Research evidence

The plan, backed by published work

Every big promise in this project, checked against peer-reviewed research. We searched the literature, then confirmed each paper is real with a working link before listing it here. The evidence supports the plan, with two honest corrections at the end.

18 papers 18 verified real 0 fabricated

In short

What the literature lets us say

The two corrections

Where the research makes us more careful

Accuracy is conditional

Every cm-level result holds only for a close, centered, square-on, marker-anchored shot. Off-angle or at distance it degrades fast. So we quote a documented working band, never a precision spec, and we never quote the Quest lab tracking figure as our field accuracy.

Thai is unproven

The latency research covers English and Japanese but not Thai, and shows vendor latency numbers are easy to misreport. So we set a caption-delay budget and test Thai end to end ourselves before we promise anything.

How to read this: each card shows the paper, its finding, the key number, and the line "what it lets us say" so the claim can be traced to a source. See the plan for how these shape the build, and the preview for what it looks like.