Prototype · Expo + Supabase
Mesa Prototype
You declare a goal, you check in daily to a private group, and the group witnesses whether you followed through. No progress bars, no scores, and no comparison to anyone else.
What it is
My read is that tracking apps lose the people who need them most, because logging every meal is work and the work is what people quit. So Mesa is built for groups of close friends rather than for a social network, and it has to work for mixed populations, including people for whom calorie and weight fixation is actively harmful.
It's Expo and React Native on Supabase, with row-level security so a group can only ever read its own data. All four tabs are built, along with six-step onboarding, seven goal types, invite deep-links, and consent gating.
The decision
I removed the progress bar on purpose.
Connecting goals to check-ins had five plausible designs, and the two numeric ones, which were goal-type-specific logging and progress-toward-target bars, are exactly what the category expects. I killed both of them on evidence rather than on cost. Showing someone a numeric target at the moment they self-report triggers the violation effect from restraint theory, where missing the number produces abandonment rather than correction. And weight goals plus numeric benchmarking plus social visibility is a documented predictor of disordered eating. Stacking all three of those together would have been the default design.
What shipped is a declaration plus a qualitative status. A check-in says "On track" or "Close," it carries whatever note the person wants to write, and it shows any values they chose to report, but there's no progress bar, no percentage of a target, and nothing comparing you to anyone else. The point I want to land is that the number, where it appears at all, is something you said rather than a score you're being measured against. Weight goals are architecturally isolated from the feed, which means structurally unable to reach it rather than hidden behind a setting, and each goal type has its own rule about whether a target can be displayed at all.
The framing that made the call obvious is that Mesa bets on social witnessing rather than social comparison. Those two look nearly identical in a mockup and behave nothing alike.
What it cost
Numeric progress bars drive engagement, so removing them moves Mesa off MyFitnessPal's high-engagement and high-churn curve toward something closer to the Focusmate and Strava tier. I want to be upfront that this is a bet and not a fact, and that it costs measurable short-term engagement to make it.
The honest problem underneath it is that the accountability research showing 1.5–3× retention is exercise evidence rather than nutrition evidence. Whether it transfers is the keystone assumption the whole model rests on, and it's unproven. I'd rather write that down now than discover it in month six.
Status & next
It runs end to end in Expo. Not yet wired: Apple Sign-In, push notifications, server-side stripping of location data from photos, and the multi-group feed. There's no store submission.
The next real step is testing that keystone assumption with one real group rather than building more surface.