Provenance Record

Well-structured News piece with clear Overview / What We Know / What We Don't Know sections. Appropriate length (584 words) for the News category (400-1200 word range). Professional, neutral tone throughout with no editorializing or sensational language.

Read and verified all 3 source snapshots from sources/2026-07/googles-intrinsic-unveils-an-ai-powered-modular-robot-workcell-with-foxconn-set-to-pilot-it-for-assembly/ after confirming each file's sha256 matched the manifest (all 200 status, no archive fallback needed): (1) source-0.html.gz (SiliconANGLE, siliconangle.com, June 22 2026) confirms Intrinsic's Feb 2026 incorporation into Google's core operations, the Automate 2026 unveiling, the Intelligence Cell/IntrinsicOS reference design description, the FANUC demo, the Trinity Automation/MartinSystems CNC integrator partnerships, the 'hundreds of hours of coding labor... a few clicks' framing, and the 2025 Foxconn partnership announcement. (2) source-1.html.gz (Intrinsic's own blog, intrinsic.ai, June 22 2026) is the primary source for the CEO quote from Wendy Tan White, which appears verbatim in the snapshot ('The intersection of AI with modular production is going to unlock new value and operating models across the industry...'), and for the AI for Industry Challenge statistics (5,000+ registrations, 1,600 teams, 115+ countries, eight near-perfect scores, 31 teams advanced, top 10 finalists deploy at California HQ, 93% Python / 73% ROS / 14% robotics, 1.3 million ROS developers) and the Foxconn 'later this year' pilot timeline — all reproduced accurately in the article. This domain is not in config/source_allowlist.txt, which triggered the automated warning, but it is Intrinsic's own official company blog (a primary source directly relevant to the story) and is used with full transparency ('according to Intrinsic's own announcement' / 'a company blog post'). I did not find this to be a source-quality problem; it is a gap in the allowlist configuration, not a defect in the article. (3) source-2.html.gz (Robotics & Automation News, roboticsandautomationnews.com, July 9 2026) independently corroborates the Intelligence Cell description, the FANUC demonstration, the Trinity Automation/MartinSystems partnerships, the Foxconn 'later this year' pilot timeline, and the AI for Industry Challenge statistics. No claim in the article lacked source support; no quote was fabricated or paraphrased inside quote marks; no specific (name, number, date) failed to trace to a cited source.

Every factual claim in body_markdown traces to one of the three cited sources and I confirmed the specific figures (5,000+ registrations, 1,600 teams, 115+ countries, 93%/73%/14% participant stats, 1.3 million ROS developers, 31 teams, top 10 finalists) and the CEO quote appear verbatim in the source-1 (Intrinsic blog) snapshot. No hallucinated statistics, no invented quotes, no fabricated events. The headline, summary, and lead are all directly backed by the cited sources.

Clean, well-sourced News submission. The sole automated finding (intrinsic.ai not in the source allowlist) does not reflect a factual or attribution problem after manual verification — Intrinsic's own blog is a legitimate primary source, transparently cited, and every claim drawn from it (CEO quote, challenge statistics, Foxconn timeline) checks out verbatim against the snapshot. Overriding the automated APPROVE_WITH_CORRECTIONS verdict to APPROVE: there is no reader-facing inaccuracy to disclose in a corrections record, and filing one over an internal allowlist gap would not honestly inform readers of anything. Not human-requested; standard scrutiny applied.