MacroLens.
A meal estimate, before you eat.

01 / THE STARTING POINT

A real question, at a real lunch.

While eating in a company cafeteria, I wanted to understand roughly how much protein, carbohydrate, and fat was on my plate. There was no food scale, the portions varied, and foods could be mixed or stacked together.

The useful moment was before eating: could I get enough information to decide how much of the meal to have?

02 / THE APPROACH

Make the next decision easier.

The experience starts with three meal photos and turns an estimate into something the person can act on. The product is focused on the decision before a meal, with room to adjust portions.

Built with AI-assisted development and vibe coding. The live web app accepts three meal photos and returns carbohydrate, protein, and fat estimates.

03 / THE TRADEOFFS

Useful estimates need honest limits.

A photo doesn’t reveal everything about a meal. Hidden ingredients, cooking oil, depth, and portion size all make estimation difficult.

• Accuracy and effort. Extra angles or a reference object might improve an estimate, but also add friction.
• Ranges and false precision. A range may communicate uncertainty more honestly than a single exact-looking number.
• Automation and correction. People should be able to correct a food or portion the model misunderstands.

These are design questions I’m exploring, rather than validated product outcomes.

04 / CURRENT PROGRESS

Three photos. Three macro estimates.

Upload three views of the same meal and tap Analyze meal. MacroLens checks the photos in the background and returns estimated carbohydrate, protein, and fat estimates.