For a nutrition product, a recipe is not just a set of macros.
It has to look like something someone would actually want to make. It has to feel specific enough to match the client's week, but varied enough that the product does not collapse into the same bowl over and over.
That is the focus of this pilot update.
What changed
We have been improving the generated recipe experience across two related surfaces:
- The quality and variety of recipe ideas.
- The food imagery that helps those ideas feel concrete.
The latest visual set pulls from backend-generated recipe library images, not mock screenshots. It includes a wider range of meal types and compositions: eggs, yogurt, cod, salmon, shrimp, wraps, chickpeas, curry, tempeh, and chicken.
That variety matters because the image is often the first test of whether a meal feels plausible.
Why it matters
Clients are more likely to engage with recipes that feel like real meals, not clinical outputs. Dietitians also need confidence that the system can produce options that are varied, culturally aware, and grounded in the client's constraints.
This work is not finished. We are still watching for repetition, ingredient bias, weak photography, and recipes that look better than they would cook.
But the direction is clear: more personal, more appetizing, and more useful in the actual weekly flow.