Treadstone 71’s tight upgrade path to turn STEMPLES Plus into a 2026‑ready analytic standard by embedding human, AI, and cultural layers inside every cell—so each matrix square becomes a living micro‑lab, not a static box.
1) Define the three data layers per cell
- Human layer (H‑Layer): SME notes, HUMINT/PEOPINT cues, red‑team hypotheses, confidence + dissent logs.
- AI layer (A‑Layer): model‑generated patterns, anomaly flags, entity/event linking, source triage scores, provenance trails.
- Cultural layer (C‑Layer): Hofstede scores, linguistic frames, semiotics, identity fusion indicators, holidays/rituals, power distance signals.
Each STEMPLES² element (Socio, Tech, Eco, Military, Political, Legal, Env/Edu, Security + “Plus”) now stores H/A/C subfields with time stamps and versioning.
STEMPLES Plus-framework to assess cyber capabilities
2) Minimal schema (per cell)
- Facts: atomic, source‑bound statements.
- Signals: weak/strong; trend slope; recency.
- Narratives: dominant frames + counter‑frames.
- Risks/Opportunities: likelihood x impact; confidence; alternatives.
- Decisions/Tasks: who/what/when; blocking assumptions.
- Citations: human notes, model IDs, cultural refs.
3) Workflow (daily)
- Ingest: scrape/collect → auto‑label to STEMPLES² cells.
- AI pre‑pass: deduplicate, cluster, score credibility, extract frames.
- Human adjudication: analysts edit, add dissent, set confidence.
- Cultural overlay: apply norms, idioms, symbolism, calendar effects.
- Publish: BLUF + cell‑level deltas; push to dashboards and briefs.
4) Training rollout (early adoption = advantage)
- Week 1 — Cellcraft Basics: how to write atomic facts, add dissent, and log provenance inside a cell.
- Week 2 — AI as Junior Analyst: prompt patterns, red‑team prompts, error catalog, model comparison.
- Week 3 — Cultural Intelligence: fast Hofstede/ASCOPE withe the Cultural Nexus Framework overlays, semiotic checklists, identity fusion cues.
- Week 4 — Decision Products: BLUFs, alternative analysis, branch‑and‑bound outlooks; live red‑cell injects.
Deliverables: a populated master matrix, a prompt kit, cultural overlay checklist, and five exemplar products (spot, warning, posture, deception readout, foresight note).
5) Governance & quality
- Rubrics: clarity, sourcing, cultural validity, alternative coverage.
- Provenance ledger: human edits and AI outputs logged per cell.
- Red‑cell cadence: weekly narrative stress tests; deception drills.
- Bias guards: banned‑phrases list, assumption register, mirror‑image checks.
6) KPIs (prove it works)
- Warning lead‑time gained (days).
- Decision latency reduced (%).
- Source efficiency (useful:total).
- Cultural miss rate (quarterly).
- Dissent adoption (accepted vs. ignored).
- Forecast Brier scores by cell.
7) Tech enablers (tool‑agnostic)
- Data: JSON schema for H/A/C fields per cell.
- Pipelines: ETL → embed → cluster → route to cell.
- Models: one general model + small task‑tuned heads (credibility, narrative frame, culture tags).
- UI: cell detail pane with H/A/C tabs; delta view; dissent stripe; export to BLUF.
8) Templates you can use today
- Cell Card (one page): Fact stack → Signals → Cultural read → Dissent → Outlook (3 branches).
- Narrative Readout: Top frames, amplifiers, audiences, counters; deception probability.
- Cultural Flash: do/don’t phrases, taboo topics, authority triggers, timing cues.
9) 90‑day roadmap
- Days 0–30: schema, ingestion, first five high‑value cells live.
- Days 31–60: cultural overlay v1, KPI baseline, red‑cell drills.
- Days 61–90: automate deltas, publish weekly outlooks, externalize two case studies.
10) Why it sets the 2026 standard
Every judgment ties to sources, model outputs, and cultural context inside the same square. Leaders get faster, culturally precise warnings with auditable reasoning and structured alternatives—repeatable, teachable, and saleable.
The Cyber Intel Training Center