Playbook / How it was made
Makeup
Everything below is read live from the article file and its brief.
Credits
| AI authors | ChatGPT-5.1, Grok 4.7, ChatGPT-5.6 |
|---|---|
| Published | 2025-11-21 |
| Last modified | 2026-09-24 |
| Template | G (evidence tier 3) |
| Charts | None |
Pass log
6 logged passes by 3 models.
| Date | Pass | Model | Note |
|---|---|---|---|
| 2026-09-22 | rewrite | Grok 4.7 | Casual take: the Stowaway survey and the bag test. Labeled as a brand survey. |
| 2026-09-23 | rewrite | Grok 4.7 | Deepened makeup around the bag you already carry, with a one-month box for the rest. Kept the 2015 ownership survey as a brand survey. |
| 2026-09-23 | voice | Grok 4.7 | Removed defensive not-exactly-80/20 lines so the lens opens an opportunity. Kept a limit only where a real figure needed one. |
| 2026-09-23 | review | ChatGPT-5.6 | Fixed audited sentence joins, unsupported or overextended claims, insider jargon, and meta prose. |
| 2026-09-23 | review | ChatGPT-5.6 | ChatGPT-5.6 correction pass: restored missing hero media where applicable, expanded sketch-like articles above 700 words, preserved existing evidence and voice, and ran sentence-by-sentence sense, link, and regression checks. |
| 2026-09-24 | rewrite | Grok 4.7 | Kept the owned-versus-used survey and restored the short kit: base, brows, and one color. Cut the pasted closer. |
Research brief
The brief was written before the outline. It is the research the article stands on. Open the full brief.
Mechanical checks
These are run on the published HTML when you load this page. They are the checks a machine can do. The judgement gates - the deletion test, audience fit, whether the concentration claim is real - are in the publish checklist and are not automated.
| Length | 1,002 words |
|---|---|
| Opens with prose | Yes |
| Outbound sources | 2 links to www.prnewswire.com |
| Link hygiene | All 3 links open in a new tab with rel set |
| Illustrative labels | 0 scenarios labeled as illustrative |
| Em dashes | None |
| AI sludge words | None |
A failed row means the article is due for another pass.