The Model — Exhibit 1

Semiconductors +91% YTD. AI workforce: 123,653 cuts.

The market priced the boom; the workforce priced the cut. SOXX closed December 2025 at $313.69 and June 2026 at $599.73 — a +91% move in six months. Over the same period, AI and tech companies cut over 123,000 jobs, with AI cited as the reason rising from 7% in January to 40% in May. This is the top-level divergence signal: not a timing call, but a structural reading.

SOXX · month-end +91% YTD 2026 · $313.69 → $599.73
AI / tech layoffs 123,653 Jan–May 2026 total
AI cited in filings 7% → 40% Jan → May 2026

SOXX breadth — who is carrying the index

MU, INTC, and AMD carry the index; NVDA and AVGO lag — narrow breadth is its own fragility tell.

Fig. 1 — SOXX price vs. monthly layoffs (dual axis)
The market priced the boom; the workforce priced the cut. Jun layoffs bar is month-to-date.

SOXX month-end closes via Yahoo Finance (ADC /api/history). +91% YTD: Dec ’25 close $313.69 → Jun ’26 $599.73. +146% trailing 12 months. Layoffs: Challenger/Gray monthly reports; layoffs.fyi; Crunchbase. Jan–May 2026 total = 123,653. Consistent with ADC AI Layoffs tracker. Jun bar = month-to-date, faded. Fragility signal, not a timing call.

The tape beneath the tape — Exhibit 1b

D(t) = M(t) − G(t). Two tapes, one number.

Exhibit 1 above shows the macro divergence (SOXX vs layoffs). This gauge formalizes it: M(t) blends market-signal stress (momentum, vol, overextension on SOXX); G(t) blends four ground-truth deteriorations (layoffs, insider selling, depreciation stretch, capex-vs-demand gap). When they diverge, D(t) widens. Endpoints: −1.80 (2025 Q3) → +4.06 (2026 Q2).

Exhibit 1c · M(t) market signal

+2.83σ — the price tape at a one-year extreme

M(t) blends SOXX momentum, realized volatility (~74% vs ~25% 2025 median), and overextension. Jun 2026 reads +2.83σ — the market-signal side of D(t).

G(t) is ground-truth strength (falls as stress rises); layoff share (~7% → ~40% AI-attributed) is the documented mover. Middle-quarter M/G path is schematic pending full series — architect data-check pending. β-weights are a calibration call, not a vibe score.

Full §6 prose — D(t) = M(t) − G(t), ground-truth vs price tape, and divergence read.