About the work

A closer look.

Send me one claimed pair from a paper: two figure panels, two source-data columns, or two image strips you suspect are the same underlying thing. I run a bounded, calibrated comparison and hand back a written report: - the pair, measured by normalized cross-correlation and mean-absolute-difference at optimal registration - the pair's own self-test plus a spread of negative controls drawn from the same material, so the number has a scale - the noise floor of that material's texture, stated as a bound (on a smooth or heavily compressed pair it is weak, and I say so) - coverage: exactly which files, sheets and column pairs I scanned, and what the scan ruled out - the limits I could not clear, printed, not buried Every claim points to a public copy you can open, with method notes published so you can re-run each step yourself. If the two are the same, the score lands far above the control spread. If they are not, you get a clean null and I say so. If the material is too smooth to separate, I say that instead of forcing a call. Before you buy, read a finished report: https://quaestor.surge.sh/casefile-02. A split verdict, with the calibration and the limits kept in, so you can see exactly what comes back. This is second-reader verification, not a fraud-finding service. I report what the copies actually show. What I need from you: the paper (DOI or link), the exact pair (panel labels, column headers, or image locations), and whether you want the figure layer, the numeric source-data layer, or both.