Vault
analysis

Regime Check — May 2, 2026

Created

Regime Check — May 2, 2026

Builds-on: regime-check-april-26-2026 Builds-on: portfolio-rebalance-april-2026 Builds-on: polly-fidelity-403b-allocation Related: macro-force-vectors-april-2026 Related: why-the-market-refuses-to-crash


TL;DR

Six days after the April 26 check, the regime is more contradictory than it was, not less. WTI broke $100, briefly hit $126 on April 30 (4-year high), now ~$101–106. And yet the S&P 500 hit an all-time high of 7,230 on May 1 with the best month since 2020. Stagflation flag still on. Powell stayed on as Fed Chair (didn't leave — he pushed back against Trump's legal attacks at the Apr 28-29 meeting). The Fed held at 3.5–3.75% with an extraordinary 8-4 dissent — the most divided FOMC since October 1992. Mag7 reported: 3 of 4 sold off despite beats on capex worries; Alphabet was the blowout (+10% on AI cloud +63% YoY).

The portfolio doesn't move. TIPS + energy is exactly the right posture for the Hormuz/inflation half. The S&P all-time high pulls toward "structural-bid masks decay" (why-the-market-refuses-to-crash) and away from "stagflation already in prices." Both views can be right at once: equities are riding the Fed put + passive flows, energy is pricing the supply shock, and bonds/credit haven't repriced either.

The interesting new signal: you're 38% on stagflation grind vs 2% creator consensus vs 20% market — you're early on the inflation second-wave from the energy pass-through. April CPI print (May 13) is the call.

The interesting new tooling: Gemini CLI now wired as a programmatic summarization backend, sidestepping the API-key 403s on the video-URL path. Backlog of stuck videos cleared. Multi-key API fallback also wired. FRED-lag fix shippedplab signals macro now fetches live CL=F futures alongside the lagged FRED close and flags divergence.

Correction notes: earlier drafts of this doc cited WTI at $99.89 (the FRED DCOILWTICO daily-close snapshot from plab signals macro). That FRED series was actually 5 days stale (last data point Apr 27, missing the entire Hormuz weekend spike). Also wrong in earlier drafts: equity-market characterization (S&P at all-time high, not stuck), Powell exit (he stayed), and the framing of the April 28-29 Fed meeting as a quiet hold (it was 8-4 dissent, last seen 1992).


Macro Deltas Since April 26 (Verified Against Live Sources)

Indicator Apr 26 May 2 Δ Read
WTI crude (CL=F front-month) $94 $101.94 spot (briefly $126 Apr 30) +$8 to +$32 intraday Hormuz escalated, not contained. Energy hedge ~13% ITM on cost basis.
WTI (FRED DCOILWTICO) $94 $99.89 (stale to Apr 27) 5 days stale; FRED publication lag during the spike
S&P 500 7,060–7,110 7,230 ATH (May 1) +1.7% to +2.4% Best month since 2020 in April. Killed the "zero-return-with-vol stagflation tell" framing from the Apr 26 doc.
CPI YoY (March print, latest available) 3.30% 3.32% +0.02pp March data flat. April CPI releases May 13 — first print to absorb the energy spike.
Fed funds target 3.5–3.75% 3.5–3.75% (held Apr 29) unchanged 8-4 dissent — most divided FOMC since Oct 1992. Miran wanted a cut; Hammack/Kashkari/Logan opposed an easing bias.
Powell status "likely last meeting" STAYED on the Board indefinitely binary flip Pushed back against Trump's legal pressure: "left me no choice." Continuity preserved.
10y–2y spread (verified vs live treasury) inverted +0.51% (10y 4.39 / 2y 3.88) uninverting Bond market not yet repricing the energy shock.
10y–3m spread +0.71% positive Recession-clock paused on yield-curve model.
HY credit spread (BAMLH0A0HYM2) 2.83% (16.8 pctile) "complacent" Credit hasn't repriced either.
UMich sentiment 53.3 low Consumer mood terrible despite ATH equity. Stagflation/uncertainty tell.
Recession P(12m) Estrella-Mishkin 15.3% low Yield-curve model says no — but the model wasn't trained on supply-side oil shocks.
Stagflation flag on STILL ON — both flags hot unchanged Inflation > 3%, slowing growth. Pass-through ahead.
10y breakeven inflation 2.48% Modestly above Fed's 2% target. Behind the energy curve.
Unemployment 4.30% (March) 4.30% (March, April releases May 8) unchanged Latest available is March data.

One-line read: the macro picture is internally inconsistent. Energy realized + Fed dissent + Hormuz unresolved + UMich at 53 should be a stagflation correction; instead the index hits all-time highs. This is the why-the-market-refuses-to-crash pattern — Fed put + passive flows providing structural bid, with the underlying decay still building. The bull and bear cases are both real. Portfolio posture (TIPS + energy + 32% US equity) participates in both.

Mag7 Earnings (Verified — All Reported Apr 28–30)

Co EPS beat Stock reaction Why
Alphabet $5.11 vs $2.48 +10% Cloud +63% YoY (consensus +47%). AI capex $185B paying off in revenue.
Apple beat +3% Posted Q2 beat after-hours Apr 30.
Microsoft $4.27 vs $4.06 −4% Beat. Azure +40%. Sold off on raised capex guidance.
Amazon $2.78 vs $1.63 +0.7% Beat hard. AWS +28%. Inline guidance penalized.
Meta $10.44 vs $6.67 −9% Massive beat. Raised FY26 capex to $125–145B (was $115–135B) on AI infra costs. Stock punished.

Pattern: 4 of 5 reported, all beat earnings, only Alphabet and Apple rallied. Meta and Microsoft bled despite beats. The market is bifurcating between "AI capex paying off" (Alphabet) and "AI capex not yet justifying itself" (Meta/Microsoft).

Read for your scenarios: this is partial dot-com scenario activation — not the AI capex unwind, but the start of the "winners and losers within AI" repricing. Your underweight US-large-cap (32% from 38.7%) and overweight small-cap value (~5% from 1.6%) are positioned for exactly this. Don't change the trade.

Hormuz Status (Verified May 1)

Read: the peace-talks-resolve-fast scenario from the Apr 26 doc is now lower-probability. Mojtaba's harder line + Trump's pause-conditional-on-full-reopening means the diplomatic off-ramp is longer than two weeks ago. Energy hedge holds; do not trim on talk-of-talks.


Prediction Market Priors (Manifold, May 2)

Scenario tag Market P Notes
hormuz_escalation (oil $150 EOY framing) 31.0% Two markets, weighted. One Hormuz market resolved this week.
recession (12m) 26.9% Aggregated from two questions.
stagflation_grind 19.9% Single thin market (Vol $402). Treat as soft signal.
dotcom_sectoral (AI bubble pop) 15.7% Two markets. AI capex still unresolved.

The Subjective vs Consensus vs Market Table

This is the most useful new view. Updated plab signals consensus with 14 days of fresh creator output (14/19 sources active, 58 items, 140 claims) plus current Manifold prices:

Scenario Your prior (Apr 26) Creator consensus Δ vs consensus Market Δ vs market
Stagflation grind 38% 2% −36pp 20% −18pp
Dot-com sectoral 18% 0% −18pp 16% −2pp
Hormuz escalation 27% 10% −17pp 31% +4pp
2008 systemic 14% 0% −14pp
Soft normal 3% 0% −3pp

Three ways to read this:

  1. Edge. Your prior baked in the macro-force-vectors three-lens framework (Crown × Brendan × Scanlon) plus the Hormuz casualty cascade you've been tracking since hormuz-to-ai-repricing-causal-chain. The creators and markets haven't fully digested the same model. You're early.
  2. Stale prior. April 26 you assigned 38% to stagflation grind partly because the March CPI print had just confirmed +90bps acceleration in a single month. That re-acceleration didn't extend (April CPI flat at 3.32%). You may need to mark stagflation down 3-5pp toward consensus.
  3. Methodological noise. "Creator consensus 0%" on most scenarios is structurally misleading — the field measures explicit scenarios_mentioned tagging, not whether creators discuss the underlying theme. Most creators talk about inflation/recession/AI without naming the scenario archetypes. The 0% is a floor, not a ceiling.

The most defensible move (revised after full verification — WTI, S&P ATH, Mag7 capex, Powell stay, 8-4 dissent):

Net redistribution: −2pp from soft-normal/crisis → +2pp into Hormuz/dot-com. Stagflation unchanged. Don't touch the portfolio, only the scenario weights for next-cycle thinking. TIPS and energy are exactly the two sleeves you want into a CPI print that absorbs $106 oil.


Asset Sentiment from Creators (Last 14 Days)

Asset Bull Neut Bear Net Read
Equity 0.7 0.0 1.5 −0.36 Creators bearish equity. Validates your 32% reduction.
Tech 2.2 1.1 2.0 +0.04 Mixed. AI bull/bear split.
Energy 2.1 0.4 1.1 +0.28 Net bullish. Supports your XLE position.

Top topics by weighted activity (last 14 days):

Topic Weight Leading-indicator content
geopolitics 11.9 1.1
policy 10.3 0.5
energy 9.6 3.0 ← strong leading signal
macro 8.9 3.4 ← strongest leading content
housing 6.5 3.4 ← unexpectedly active
recession 5.6 3.4 ← single creator, watch source

Housing flagged at 3.4 leading-indicator weight is the surprise. Worth a one-shot deep dive next week to see what's driving it. Could be permits, building starts, a mortgage-rate signal, or a sector-specific creator anomaly.


Portfolio Simulation — Updated Allocations

Both ryuhei-holistic.yaml and the new household-holistic.yaml reflect the executed April 26 rebalance. Re-run against world_2026_full (10K Monte Carlo paths):

Allocation Mean final Median final P5 (worst) Mean DD P95 DD
Ryuhei holistic ($165K) $77,007 $74,426 $43,101 27.8% 45.7%
Household holistic ($215K) $75,637 $72,897 $41,021 30.4% 49.6%

(Initial value normalized to $100K for both so they're directly comparable.)

Surprise finding: Polly's account, even after rebalancing to the three-fund mix, adds equity-concentration risk to the household in the simulation. Mean drawdown gets ~2.6pp deeper, P5 outcome ~5% worse.

The mechanic: Polly's 70% FFFHX target-date sleeve is ~54% US equity, which dilutes your heavier hedges (XLE, IAU, gold miners) when blended in at her 23% household weight. The TIPS bump (FIPDX adds ~5.8pp to household FIPDX exposure) only partially offsets it.

Implication: the polly-fidelity-403b-allocation doc was right that her account needed something (100% SPAXX was bleeding $600/yr in real terms), but the three-fund Path A is correct for her risk tolerance, not necessarily optimal for the household. If you wanted to fully tune to household scenario coverage, her FIPDX would be 35-40% instead of 25%, with a small XLE-equivalent if BrokerageLink ever opens up. Not actionable today — let her account settle, revisit at next regime check.

The 2008 historical stress (limited because most modern ETFs didn't exist in 2007-09) showed −42.4% household vs −36.3% Ryuhei-only. Same direction. Take it as directional only, not load-bearing.


Tooling Updates Made This Session

Two infrastructure improvements landed in tools/portfolio while doing this check, both addressing the Gemini summarization brittleness exposed last run:

1. Multi-key API fallback

gemini.py now reads GEMINI_API_KEY, then GEMINI_API_KEY2, ... up to N. On 403 PERMISSION_DENIED or 429 RESOURCE_EXHAUSTED, the wrapper falls through to the next key. Useful for quota dispersion across personal API keys. Wired via the new _FallbackGeminiClient class.

2. Gemini CLI as alternate backend

New _GeminiCLIClient shells out to the locally-installed gemini CLI (currently v0.40.1, OAuth-personal authenticated). Selectable via:

plab signals summarize --backend cli   # transcript-only, OAuth, ~60 RPM
plab signals summarize --backend api   # default, video URL mode, ~5-15 RPM

The CLI path uses gemini-3-flash-preview (Gemini Code Assist for Individuals consumer tier), which has a higher RPM ceiling than the API key tier. Tradeoff: no native schema enforcement, transcript-only (no YouTube video URL processing), and Google has flagged programmatic CLI use as a policy gray zone — keep it polite, don't run 24/7.

Backlog cleared this session: 19 transcripts that were 403'ing on the API path went through cleanly via CLI. All 14/14 succeeded on the final batch.

3. FRED-lag fix shipped in plab signals macro

DCOILWTICO is FRED's daily Cushing spot price sourced from EIA, and during the Hormuz spike the FRED publication lag was 5 days, not 1–2. The first draft of this doc cited $99.89 from this series while live spot was already $106+.

Fixed today:

Output now reads:

WTI crude (FRED close 2026-04-27)              $99.89    5d stale
WTI front-month (CL=F live, 2026-05-01)       $101.94    Δ +2.05 (+2.1%) — FRED is stale

If yfinance is ever flaky we still ship the FRED reading — the fallback is best-effort, not load-bearing.

4. Other macro readings — verified against live sources

Cross-checked the rest of the snapshot against the original sources to catch any other lag:

Indicator FRED snapshot Live verification Match?
10y-2y spread +0.51% 10y 4.39 / 2y 3.88 = +51bps (May 1)
HY OAS 2.83% 2.83% (April reading)
CPI YoY 3.32% March data; April releases May 13 ✓ (latest available)
Unemployment 4.30% March data; April releases May 8 ✓ (latest available)
Stagflation flag both hot Confirmed via UMich + CPI
Recession P (E-M) 15.3% Yield-curve-derived, model unchanged

The other indicators are all up to date — only WTI has the publication-lag problem because it moves daily and FRED's EIA upload was delayed.

5. The 470 transcript backlog isn't yet cleared

The deferred queue is 470 videos that need transcripts pulled (separate from the summarize step). YouTube transcript scraping rate-limits to 30/run by design. To work through it, run plab signals fetch on a daily schedule for ~16 days, or accept that the consensus dashboard operates on partial coverage. Not blocking the regime check.


What to Watch Next Two Weeks

  1. April CPI print (May 13). The single most important data point on the calendar. First print to absorb the late-April energy spike. Flat-or-down = your stagflation prior is too bearish; up = thesis confirmed and you're early.
  2. April jobs report (May 8). Unemployment, payrolls, wage growth. If unemployment ticks toward 4.5% with payrolls weak, recession + stagflation both strengthen. If labor stays tight, "no landing with inflation" view wins.
  3. Hormuz status. Iran's new proposal is on Trump's desk. Either accepted (oil falls fast, energy sleeve gives back), counter-offer (regime continues), or rejected (escalation toward $115+). All three are live.
  4. Whether the S&P 500 ATH holds. First 5%+ pullback since the run would tell you the structural bid is being tested. If it holds and we keep ripping, that's the why-the-market-refuses-to-crash thesis fully validated. Gives the FOMC dovish-leaning members more cover.
  5. Mag7 capex follow-through. Did Meta's −9% turn into a sector-wide repricing or just a single-name pullback? Watch SMH and SOXX through the next two weeks. Concentration in Mag7 is the dot-com scenario's mechanism — early signs would be 2-3 names breaking with the cohort.
  6. NVIDIA earnings (late May). The capex spend the other Mag7 are racking up flows directly to NVIDIA's revenue. NVDA print is the cleanest read on whether the AI capex cycle is plateauing or still accelerating.
  7. Housing leading-indicator follow-up. Why are creators flagging it at 3.4? One-week deep dive next time.
  8. Powell stayed — Warsh isn't a near-term variable anymore. Reframe: the bigger risk is now that the 8-4 dissent persists into the June meeting (which DOES have a dot plot). Watch Miran/Hammack/Kashkari/Logan public comments through May for signals.

What This Doesn't Recommend


Files Changed This Session


Sources

Tooling pulls (live this session)

External verification (web)

Internal