Content Quality: News article of 997 words (News range 400-1200) with a clear structure: lead with the three dated facts, what the pages say about versions, what a caller sees, tooling and a Python client check, the Azure/AWS comparison, and what the sources leave open. Facts are attributed to the specific page each time (blog, guide, release notes, GitHub release, InfoQ). The Herald's own package comparison is labelled as such and its scope limits are stated (Go, Java, Node.js, C#, Ruby and PHP not checked; AWS services other than EC2 not examined).
Source Verification: Read all 8 snapshots from sources/2026-11/<slug>/ (source-0..7.html.gz); the sha256 of each decompressed snapshot matches the manifest. All 8 returned 200 with no archive fallback; suspicious_patterns is null for all 8 (no prompt-injection flag to judge). I also fetched all 7 distinct live URLs with curl and a browser User-Agent on 2026-11-14 and the key strings were present and identical to the snapshots (guide and release notes both still 'Last updated 2026-11-10 UTC'). HEADLINE: (a) 'Compute Engine API gets date-based versions' - blog (source-1): 'API multi-versioning for Compute Engine, now in general availability' and 'available today across all Google Cloud regions, with 2026-09-01 as the initial stable version and 2026-10-01-preview available for preview testing'; guide FAQ: 'IBV API is enabled by default with Compute Engine API', so usable by all customers without opt-in; /v1 stays the default. (b) 'release notes say GA on September 29' - source-0 entry dated 'September 29, 2026' (above the September 28 entry, below the September 30 one): 'Interface-based versioning (IBV) for the Compute Engine API is now generally available.' and 'Generally available: Compute Engine API version 2026-09-01 is now generally available.'; the entry also lists 2026-10-01-preview as 'now available in Preview'. The GA statement is for IBV and for 2026-09-01, as the article words it; the article does not call the preview GA. (c) 'blog is dated November 5' - source-1 byline shows 'November 5, 2026' and the page metadata has datePublished 2026-11-05. (d) 'Python client 1.55.0 removed fields' - supported by a cited source, not only by the Herald's check: GitHub release page (source-5, time attribute 2026-10-01T16:02:52Z) lists under Bug Fixes 'all AggregatedList RPCs return_partial_success fields removed' and 'remove several enums and fields that no longer exist in the API' (no names), and under Features 'update Compute v1 API to version 2026-09-01'. The 199-vs-0 count is not in the headline or summary; the body labels it 'The Herald compared ... itself'. SUPPORT-TERMS DISAGREEMENT: blog says 'They have no end-of-support date today.' (stable versions) and 'remains fully supported' (/v1); guide FAQ says 'Stable versions are maintained in perpetuity under the standard Google Cloud deprecation policy.' Both quotes verbatim on the exact pages; the guide FAQ has no link at 'deprecation policy' (checked in raw HTML), both pages say 'Last updated'/dated 2026-11-10 UTC (guide) and November 5 (blog); I fetched Google's cloud.google.com/terms/deprecation and the 'Creating API requests and handling responses' page (linked from the blog) and neither states a support term for date-based versions (the deprecation policy page does not mention IBV or date-based versions at all; the requests page mentions IBV only in passing, with no support or deprecation wording), so no third page reconciles them. The two wordings are compatible readings (no end date announced today vs maintained under the general deprecation policy); the article only says the pages 'word some points differently' and that it found no page reconciling the wording; it does not assert a contradiction. The summary says the same ('word support terms differently'), which is accurate. UNMENTIONED-IN-BLOG ITEMS: release notes 'The returnPartialSuccess query parameter is no longer available in aggregatedList and list methods.' verbatim; blog text has no 'removed/no longer' wording and describes only the default change ('defaults returnPartialSuccess to false' in /v1, 'partial success is enabled by default' in date-based versions), so 'The blog does not mention that the parameter is removed' is correct. Preview quota change verbatim: 'the quotas field is no longer available in responses from the projects.get, regions.get, and regions.list methods' (three methods) and 'use the Cloud Quotas API'. gcloud: source-4 under 588.0.0 (2026-10-06), Compute Engine: 'Froze development of the alpha and beta Compute Engine release tracks. Existing commands in these release tracks remain supported and functional.' and 'new Compute Engine preview features will launch in gcloud preview compute going forward.' - verbatim; the article's quote strings match. Guide (source-2) quotes verified: 'IBV supersedes CBV', 'continue to work as before', 'you won't be able to access new features available in IBV API', 'are planned with quarterly releases', 'can be released at any time', 'We may change, refine, or remove preview features while promoting them to a stable release.'; guide states no preview lifespan (correct); release-notes header 'Latest API version: v1' present. Blog quotes verified: 'Stable versions don't receive breaking changes after they're published.', 'a predictable one-year lifespan', 'continues to serve as the default if no date-based version is specified', 'tighten overly broad legacy IAM permission checks and introduce safer configuration defaults', 'Upgrade your client library dependency as usual–no import path rewrites required.'; X-Goog-Api-Version / $apiVersion on the /compute/v1/ path, gcloud preview compute and Terraform google vs google-beta statements all match. The article's claim that the blog and InfoQ do not mention September 29 is correct (no 'September' in either). PYTHON PACKAGE RE-RUN: downloaded the wheels google_cloud_compute-1.54.0 (uploaded 2026-09-17) and 1.55.0 (uploaded 2026-10-01) from PyPI (pypi.org is neither cited nor on the allowlist; it was used only for the Herald's own check and the article does not cite it): google/cloud/compute_v1/types/compute.py has 199 classes containing return_partial_success in 1.54.0 (all 199 are *Request classes; 597 text occurrences) and 0 in 1.55.0; gapic_metadata.json has 0 apiVersion entries in 1.54.0 and 129 entries all equal to '2026-09-01' in 1.55.0; the compute_v1 package name is unchanged and no class was removed (16 classes added). The article's figures reproduce. INDEPENDENT OUTLET: InfoQ (source-3), 'Google Cloud Moves the Compute Engine API to Immutable Date-Based Versions', Steef-Jan Wiggers, Nov 14, 2026 (datePublished 2026-11-14T07:27:00+0000); every InfoQ quote in the article is verbatim: 'has no end-of-life date', 'no complete list of behavioral differences between /v1 and 2026-09-01 beyond the examples above', 'the last of the three largest cloud providers to let callers pin an API version on a request', the Azure/AWS sentence, and 'no path by which a preview version becomes stable'. InfoQ largely retells the blog, so independent sourcing is one outlet (InfoQ) plus the Microsoft and AWS documentation used only for the comparison; the rest are Google's own pages. The Stripe comparison is not repeated in the article (Stripe docs are not allowlisted and not cited). AZURE/AWS (Rule 14): source-6 Microsoft Learn: 'Both require an api-version query-string parameter.' (said of Azure Resource Manager provider APIs and the classic deployment model) with example requests api-version=2014-04-01-preview and api-version=2016-02-01 - verbatim. source-7 AWS EC2 Query reference: 'Version: The API version to use. For the Amazon EC2 API, the version is 2016-11-15.' - verbatim, listed as a component of a request and shown as Version=2016-11-15 in the examples. InfoQ's 'does not expose version selection on a request' vs AWS's Version parameter: the statements are not strictly contradictory (a required Version parameter with a single value is not obviously 'selection'), and the article neither asserts a contradiction nor reconciles them: it reports both, says the cited pages do not relate them, and states that AWS services other than EC2 were not examined. This satisfies the disagreement rule. CLAIM-ACCURACY SCAN: ran the quantifier grep on title, summary and body; every hit tested against all snapshots (both, each, all, only, last, any, none): 'Both require' (Microsoft, verbatim), 'last of the three largest' (InfoQ attribution), 'each client-library release' (blog and guide), no unattributed absence claim beyond the explicitly bounded 'the Herald found no cited page that ...' statements, which I confirmed hold for the 8 cited pages (and for the two extra Google pages I checked). Dates: calendar dates only as printed by each page; the article states it uses the dates as the pages print them; the GitHub timestamp is quoted as the page's time attribute in UTC; no time-zone-sensitive comparison is made. Derived figures: 199 and 0 recomputed; 129/0 apiVersion entries not stated; 'three methods' for the quotas field recounted. Anthropic disclosure: not required - Anthropic and Anthropic products are not in the story, headline, summary or lead, and no Anthropic credit line appears. Allowlist: docs.cloud.google.com, cloud.google.com, github.com, learn.microsoft.com, docs.aws.amazon.com and infoq.com are on config/source_allowlist.txt; no additions needed (cloud.google.com/sdk/docs/release-notes redirected to docs.cloud.google.com, both allowlisted). Every body URL is in article.sources and vice versa.
Factual Accuracy: No errors found. Headline, summary and lead each trace to a cited page: the September 29 GA entry (release notes), the November 5 blog date (blog byline and metadata), the 1.55.0 field removals (GitHub release notes), and the differing support wording (blog vs guide FAQ). Every quoted string is verbatim on the page it is attributed to. The Herald's own package comparison reproduced exactly (199 request classes with return_partial_success in 1.54.0, none in 1.55.0; apiVersion 2026-09-01 in 1.55.0's gapic_metadata.json only).
Overall Assessment: Accurate, well-attributed News article whose headline dates, quotes, the Herald's own package comparison and the Rule 14 cross-provider statements all verify against the raw sources; no corrections needed. APPROVE. Featured candidacy: 1 - solid, accurate coverage of a cloud API versioning change for platform and infrastructure developers