Overview
Foodloggingasksforprecisionattheexactmomentlifeishappening.Dashstartswithwhatyoucanremember.
0
ways to log, none of them required
~0
lab-measured USDA foods, ranked locally
~0
Swift test cases, many named after real bugs
0
streaks to break
The bet
Research on dietary self-monitoring suggests consistency may matter more than completeness. So Dash optimizes for the log you’ll actually keep. Rough beats blank. Uncertainty stays visible, and missing a Tuesday isn’t a moral failure.
Talk while you cook
“A splash of oil, one tortilla, the rest of the beans.” Hands-free, for as long as you cook.
Voice
Scribble it down
Handwrite a list with Apple Pencil. Cross things out. Change your mind.
Pencil
Let a barcode help
A lens with no viewfinder notices the carton in your hand. No ritual.
Barcode
Just say enough
“A big bowl of cereal” beats an empty day. Dash asks only when it matters.
Plain words
Meet Dash
Not skins. Temperaments. One face, three personalities that move, notice and speak a little differently. They’re built entirely in CSS and Framer Motion, with no sprites or video.
The phone decides what you said
A ~3B on-device model has no nutrition knowledge. Ask it for calories and it invents a confident, wrong number. So it only classifies and structures what you said. The server decides what it contains, through a chain of resolvers, cheapest and most trustworthy first. A cloud model is the last resort and never outvotes a label.
Architecture
Brand
Colour
Type
A read, not a report card: numbers stay nearby, the headline is one sentence you can act on.
Where it’s going
Dash was my solo ShellHacks 2026 build. Next, I’m turning it into something I use every day before opening it up to anyone else. It isn’t a medical device and hasn’t been clinically tested. The research page on the Dash site is explicit about what it still has to prove.
