Portion control · at the moment of the scoop
Weigh every scoop.
01 — The leak
more food in the heaviest Chipotle bowl than the lightest. Same bowl. Same price. That gap is your margin.
Wells Fargo, 75 bowls weighed, 2024
02 — Under the hood
How it sees the bowl.
One camera over the line sees the bowl top-down.
The model outlines every ingredient in real time.
Area × the bowl-rim scale → grams, per ingredient.
Compared to the recipe. Off by more than tolerance? Flag it.
keep scrolling ↓
03 — On the pass
Every bowl, checked. Live.
Chicken Rice Bowl
recipe #14next
Add ~0.6 oz chicken to hit target.
scroll to build the bowl ↓
04 — The software
Downtown
lunch service · liveStations
Recipes & ratios
06 — On the go
Every store, in your pocket.
Live in-spec rate, flagged over-scoops, and margin recovered — every location, one screen. A push the moment a line drifts.
Margin recovered · today
$0
Stores
05 — Beyond the line
One check. Everywhere food moves.
Any line
Taco truck to flagship.
Camera glasses
The cook's own view is the scanner.
Dock to shelf
Right item, right box.
07 — North star
Weigh every portion on the line.
The whole smart kitchen.
Every camera in the building.
The north star
A weight-and-vision layer for everywhere food and goods are portioned, packed, and moved.
restaurants · ghost kitchens · grocery delis · co-packers · distribution · pharma fill · manufacturing QA
Our own honest scorecard — warts on
Validated as promising, not proven.
13/16
promising
Dated why-now and proven feasibility carry it. The open questions are a real moat and committed demand — no signed pilots yet. We'd rather show you that than hide it.
See the full scorecard →Hide the full scorecard
crucible run 20260714-063255 · re-validated, 11→13
Where it lost points
- market 1/2 — Reachable 3–9-unit segment is real but shrinking (−2.3% in 2025) and smaller than top-down TAM implies.
- demand 1/2 — Real dated pain (52% of operators cite food cost #1) but evidence is survey-tier — no LOIs or paid pilots yet.
- moat 1/2 — Weak on every power today — no cornered IP vs PreciTaste's 40+ patents, no network effects; counter-positioning is still a hypothesis.
Biggest risk, verbatim
The "underserved sub-10-unit segment" thesis the whole re-check hinges on is now contested, not confirmed: the red-team surfaced at least two additional incumbents already operating in the exact niche PreciTaste supposedly won't chase, and 3 of 3 sampled personas (VC, staff engineer, and the target 3-unit-operator buyer) independently leaned no — each citing that a funded, referenceable competitor plus a near-zero-cost manual fallback (scoops, scales, spot-checks) means Line Ledger has to beat two alternatives before an operator writes a check.