---
type: "synthesis"
sources: ["attention"]
tags: ["epistemics", "evidence", "verification"]
related: ["cross-agentic-ai-enabler-and-destroyer", "cross-relationship-metrics-over-vanity"]
id: "cross-proprietary-evidence-epistemics"
sourceVaultSlug: "hbr-seg-attention"
originDay: 4
articleStem: "hbr-seg-attention"
sourceUrl: "(unified vault: 8 sources)"
sourceTitle: "HBR — Demand Ⅰ-B · Attention economy / GTM shift & habit-moat"
---
A methodological warning that applies to the *entire* corpus: **the behavioral principles are well-grounded, but many headline statistics are proprietary, single-source, or unverified.** A downstream agent should answer with a two-layer verdict — endorse the mechanism, flag the magnitude.

Representative examples:
- **A007:** the 'six-week' capability cycle ([[claim-capability-depreciation]]), M365 Copilot's 3.3% ([[claim-invoked-ai-ignored]]), and especially the [[claim-instant-checkout-failure|Instant Checkout case]] — externally *unverified*.
- **A069:** vendor-reported Cyber Week/Adobe/Anthropic figures behind [[claim-tipping-point-2025]]; the strong verbs ('destroy', 'race to the bottom') are extrapolations — mechanism solid, magnitude forward-looking.
- **A070:** the 70/18/37 churn stats are proprietary; only the 9–15% attention lift comes from the study ([[claim-timing-content-equivalence]]).
- **A065:** 'half fake / third misrepresent' likely proprietary ([[claim-trust-eroding-despite-growth]]).
- **A090:** 15–20% productivity 'runs hot' vs. McKinsey's central 3–15% ([[evidence-productivity-benchmarks]], [[claim-productivity-boost]]); the 50k/1M agentic case is one unverified engagement ([[evidence-agentic-scale-caveats]]).
- **A068:** the '30-fold' production increase is uncorroborated strategic rhetoric ([[question-supply-chain-limits]]).

The pattern is remarkably consistent: **thought-leadership pieces pairing a robust directional argument with illustrative, hard-to-verify numbers.** The A090 vault models the correct posture best — separate the claim note from an [[evidence-productivity-benchmarks|evidence]] note. Apply that discipline everywhere: cite the principle confidently, attribute the figure, and name it author-provided when precision matters.