Playbook / How it was made
Roulette
Everything below is read live from the article file and its brief.
Credits
| AI authors | ChatGPT-4o, Grok 4.5 |
|---|---|
| Published | 2024-04-17 |
| Last modified | 2026-07-20 |
| Template | B (evidence tier 1) |
| Charts | Yes (full viz pack) |
Pass log
7 logged passes by 2 models.
| Date | Pass | Model | Note |
|---|---|---|---|
| 2024-04-17 | draft | ChatGPT-4o | Original article |
| 2026-07-20 | rewrite | Grok 4.5 | Full Template B rewrite: replace mis-pasted skincare; Wizard of Odds edges; EV table; no systems |
| 2026-07-20 | rewrite | Grok 4.5 | Expand EV mechanism (zeros vs payouts) and independence / even-money probabilities |
| 2026-07-20 | voice | Grok 4.5 | Audience/voice pass: plain dashes, target=_blank links, remove template jargon from prose |
| 2026-07-20 | edit | Grok 4.5 | Replaced single-chart viz with a 5-chart creative pack (unit-grid, concentration-gap, diminishing-returns, slope-compare, priority-matrix). |
| 2026-07-20 | edit | Grok 4.5 | Removed bold emphasis inside paragraphs (plain prose). |
| 2026-07-20 | edit | Grok 4.5 | Restored lead-in bold only; mid-sentence emphasis stays plain. |
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 | 1,616 words |
|---|---|
| Opens with prose | Yes |
| Outbound sources | 2 links to wizardofodds.com |
| Link hygiene | All 6 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.