Every claim traces back.
The Compact's methodology was synthesized through three structured research waves. Twelve source-extraction documents. Eighteen parallel research agents. One synthesis brief. The audit trail is open.
Sources extracted
12 · across enterprise-internal + external public
Synthesis waves
Wave 1 → Wave 2 → Wave 3
Lives at
research/academy/sources/*.md
I · The methodology
Three waves. One audit trail.
Each wave produced specific artifacts on disk. Wave 1: twelve source-extraction documents. Wave 2: a synthesis brief plus the F-08 Compact spec. Wave 3: the Universal Layer authorship that closed the frontline-asymmetry gap.
i.
Wave 1 · Source Extraction
Twelve parallel research agents — four mining internal enterprise sources (org-wide AI fluency framework, engineering fluency site, knowledge-base ecosystem, messaging channels), eight mining external public sources (Anthropic AI Fluency Index, NIST AI RMF, banks regulators, CUs+community banks, insurance, healthcare, credentials and pedagogy research, competitor academies). Each agent produced a structured source-extraction document.
Output: 12 files · ~37,000 words · 174 cited URLs
ii.
Wave 2 · Synthesis
Six synthesis agents read all twelve Wave-1 outputs and produced spec-shaped deliverables. Research brief at research/21-academy-improvement.md plus F-08 sections by industry-specialist (banks, CUs+community banks, insurance+healthcare), Sentinel adaptation, pilot offer amendment. Spot-checked between waves to verify citations resolve and claims are specific.
Output: Research brief · F-08 v1.0 · pilot offer v1.1
iii.
Wave 3 · Universal Layer
One synthesis agent identified the structural gap in F-08 v1.0: the workforce-wide layer was missing. Wave-3 dove into CX AI Fluency programs, Engineering AI Fluency programs, GTM dashboards, and subsidiary AI toolkits. The Frontline Asymmetry commercial wedge emerged. F-08 v1.1 includes §3.5 Universal Layer; the Five Postures naming was committed.
Output: F-08 v1.1 (Universal Layer · 7K words added)
II · The twelve Wave-1 sources
Click any source to read its findings.
Each source extraction follows the same template — pedagogy, credentialing, simulation, evidence, regulated-industry notes, surprises, citations.
III · The synthesis verdict
The one-paragraph verdict.
From research/21-academy-improvement.md — the Wave-2 synthesis brief that integrates the twelve Wave-1 extractions into the Compact's load-bearing thesis.
The Compact's defensible identity
research/21-academy-improvement.md · §1 verdict
The Compact's most defensible identity is not 'AI fluency' and not a competitor to AIGP — it is an evidence-graded, regulator-defensible apprenticeship for the AI-Native Architect, structured on ISO 42001's ladder, scaffolded to NIST AI RMF, and credentialing artifacts produced against the learner's own employer's real AI systems.
i.
The dominant credentialing logic in the entire surveyed market is empirically obsolete. Per Holton (1996) and Alliger et al. (1997), every existing cert sits at Kirkpatrick L2 with r=0.18 weak correlation to behavior change. The evidence-portfolio thesis is what the pedagogy literature has been waiting for.
ii.
The word "fluency" is contaminated in real time. The Compact must reposition as "AI-Native Architect," not "AI fluency."
iii.
Artifact production suppresses discernment (Anthropic's 9,830-conversation Index found 3.1pp / 5.2pp penalties). A Compact whose premise is "produce evidence" must explicitly teach "discernment under artifact pressure" or it trains learners into worse judgment.
iv.
No competitor academy in the 15+ programs surveyed trains the frontline. The Frontline Asymmetry is the Compact's clearest commercial wedge. The Universal Layer + Five Postures architecture closes the gap.
IV · The discipline
Every claim traces back.
The research was structured to find disagreement first. Every Wave-1 source-extraction document includes a mandatory Surprises section — what the agent expected to find vs. what it actually found, where the previous research framework breaks. Surprises that contradicted the operating thesis were preserved, not filed away.
Three significant ones — the contamination of the word "fluency," the discernment-suppression effect in Anthropic's data, the Frontline Asymmetry — reshaped the Compact's architecture in Wave 2 and Wave 3.
This is the discipline the methodology asks of its learners: name what could be wrong, identify missing context, state which evals you'd write to falsify your current approach. The research process modeled the practice. The practice teaches the research.
Honest limits remain. Some canonical internal URLs returned NO_PERMISSION_OR_DELETED during Wave-1 access; reconstruction proceeded from indexed mirrors. Certain dashboard URLs are auth-gated; content was reconstructed from rollout messages and pre-cohort meeting prep. These are noted in the source files; claims that depend on the reconstruction are flagged as such.
Every claim in F-08 traces to a Wave-1 source. Nothing here was conceived from imagination alone.
The research process modeled the practice; the practice teaches the research. The audit trail is the methodology made visible.
⛬ Find disagreement first; the agreement that survives is the methodology.