Methodology
How we read. Source tiers, two-pass investigation, statistical-claim qualifiers, corrections.
What this lens looks for
The publication’s own methodology, examined in public. Articles through this lens are about how a verdict was reached, not just what the verdict was.
When we apply it
- A reader asks how a published verdict was sourced
- A change in research, sourcing, or correction protocol warrants disclosure
- A common reasoning trap (selection bias, denominator drift, citation laundering) shows up across multiple recent announcements
The standing rules
Source tiers. Every claim is sourced to a tier:
- T1: peer-reviewed work, regulatory filings, earnings call transcripts, official vendor releases, AIRS canonical data
- T2: reputable trade press, vendor blogs with disclosed methodology, named LinkedIn posts
- T3: anonymous social, Reddit, X. Excluded as standalone evidence; corroboration only.
Every analytical claim is anchored at T1. T2 and T3 sources are leads, walked back to a T1 primary before they enter the queue.
Two-pass investigation. No story enters the queue on a single T2 or T3 signal. Pass 1 surfaces the lead. Pass 2 walks it to a primary source. The article either ships with the primary in hand or does not ship.
Statistical-claim qualifiers. Any draft that prints β, p, Cohen’s d, R², CFI, or RMSEA carries an in-sample qualifier nearby. A coefficient without the sample is theatre.
Corrections. When the record changes, the corrections log carries the change with date, source, and what was wrong.
Articles through this lens
-
Two Weeks in Review (W21-W22): The Baseline for How We Read AI Adoption
This is the first two-weeks-in-review for The Hype Check and the baseline for our method. We score news against customer-facing levers, then separate signal from announcement noise.
-
Welcome to The Hype Check
Phase 1 scaffold. The publication is live; the first real article ships in Phase 2 (June 2026).