The latest articles on DEV Community by Livo Reviewer (@livoreview). https://dev.to/livoreview https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4046806%2F8470d4b4-2f74-409c-ba27-7f72172c063b.png https://dev.to/livoreview en Livo Reviewer Sat, 25 Jul 2026 11:52:45 +0000 https://dev.to/livoreview/same-first-principles-food-prompt-many-llms-model-lab-charts-animal-vs-plant-calorie-shares-1b48 https://dev.to/livoreview/same-first-principles-food-prompt-many-llms-model-lab-charts-animal-vs-plant-calorie-shares-1b48 <p>I built a small static experiment: give several frontier models <strong>the same hard question</strong> about food, force a fixed output shape, and put the results side by side.</p> <p><strong>Live lab:</strong> <a href="https://nutrition.livo.community" rel="noopener noreferrer">https://nutrition.livo.community</a></p> <h2> What it is </h2> <p><strong>Model Lab</strong> is a personal reasoning experiment, not a product and <strong>not medical advice</strong>.</p> <p>Every model gets one core prompt. Sole objective: <strong>metabolic optimal</strong> for a healthy young adult (peak hypothesis, not "people do fine" or guideline aesthetics). Method is locked to <strong>first principles + logic only</strong> — studies and guidelines are out of bounds as authority so the comparison stays fair.</p> <p>Mandatory answer shape:</p> <ol> <li> <code>## Plate description</code> (bullets first)</li> <li> <code>## Substantiate plate choice</code> (short)</li> <li> <code>CALORIE_SHARES</code> footer (<code>animal</code> / <code>plant</code> % of calories + constituents)</li> </ol> <h2> How to read the site </h2> <ul> <li> <strong>Carousel:</strong> each card is one model outcome (or the joint debate fable on the first card). The fable is <em>not</em> labeled as the correct answer — it is a Grok × Claude write-up under the same prompt contract.</li> <li> <strong>Chart:</strong> Artificial Analysis <strong>Intelligence</strong> score vs <strong>animal/plant % of calories</strong>. Animal share is drawn in red so you can scan the spectrum quickly.</li> <li> <strong>Discuss:</strong> optional browser-side OpenRouter key if you want to run more models yourself (key stays in localStorage only).</li> </ul> <h2> Why this is interesting for builders </h2> <p>Same system + user prompt, same format constraints, many vendors. You get a clean view of how models diverge when you strip citation theater and force a concrete plate + calorie split.</p> <p>Static captures live in <code>responses.json</code>; the site is plain HTML/CSS/JS (no framework).</p> <p>Again: <strong>hypothesis comparison only. Not a diet prescription.</strong></p> <p>If you try it, curious which models land nearest each other on the animal/plant axis — and whether that correlates with intelligence scores at all.</p> ai llm webdev