kg Food Log: Reveal the molecules in your foods
Food tracking via voice with molecular breakdown, but Gemini approximations undermine trust.

Gemini-powered food inference with micronutrient detail, but missing verified nutrition data accuracy.
Biohackers, health-conscious users interested in nutrient optimization beyond calories
MyFitnessPal · Cronometer · USDA FoodData Central
This is my molecule tracking web app.
It tracks calories like traditional apps, but goes into much higher detail on micronutrients like vitamins and amino acids.
It is powered by Google Gemini, so you can just type a rough list of foods for each meal and it will infer the exact foods you meant. No more scanning barcodes!
I personally built this because the biochemical nature of foods interests me, and it is much faster to use this tool than looking up every food manually.
I hope you will find it useful, and let me know of any feature requests.
Thanks, Enzo
Food tracking via voice with molecular breakdown, but Gemini approximations undermine trust.
Natural language food logging is convenient, but MyFitnessPal and MacroFactor do this now.
Gemini-powered food extraction via natural language, but MyFitnessPal and Cronometer dominate.
Clean, no-nonsense calorie and macro lookup with a searchable, categorized table sourced from USDA FoodData Central — perfect for quick checks while building a meal plan. The UI shows attention to detail (category filter, inline macros, simple search), but the idea is utility-first rather than novel; adding an API, CSV export or portion-size conversion would push this from handy to indispensable. It's a useful complement to MealJar, not a replacement for established calorie databases or trackers.
Voice logging beats database searching, but Cal AI and MyFitnessPal already have voice features.
Photo-to-macros logging and a context-aware workout engine are the project's real selling points — snapping a meal instead of digging through a database and having workouts adapt to your current equipment are tangible time-savers. The landing page is clean and trustworthy, but there’s no evidence on the page about CV accuracy, model safety, or sample workout flows; those are the two things that will make or break this in practice.