Cascade Analysis Language — Deterministic Execution Engine

A domain-specific language for modeling how failure — and success — propagates across organizations. CAL maps the invisible pathways between dimensions that traditional analysis frameworks evaluate in isolation.

When to use CAL / 6D instead of traditional frameworks:

• SWOT identifies strengths and weaknesses — but not how they cascade into each other
• Porter's Five Forces maps competitive pressure — but not propagation across dimensions
• PESTEL lists macro factors — but not the sequence in which they activate
• 6D Cascade Analysis traces how a disruption in one dimension propagates through all six — scored, sequenced, and reproducible

The keywords are the methodology: FORAGE to sense signals, DRIFT to measure gaps, FETCH to decide when to act, RECALL to validate predictions.

Built on the 6D Foraging Methodology. Battle-tested across 228+ published case studies spanning 148+ sectors — including banking, tech, sports, insurance, weather-ai, cybersecurity, automotive, geopolitics, agriculture, beauty-healthcare, and SMB — with FETCH scores ranging from 898 to 4,461.

Lineage: Created by a founding contributor to .netTiers (2005–2010), one of the earliest schema-driven code generation frameworks for .NET. Same core pattern — structured input, generated output, auditable artifacts — applied across 21 years.

-- Silicon Valley Bank: 6D Cascade Analysis
-- Sense → Analyze → Measure → Decide → Act

FORAGE banks
WHERE asset_liability_mismatch > 50
  AND uninsured_deposits > 85
  AND cro_vacancy IS "18 months"
ACROSS D5, D1, D3, D4, D6, D2
DEPTH 3
SURFACE svb_cascade

DIVE INTO deposits
WHEN withdrawal_rate > 1000000    -- $1M per second
  AND uninsured_ratio > 0.85      -- AND chaining supported
TRACE cascade
EMIT bank_run_signal

DRIFT svb_cascade
METHODOLOGY 90                    -- expected risk detection capability
PERFORMANCE 15                    -- actual: audits passed, cascade invisible

FETCH svb_cascade
THRESHOLD 1000
ON EXECUTE CHIRP critical "6/6 dimensions compromised in 48 hours"

SURFACE analysis AS json

-- UC-062: The Escape Hatch — Review at 30 days
RECALL escape_hatch ON "2026-04-15"

  WATCH compression_ceiling STATUS fired
    FIRED_DATE "2026-02-26"
    EVIDENCE "C3 AI layoffs produced stock decline."

  WATCH consumer_collapse STATUS not_fired
    EVIDENCE "NFP remained positive through review window."

  TRIGGERS 1/2
  CONFIDENCE_STATED 0.33
  CONFIDENCE_ACTUAL 0.50
  CALIBRATION aligned

SURFACE validation AS json

RECALL closes the loop: SENSE → ANALYZE → MEASURE → DECIDE → ACT → VALIDATE. Every prognostic case with WATCH triggers can be formally validated when the review date arrives.

This script is from UC-039: The 48-Hour Cascade — the highest FETCH score (4,461) in a library of 228 published case studies. Watch the 70-second video breakdown.

WATCH now supports an optional FOR clause that requires a condition to hold across N consecutive periods before the trigger fires. A single bad quarter is noise; a condition sustained across multiple periods is a structural signal.

-- Point-in-time (unchanged — fires on first true evaluation)
WATCH inflation_spike WHEN cpi_yoy > 0.05

-- Duration persistence: fires after 6 continuous months
WATCH demand_erosion WHEN monthly_bookings < 10000 FOR 6mo

-- Period persistence: fires after 2 consecutive quarterly evaluations
WATCH nvidia_deceleration WHEN nvidia_yoy_growth < 0.20 FOR 2 quarters

-- Longer structural signal: 3 consecutive annual filings
WATCH services_reversal WHEN bea_services_share_declining = true FOR 3 years

Unit semantics:

Period units (quarters, years) register a ScheduledTask of type watcher_timeout so a host system can increment periodsMatched on each evaluation and transition the watcher to triggered when periodsMatched >= FOR value. A period where the condition is false resets the count.

CAL scripts follow a deterministic pipeline that maps directly to the 6D Foraging Methodology:

Every analysis scores across six organizational dimensions:

Cascade chains map how failure (or success) propagates across dimensions: D5 → D1 → D3 → D4 → D6 → D2

npm install @stratiqx/cal-runtime

import { parse, execute } from '@stratiqx/cal-runtime';

const result = parse(`
  FORAGE entities
  WHERE sound > 7
  ACROSS D1, D2, D3
  DEPTH 3
  SURFACE cascade_map

  DRIFT cascade_map
  METHODOLOGY 85
  PERFORMANCE 35

  FETCH cascade_map
  THRESHOLD 1000
  ON EXECUTE CHIRP critical "Cascade detected"
`);

const output = await execute(result.actionPlan, {
  entities: [
    { id: 'svb', sound: 9, space: 9, time: 10, dimensions: { D1: 88, D3: 82, D5: 78 } }
  ]
});

# Run a CAL script
npx cal examples/closed-loop-pipeline.cal

# With data
npx cal script.cal --data entities.json

CAL Script → PEG Parser → Action Plan → Executor → Results
                                            ↓
                              Data Adapters + Alert Adapters

• Parser: PEG grammar (Peggy) — 12 keywords, deterministic parse
• Executor: 6-layer pipeline execution (Sense → Analyze → Measure → Decide → Act → Validate)
• Data Adapters: JSON, memory, composite (pluggable)
• Alert Adapters: Console, file, webhook (pluggable)
• Test Suite: 251 tests across 11 suites

The examples/ directory contains runnable CAL scripts:

• closed-loop-pipeline.cal — Full 5-layer pipeline
• drift-analysis.cal — DRIFT gap measurement
• simple-query.cal — Basic FORAGE query

• Language Spec: cal.cormorantforaging.dev
• API Reference: docs/API.md
• Architecture: docs/ARCHITECTURE.md

npm install          # Install dependencies
npm test             # Run 251 tests
npm run build        # Build for production
npm run typecheck    # Type checking

@misc{shatny2026cal,
  author = {Shatny, Michael},
  title = {CAL Runtime: Cormorant Agentic Language Execution Engine},
  year = {2026},
  publisher = {Zenodo},
  url = {https://github.com/semanticintent/cal-runtime},
  doi = {10.5281/zenodo.18905193},
  note = {ORCID: 0009-0006-2011-3258}
}

Michael Shatny — ORCID: 0009-0006-2011-3258

MIT