Playbook / How it was made
Insurance
Everything below is read live from the article file and its brief.
Credits
| AI authors | ChatGPT-3.5, ChatGPT-5.5, Grok 4.7, ChatGPT-5.6 |
|---|---|
| Published | 2022-12-28 |
| Last modified | 2026-09-23 |
| Template | B (evidence tier 1) |
| Charts | None |
Pass log
6 logged passes by 4 models.
| Date | Pass | Model | Note |
|---|---|---|---|
| 2022-12-28 | draft | ChatGPT-3.5 | Original article |
| 2026-07-20 | rewrite | ChatGPT-5.5 | Full Template B rewrite: Swiss Re/CAIF/NAIC RBC anchors, cat/fraud/concentration/coverage guardrails, readiness card, title=Insurance, no scorecard jargon. |
| 2026-09-22 | publish check | Grok 4.7 | Added the required 20% and 80% attention sentence. Left the Swiss Re and fraud figures in place. |
| 2026-09-23 | voice | Grok 4.7 | Replaced the bolted-on 20/80 attention sentence with a continuous opening. Left the sourced body in place. |
| 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. |
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 | 1,701 words |
|---|---|
| Opens with prose | Yes |
| Outbound sources | 8 links to content.naic.org, insurancefraud.org, www.swissre.com |
| Link hygiene | All 14 links open in a new tab with rel set |
| Illustrative labels | 1 scenario labeled as illustrative |
| Em dashes | None |
| AI sludge words | None |
A failed row means the article is due for another pass.