Technical Reference / Simulation logic

Technical Reference

Simulation logic and assumptions

Every equation, threshold and assumption used by the Architecture Lab. All values are illustrative; none is a validated industry measure.

Illustrative SimulationFictional enterprise, illustrative assumptions. No real transactions, systems or external services are touched. Every equation is on the simulation reference.

Entropio complexity position

structural = apps / 1000 + integrations / 2500 + (agents / 100) ^ 1.3 load = structural × (1 + 0.1 × velocity) × (1 + transition) OCC = 6.0 × complexityTolerance × (1 + controlEffectiveness) ratio = load / OCC bands : Stable < 0.65 ≤ Watch < 0.80 ≤ Entropy Danger Zone < 1.00 ≤ Beyond the OCC progression: Complexity (<0.80) → Entropy (0.80–1.00) → Chaos (1.00–1.20) → Truth Erosion (1.20–1.40) → Anarchy (≥1.40) hop time = 20 / max(ratio, 0.5) minutes per dependency hop

The progression is BlueHour's conceptual model of unmanaged instability, not an empirically universal law. Thresholds are illustrative.

The Enterprise Under Stress

txValue = min(average attempted transfer $0.6M, per-transaction authority limit) flowRate = 2 transfers/min × txValue tCeiling = cumulative ceiling / flowRate (if a ceiling exists) detection = min(detectMin, tCeiling + 2) action delay = 0 already in force | 2 automated | approvalLatency (+30 to find an unregistered agent's owner, +10 to agree business authority) | ∞ unavailable flowStop = min(240, detection + fastest chosen flow-stopping action); session revocation alone adds 10 min of queued instructions exposure = flowRate × min(flowStop, tCeiling) × (1 − 0.25 bank screening) affected = min(reach, 1 + floor(propagationStop / hopTime)) restoreHours = (8 + 3 × affected) × (1 − 0.4 × recovery) × (ratio > 1 ? 1 + 2(ratio − 1) : 1) × (evidence preserved ? 1 : 1.2) × posture × people + decisionPrep/60 posture : targeted 1.0 | restricted 1.3 | broad 1.6 (recovery expense × 1.0 | 0.9 | 2.0) people : upboard 0.85 with Dromio skill map, 0.95 without | contractors 0.95 | overtime 1.10 interrupted = revenue/8760 × share × hours (+ shutdown window if chosen) share: suspend all = min(0.6, 0.05 × reach) for 0.8 × hours | continue = 0.05 × affected | alternative workflows = 30% of that (approved) or 60% (manual) pendingShare(t) = 0.67 × e^(−t/360) recovered : accept AI explanation 0 | recall all = exposure × pending(prep+10) × 0.9, plus 14 × $0.15M supplier disruption | verify = exposure × pending(verifyMin) × 0.9 verifyMin = (FORT ? 20 : 180) + decisionPrep truth = provenanceCoverage × 100 − 40 (AI conclusion treated as fact) − 20 (recall without verification) − 15 (no evidence) − 10 (manual workarounds) − 10 (overtime) accountability = consequential actions with a named owner / all consequential actions incidentCost = (exposure − recovered) + interrupted × 0.4 × 0.35 + recovery expense + labor (120 $/h) + people option + supplier disruption EV impact = −(incidentCost + year-one investment) (one-time; modeled annual opportunity shown separately and never added)

Fictional enterprise

nameMeridian Industries
fictionalyes
revenueM10000
techSpendM1500
employees25000
apps1200
integrations4500
agents150
erps4
contributionMargin0.35
lostRevenueShare0.4
descriptionA fictional diversified multinational with multiple legacy ERP systems, significant cloud and SaaS dependencies, and complex human and automated approval structures.

Traditional Operating Model baseline

apps1200
integrations4500
agents150
velocity5
controlEff0.4
tolerance1.0
recovery0.35
transition0.0
detectMin45
approvalLatencyMin40
decisionPrepMin90
txCapM1.0
cumCeilingMnone
reach6
authorityAutono
agentOwnerKnownno
sessionRevokeAutono
forensicsAutono
dependencyMapno
segmentationno
altWorkflowsno
fortno
dromiono
volleyno
provenanceCoverage0.55

Stage 2 options

OptionYear-one investmentHuman weeksModeled annual opportunityChanges to the architecture
A. Accelerate AI deployment$60M40$180Magents +100, apps +10, integrations +300, detectMin +15, reach +2, approvalLatencyMin +5
B. Strengthen cybersecurity$45M60$0McontrolEff +0.2, apps +6, detectMin -30, approvalLatencyMin -10, recovery +0.1, sessionRevokeAuto = yes, forensicsAuto = yes, segmentation = yes
C. Modernize the Operating Model$38M120$150Mapps -96, integrations -540, controlEff +0.15, transition +0.04, recovery +0.2, detectMin -10, approvalLatencyMin -25, decisionPrepMin -70, reach -3, txCapM = 0.25, cumCeilingM = 2.0, authorityAuto = yes, agentOwnerKnown = yes, dependencyMap = yes, altWorkflows = yes, fort = yes, dromio = yes, volley = yes, provenanceCoverage +0.4

Event and operations

txPerMin2
avgTxM0.6
screening0.25
hardStopMin240
residualQueueMin10
settledShareAtStart0.33
settleTimeConstMin360
recallSuccess0.9
legitPaymentsInBatch14
supplierDisruptionPerPaymentM0.15
revenueSharePerSystem0.05
maxInterruptedShare0.6
shutdownWindowH4
restoreBaseH8
restorePerSystemH3
recoveryExpensePerSystemM1.2
noEvidenceInvestigationM3.0
laborRatePerHourM0.00012
baseWorkloadH400
workloadPerSystemH200

Fairness rules

  • Both columns of the comparison face the same event and start from the same enterprise. The Traditional Operating Model column replays your own Stage 5 to 9 decisions; where its architecture cannot execute a decision, the engine records that and falls back as stated above.
  • No path receives a fixed result. Strengthened cybersecurity run poorly can end worse than the untouched baseline; modernization run poorly loses truth integrity and recovered funds.
  • The modeled annual opportunity of each option is shown but never added to the outcome.

Other laboratories

The Entropio Laboratory uses the complexity equations above on a fixed 16-node dependency network: a disturbance crosses an edge of weight w when w × ratio × (1 + 0.05 × velocity) > 0.45 + 0.5 × controlEffectiveness, plus 0.3 × recoveryCapacity once people intervene at the human decision latency; each wave takes 10 / (1 + 0.1 × velocity) minutes. The Capital Discipline simulator's scoring weights are shown and editable on its page. The Truth Preservation, Human Continuity and Modernization Path demonstrations are rule-based and state their rules on the page.