# Campaign creation demonstration

Three drafted creative directions for a fictional, unbranded reusable aluminium bottle. No client, launched ad, commercial result or model-performance claim is implied.

Read `brief.md`, then `directions.md`, then `evaluation.md`. The full structured copy and editorial scores are also in `campaigns.json`. `prompt.md` supplies reusable instructions; it was not run as a separate model evaluation.

Reproduce the offline mechanical checks with Python 3.9+:

```sh
python3 validate.py
```

The actual result is recorded in `validation.json`: 3 directions, 5 formats per direction (feed, story, search, email, landing), 15 search copy lengths checked, and 3 editorial score totals checked. These checks validate structure and arithmetic; they cannot validate market appeal or prove claims.

Suggested conceptual image: `/images/mindverve-campaign.webp` in the accompanying website. It is not copied into this bundle and is not evidence of a manufactured product or its specifications. Use it only as a labelled concept visual.

Drafting and selection are editorial work. Only the mechanical validator was executed as code. Nothing was sent to an advertising platform; no credentials, network, dependencies, databases, spend or live audience were used. The script writes only `validation.json` beside itself.

Website-friendly exports: `angles.md` duplicates the direction text; `editorial-review.md` duplicates the evaluation; `ads.csv` contains every copy field by direction and channel; `rubric.csv` contains the subjective scores and their rationale. These are presentation exports of `campaigns.json`, not additional tested concepts.
