Provenance Record

Well-structured News piece (900 words, within the 400-1200 range) following the standard Overview / What We Know / What We Don't Know / Analysis format. Clear, technically precise explanation of a router-not-a-pooler product, with the workload-level-concurrency distinction reinforced multiple times throughout the body.

All 6 sources fetched successfully (200 status, no archive fallback) and read in full from the gzipped snapshots at sources/2026-09/nvidia-releases-personal-ai-router-an-open-source-tool-that-pools-local-ai-inference-across-a-home-network/: source-0.html.gz (NVIDIA developer blog, primary source), source-1.html.gz (MarkTechPost coverage), source-2.html.gz (GitHub README.md), source-3.html.gz (docs/getting-started.mdx), source-4.html.gz (docs/overview.mdx), source-5.html.gz (GitHub releases page). Every direct quote in the article was checked verbatim against its snapshot, including the bolded 'not' in the README quote (confirmed present as <strong>not</strong> in the underlying rendered markdown, not an invented emphasis). The 18-minutes vs 8-minutes-48-seconds benchmark figures, the Qwen 3.6 35B A3B model name, the six-digit PIN pairing flow, the mDNS/MTLS security details, the supported-hardware list, the OS/architecture support matrix, and the v0.1.1 (Aug 28, 2026) / v0.1.0 (Aug 26, 2026) release dates all match their cited sources exactly. No hallucinated quotes, no misattributions found.

No unsourced or fabricated specifics detected. Every number, date, model name, and direct quote traces to one of the six cited sources. The docs/overview.mdx source, while listed in article.sources, is not independently hyperlinked inline in the body text (it corroborates rather than being directly quoted) -- this is a minor sourcing-hygiene note, not a factual problem, since its content (PIN pairing, node/cluster/engine vocabulary) is consistent with and reinforces claims already sourced to the other five links.

Clean, thoroughly sourced submission. All six sources were fetched and read directly, every claim and quote verified against the underlying snapshot text, and no fabrications, misattributions, or duplicate coverage found. The one headline-wording concern is adequately defused by the article's own summary and body content, so it does not warrant APPROVE_WITH_CORRECTIONS. Approved as-is.