Content Quality: Well-structured News piece (777 words, within the 400-1200 range) with clear Overview/What We Know/What We Don't Know/Analysis sections, appropriate technical depth for a developer-infrastructure story, and a fair 'What We Don't Know' caveat about the unverified social-media wording.
Source Verification: Both snapshots read in full from disk (gunzip + text extraction) rather than re-fetched: sources/2026-08/cloudflare-workers-add-inbound-tcp-sockets-bringing-grpc-support-to-the-serverless-platform/source-0.html.gz (Cloudflare blog, sha256 03bf5da... matches manifest) and source-1.html.gz (InfoQ, sha256 d114e5c... matches manifest). Both fetched at 200 status, no archive_fallback, no suspicious_patterns entries in manifest.json (both null) — the manifest.json was checked and contains no flagged patterns to evaluate. Every specific claim (connect(socket) handler, Spectrum routing, Durable Object/Container forwarding via getTcpPort(8080).connect(), gRPC-to-gRPC-web translation limited to unary/server-streaming, Cap'n Proto/Cap'n Web internal usage, gRPC-web used internally since 2020, UDP roadmap mention) was cross-checked against source-0 and confirmed present and accurately represented, EXCEPT for three quote-marked passages detailed below and in findings.
Factual Accuracy: The article's core claims are accurate and well-supported. However, three quote-marked passages attributed to 'Cloudflare's own announcement' / 'Cloudflare's post' are not Cloudflare's own wording: (1) the Overview's lead quote ('since the platform launched in 2017, a Worker could open outbound sockets to a database or a service, but it could only be a server for HTTP') is verbatim from InfoQ's article, not present anywhere in the Cloudflare blog snapshot; (2) 'web platform APIs like fetch() do not expose' is InfoQ's phrasing (Cloudflare's own text reads 'don't provide this control'); (3) 'which is why gRPC-web exists' is InfoQ's unquoted editorial paraphrase, misattributed to Cloudflare and factually swapped — Cloudflare's post attributes that 'why X exists' reasoning to the WebSocket API, not gRPC-web. In all three cases the underlying substantive claim is true and independently corroborated (by InfoQ directly, and consistent with Cloudflare's own framing of the change), so this is a source-attribution error rather than a fabricated event — the specific ask from the editorial desk to verify the '2017 launch / first time as a non-HTTP server' framing is accurate in substance but was sourced to the wrong outlet in the article text. All other direct quotes attributed to Cloudflare (Spectrum description, Cap'n Proto/Cap'n Web statement, 'and when we ship things...', 'work closely with a smaller set of developers...', private beta signup line, UDP roadmap line, 'since 2020') were verified verbatim against source-0. The 'is now available' quote attributed to Cloudflare's social copy is correctly sourced to InfoQ's reporting of that social copy, not the blog post — no issue there.
Overall Assessment: Substantively strong, accurate, well-sourced News piece on a genuinely new story. The headline, summary, and central factual claims all hold up against both source snapshots. The flaw is narrow and mechanical: three passages quote-mark InfoQ's paraphrase and attribute it to Cloudflare's blog post instead of InfoQ. This is a source-attribution error, not a fabricated fact or an unsourced headline/summary claim, and is fully correctable with a public corrections note. APPROVE_WITH_CORRECTIONS.