10Y yield into term premium and expected short-term rates.
Switching buffer (hysteresis): 0.05 pct pts
Classified on: Value | Daily | Current value: +2.7800
| State | Boundary | Likelihood ratios |
|---|---|---|
| high ◀ (no change) | Value > 2.0 | not set (display only) |
| moderate | 0.5 < Value ≤ 2.0 | not set (display only) |
| low | Value ≤ 0.5 | not set (display only) |
Length of software task (human time) the best frontier agent completes with 50% success, per METR. Catalogue signal A1 (capability trajectory). Note METR's suite is unreliable above ~16h.
Sum of Microsoft, Alphabet, Amazon and Meta capex guidance for the current calendar year, updated after each earnings round. Catalogue signal A2.
Monthly announced US job cuts where AI is the stated reason (Challenger, Gray & Christmas).
Share of US businesses using AI to produce goods or services (Census Business Trends and Outlook Survey, biweekly). Diffusion, as opposed to frontier capability.
Pre-committed action triggers. Written before data is observed to prevent post-hoc rationalisation.
| ID | Condition | Action | Status |
|---|---|---|---|
| BL1 | S5 > 0.35 | Structural disruption > 35% - review credit-sensitive real estate exposure | Not triggered |
| BL3 | S3 > 0.25 | AI abundance > 25% - review equity concentration; increase hard asset allocation | Not triggered |
| BL4 | BAMLC0A0CM > 2.0 AND T10Y2Y < 0 for 182+ days | IG OAS > 200bps AND 10Y-2Y inverted > 6 months - classic recession hedge review | Not triggered |
| BL5 | S6 > 0.15 | Hard landing > 15% - review liquid emergency reserve; do not ignore | Not triggered |
| T10Y2Y | 10Y-2Y Treasury spread < -0.5 | Yield curve deeply inverted - recession signal elevated | Not triggered |
| DRSFRMACBS | Mortgage delinquency rate (single-family) > 3.5 | Mortgage DQ above pre-2008 level - immediate review | Not triggered |
| LNS14027662 | Unemployment, bachelor's degree+ > 4.0 | College-educated unemployment above 4% - monitor S5 signal | Not triggered |
| ATLFED_WAGE_BACH | Wage growth, bachelor's degree+ (Atlanta Fed) < 2.5 | Educated-worker wage growth below 2.5% - structural disruption signal | Not triggered |
| BAMLC0A0CM | US investment grade OAS (ICE BofA) > 2.0 | IG spreads above 200bps - recession hedge review triggered (bright line Section 8) | Not triggered |
| EDU_UR_RATIO | College / high-school unemployment ratio > 0.80 | BL6: College/HS unemployment ratio above 0.80 - escalate monitoring; structural disruption flag | Not triggered |
Generated 2026-10-16 - AI Scenario Monitor (method v2)