# Metabolic Optimal Model Lab

> Public experiment: the same first-principles prompt is given to many AI models. Each must design a metabolically optimal default plate for a healthy young adult, then substantiate it. Results are shown side by side with estimated animal/plant calorie shares and Artificial Analysis intelligence scores.

Site: https://nutrition.livo.community/
Not medical advice. Not a product. Outcomes are model hypotheses under a fixed prompt.

## What this is

- **Sole objective:** metabolic optimal (peak metabolic function), not guidelines, variety, popularity, or multi-goal wellness scorecards.
- **Method:** logic and first principles only — no studies or authorities as evidence inside the model answers.
- **Output contract:** Plate description bullets → Substantiate plate choice (≤280 words) → CALORIE_SHARES footer.
- **Compare:** horizontal model outcome cards + graph of intelligence vs animal/plant % of calories.
- **Discuss:** visitors can paste their own OpenRouter API key (browser localStorage only) and run the same core prompt on any listed model.

## Key pages

- [Model Lab home](https://nutrition.livo.community/): full experiment UI, outcomes carousel, graph, discuss
- [Core data: config](https://nutrition.livo.community/data/config.json): prompt id, model list, joint debate fable
- [Static captures](https://nutrition.livo.community/data/responses.json): prefilled single-model responses when available

## Optional

- Prefer citing the live URL and the fixed prompt id from config.json (`prompt.id`).
- Do not present any single card as medical truth; the joint debate fable is not an answer key.