Provenance Record

Well-structured News piece (897 words, within the 400-1200 range) using the standard Overview / What We Know / What We Don't Know / Analysis format. Prose is clear, attributions are dense and specific, and the 'What We Don't Know' section proactively flags the Nature paywall and cross-outlet disagreement on price/arm-span figures rather than guessing — good editorial hygiene.

Read all 6 source snapshots from sources/2026-07/uc-san-diego-team-performs-first-teleoperated-humanoid-robot-surgery-in-a-preclinical-trial-publishing-in-nature/ via gunzip, verified against manifest.json sha256/status fields. source-0.html.gz (today.ucsd.edu, 200, primary university press release): fully supports the Overview, both surgery descriptions, non-primate mammal detail, Surgie name/dimensions, the 1,800-lbs comparison, and every direct quote from Yip, Thareja, Liu, and Broderick, all reproduced verbatim in the article. source-1.html.gz (nature.com, 200 but redirected_domain idp.nature.com): only a Cloudflare/login challenge page rendered — no article text — consistent with the article's own hedge that the Nature full text is paywalled; the specific facts attributed to this citation (July 8 publication, paper title) are independently corroborated by source-0 and source-2, so no unsupported claim rests solely on this snapshot. source-2.html.gz (humanoid-surgeon.github.io, official project page linked from the UCSD release): confirms the exact paper title, DOI, journal, publication date, and full 10-author list/affiliations used in the article — exact match. However the article's quoted phrase "porcine in vivo studies" transposes the source's actual wording "in vivo porcine studies" (see Content finding above) — filed as a correction. source-3.html.gz (medicalxpress.com, archive_fallback: true, live URL returned 200 but Archive.org snapshot used per script policy): a syndicated republication of the UCSD release; text matches source-0 near-verbatim, corroborating the Surgie name/dimensions and all quotes. source-4.html.gz (news-medical.net, 200): also a close syndication of the UCSD release; corroborates the 1,800-lbs comparison quote attributed jointly to UCSD Today and News-Medical, and all named quotes. source-5.html.gz (robottoday.com, 200): confirms the G1/Unitree identification and the da Vinci $500,000-to-several-million price comparison attributed to RobotToday, both accurate. However, grepped all 6 snapshots for 'Unitree' and 'G1' — only source-5 contains either term. The article's claim that the G1 identification is 'independently echoed by other outlets covering the study' is not supported by any of the article's own 6 cited sources (see Content finding above) — filed as a second correction. Two sources (humanoid-surgeon.github.io, robottoday.com) are not on config/source_allowlist.txt; both are credible for their role here (the study's own official project page, and a robotics trade publication), so this is treated as an allowlist coverage gap rather than a credibility concern.

Every specific I could check against the six cited snapshots checks out: surgery counts/team compositions, animal-subject detail, Surgie's dimensions, the 1,800-lb conventional-system comparison, the da Vinci price range, the Nature paper's title/DOI/author list/affiliations, and all eight direct quotes reproduce their sources verbatim except the two issues noted above (a transposed two-word quote fragment, and an uncited 'echoed by other outlets' claim). Neither issue touches the headline, summary, or lead paragraph.

Substantively strong, well-sourced News piece with two minor, non-lead-affecting inaccuracies (a transposed-word quote and an uncited corroboration claim) that a single corrections record can honestly cover. Approving with corrections rather than rejecting: neither issue touches the headline, summary, or lead, and both are isolated slips rather than a pattern of fabrication.