The carousel held. The pump head did not.
Five hundred rotations across six slots with no misalignment, and two seal failures on the pump head that sent us to a different supplier. Photographs, the failure log and what it costs us in weeks.
JOURNAL
Including what
did not work.
We promised on the product page to publish the dispensing results when they exist, pass or fail. This is where that promise gets kept, alongside the monthly build update founders already receive by email.
AllBuild updatesTest resultsScienceCompany
First 0.2 ml accuracy run: ±7%, not ±5%.
We said we would publish this whether it passed or not. It did not. Here is the histogram, the two causes we have identified and the date of the next run.
Why we dropped a formula we had already designed.
One ADAPT candidate cleared the evidence review and failed the interaction matrix against two CORE formulas. It is not in the 29 and it is not coming back until the matrix says otherwise.
What the pilot actually taught us.
Five people, ninety days, and the three assumptions we got wrong about how a supplement fits into a morning. Two of them changed the hardware.
443 of 9,730 pairs. Why that number is small.
Every ingredient pair we have not checked is unchecked, not safe. What it takes to document one, and the order we are working through them in.