Preview, Beta, GA: What Vendors Said vs What Coverage Said

A new ledger comparing vendor availability language to media coverage across 36 dated AI announcements from 17–22 August 2026 found that 21 of 26 matched, with all five mismatches showing coverage upgrading or adding status words beyond what vendors stated. The analysis, which read vendor and coverage wording verbatim, found that only GitHub's Pre-release License Terms provide a contractual definition, leaving 30 of 36 rows with undefined launch-stage terms.

Preview, beta and general availability are the three words an AI vendor uses to tell you how much to trust a feature, and they are also the three words coverage most often changes on the way to a headline. This ledger puts the two side by side for every dated vendor announcement we could find between 17 and 22 August 2026: the vendor’s status word, read verbatim from its own page, next to the word the coverage used, read verbatim from the outlet. The question matters because a status word is often the only contractual signal a buyer gets. “Public preview” at GitHub means the feature is provided as-is and may be withdrawn without notice. “Out of beta” at Anthropic means a request header is no longer required. “GA” in a headline about Cursor means, as it turns out, nothing Cursor said. Thirty-six rows, one claim per cell, with every source linked so the comparison can be re-run. The dataset holds labs and developer platforms, coding tools, distribution layers, martech and ads platforms, and three control rows a regulator and two securities filings where the vendor’s wording is definitive by construction. Rows where vendor and coverage agree stay in; so do the ten rows where no independent coverage could be located at all. - 0136 rows, read from the published pages.The vendor word was read, not inferred, for all 36 rows; three cells carry a caveat the table states in place, two of them because the wording was carried from the batch’s own earlier fetch rather than re-pulled here. The coverage word was read for the 26 rows where an independent outlet was located, two of which rows 24 and 27 record a sampled characterisation rather than one pinned URL. Ten rows have no coverage side at all. - 0221 of 26 agree; the 5 mismatches all point the same way.Three rows show coverage upgrading a hedged or unlabelled vendor word to “GA”, “unveils”, “rolls out” ; two show a coverage word the vendor’s own copy does not carry “warns” for the FTC or does not confirm “beta” for HubSpot . Zero rows show coverage hedging below the vendor. - 03Ten rows had no independent coverage word at all.A Cursor changelog entry, a Codex beta, a Copilot settings change, a preview-tagged CLI release, an eve default change, a Gateway label, a floating alias, a Shopify deprecation and two Vercel changelog entries drew no sampled outlet using its own status word. Recorded as an absence, not filled. - 04The three seed rows hold on re-fetch; one is no longer a mismatch.Cursor’s changelog uses neither “GA” nor “general availability” — but no outlet headline carrying the firmer word could be pinned either, so the row is recorded as coverage not located rather than scored as an upgrade. OpenAI says “previewing” and “currently being tested with early customers” where headlines say “Unveils”. HubSpot’s KB page carries no “beta” in its raw HTML body; a rendered token near the title could not be resolved, so the label is attributed to trackers, not to HubSpot. - 05Almost nobody defines the words they use.Only GitHub’s site-wide Pre-release License Terms give a row a quotable contractual meaning. Anthropic, OpenAI, Cursor, DeepSeek, Z.ai, Meta, xAI, Vercel and HubSpot publish no launch-stage glossary for the items in this table, so 30 of 36 contractual-meaning cells open with “not defined”, four read “not applicable”, one records no vendor glossary at all, and one row 15 rests directly on a published definition. GitHub’s terms are the only published definition cited anywhere in that column — by rows 10, 11 and 15. 01 — MethodologyTwo words per row, both read , not remembered. The unit of this ledger is a dated announcement, not a product. An announcement qualifies if a vendor published it on its own surface blog, changelog, release notes, GitHub Releases, an SEC filing or a regulator’s press release with a date inside 17–22 August 2026, and if it carries, or conspicuously lacks, an availability word. The population is the set of such items named in our own batch anchor document for the week, which was itself built by primary source sweeps; this table did not go looking for announcements beyond that list. What was collected. Every dated 17–22 August 2026 AI-vendor or platform availability announcement in our week’s anchor list across labs and developer platforms, paid media, ecommerce and martech, and agentic and developer tooling, plus three business and policy items with a stated status word the FTC, NVIDIA and Marvell . Total: 36 rows from 18 vendors, regulators and filers. How each row was built. 1 The vendor’s own page was fetched directly, by a plain fetch or, where the direct fetch returned a 403 or an empty body OpenAI, the FTC , by a reader-mode fallback that strips page chrome but keeps body text; the availability wording was copied verbatim, or recorded as “no status word”. 2 One or two pieces of independent coverage were located by targeted web search and fetched the same way; their status word was recorded verbatim with its own URL and date. 3 The two words were classified as same coverage matches or fairly paraphrases the vendor , upgraded coverage used a stronger or more definite word , not in vendor copy coverage used a term that the vendor’s own published copy does not carry, or does not confirm, and that is not a defensible synonym , downgraded coverage hedged below the vendor , or none located no independent coverage word found . Data as of 22 August 2026. Every page was fetched directly and the wording recorded as it read at that fetch. Live vendor pages can be edited after publication; a later re-fetch that disagrees with a cell is a dated discrepancy, which is itself useful signal. Read versus inferred. Every vendor word in the table was read from a published vendor record rather than recalled or summarised, and 34 of the 36 were read from a page fetched for this table. Three cells carry a caveat, and say so in the table: row 22’s “coming soon” wording for Z.ai’s direct API is carried from the anchor document’s verified citation of Z.ai’s docs, not re-fetched here; row 23’s OpenRouter alias record comes from the batch’s earlier API pull, not a fresh pull; and row 25’s HubSpot cell reports a raw-HTML read no “beta” in the body alongside a rendered token near the title that could not be resolved as a badge or a conversion artefact. Two classifications are explicitly inferred : row 12’s “GA by absence of a preview disclaimer” and row 6’s “v1.0 means stable”, both flagged as inferences in the contractual-meaning column. Row 9’s coverage side rests on a circulating editorial characterisation documented independently; no single outlet headline using “GA” was pinned, so the row is scored none located rather than upgraded. Coverage was sampled, not censused. One to two outlets per row via targeted search. A row marked “same” therefore proves only that the sampled outlets did not change the word, not that no outlet anywhere did; a row marked “none located” means no independent outlet was found in this search, never that none exists. Two rows were scored down, not up. Earlier working notes for this batch scored the Sol price-cut entry row 17 and v0 Connect row 18 as mismatches. Neither survived: no independent outlet applying its own status word to either page was located, and the “GA” framing attached to v0 Connect came from an in-house draft corrected before publication, not from an outlet. Both are recorded here as “none located”, the more conservative reading, which lowers the mismatch count rather than raising it. What was excluded, and why. a Items on the anchor document’s own unconfirmed or do-not-publish lists, since there is no availability language to audit if the event itself is unconfirmed. b Items the anchors date outside the window Cloudflare’s Pay Per Use rename, TikTok’s May and June features, the GSA and Anthropic integration reporting . c Items retired by the batch’s dedup rulings unless the wording angle specifically was new the glm-latest alias and Cursor cleared that bar . d Three in-window items with no availability axis at all: the MCP Rust SDK Tier 1 promotion, GitHub’s outage postmortem and CrewAI’s SSRF fix, which are maturity or security events, not preview-beta-GA events. Known limitations. Sampling, as above. Reader-mode fallbacks for two vendor pages, cross-checked against the anchor document’s independent fetch of the same pages. The contractual meaning column is populated only where a vendor publishes its own launch-stage definitions GitHub’s Pre-release License Terms; a general Microsoft Learn preview convention for the LinkedIn row and reads “not defined” everywhere else rather than borrowing another vendor’s definition. The table measures whether two published words match; it says nothing about whether either word was the right one. The five match categories were fixed before any row was scored. They are deliberately about words, not about whether the feature works: a row scores “same” if the outlet repeated or fairly paraphrased the vendor, whatever either party’s word was. Coverage matched The outlet used the vendor’s word or a defensible synonym for the same real-world state. Includes rows where neither side used any maturity word at all, such as a routine version bump treated as routine by both. Coverage went firmer The outlet used a stronger or more definite word than the vendor. Two are soft: a blanket “GA” headline sweeping in a tool that was never in beta, and a completed verb for a progressive rollout. The third is the OpenAI row, where an early-customer test became an unveiling. A word not found in vendor copy The coverage word is not confirmed in the vendor’s own published copy and is not a defensible synonym for what does appear — absent outright on the FTC release, unconfirmed for HubSpot, whose page carries an unresolved rendered token. Both rows involve a vendor hedging a draft policy statement; an unlabelled feature page and an outside party filling the gap with a more decisive word. No independent coverage word No outlet using its own status word for the specific item was found by targeted search. Nine of the ten are developer-tooling, gateway or model-router entries. The absence is recorded as an absence; nothing was filled by inference. Downgraded: 0 rows. 02 — What the Words MeanWhere a vendor actually defines preview, beta and GA. Three published definitions anchor this section, and none of them belongs to a lab. Only GitHub’s populates a cell in the table: Google Cloud’s is the reference ladder the rest are read against, and Microsoft Learn’s convention is background for a single row. Google Cloud’s launch-stage page states that at Preview “products or features are ready for testing by customers”, that “no SLAs or technical support commitments are provided” and that “The average Preview stage lasts about six months”, while General Availability products “are open to all customers, ready for production use, and covered by a Google Cloud SLA, where applicable” cloud.google.com/products, launch-stages section . GitHub’s site-wide Pre-release License Terms https://docs.github.com/en/site-policy/github-terms/github-pre-release-license-terms define the umbrella: “‘Pre-release’ means software, online services, and additional products and features that are not yet generally available, such as private preview, public preview, early access, technical preview, or similar versions,” provided “‘AS-IS’, ‘WITH ALL FAULTS’ and ‘AS AVAILABLE.’ GitHub may change or discontinue Previews at any time without notice.” Microsoft Learn’s general convention, used as background for the LinkedIn row rather than as a single-URL quote, is that preview features come without a service-level agreement and are not recommended for production workloads. Everything else in the table is undefined by its author. That is not an accusation; it is the reason the column exists. When Anthropic writes “is out of beta on the Claude API”, the release notes make clear that the operational meaning is that the anthropic-beta header is no longer required and that requests still sending it keep working. It is a technical statement, not a service commitment, and the table records it as such. Preview vs General Availability Preview: ready for customer testing, no SLA or support commitment, average stage about six months. GA: open to all customers, ready for production, covered by an SLA where applicable. The clearest published ladder in the set, and the one most coverage silently assumes every vendor shares. No row in this table cites it: it is the yardstick the others are read against, not a populated cell. Pre-release License Terms Any private preview, public preview, early access or technical preview is “not yet generally available”, provided as-is, with all faults, and may be changed or discontinued without notice. This is what “public preview” means for Copilot in Slack and Teams rows 10–11 and for a GitHub release tagged Pre-release row 15 . Preview convention background Preview features are provided without a service-level agreement and are not recommended for production workloads. Cited as general convention, not as a quote from one canonical page, which is why the LinkedIn row 31 carries it as background only. Not defined Anthropic, OpenAI, Cursor, DeepSeek, Z.ai, xAI, Meta, Vercel, OpenRouter, JetBrains, Shopify and HubSpot publish no launch-stage glossary for the items in this table. “Out of beta”, “-Exp”, “experimental”, “developer preview” and “now included” each mean what the surrounding sentence says and no more. 03 — The DatasetThe complete 36-row ledger. Every row is below, including the 21 where coverage and vendor agree and the ten where no coverage was located. Words in curly quotes are verbatim from the linked page. “Match” uses the five categories above. A handful of commercial SaaS and martech destinations carry nofollow ; everything else links plainly. Outlets are named with their URLs because each row is a citation, not a criticism: the point is that the two words can be checked, not that a particular desk got one wrong. | | Vendor · announcement | Vendor page · date | Vendor’s status word verbatim | Coverage · date | Coverage’s status word verbatim | Match | Contractual meaning | |---|---|---|---|---|---|---|---| | Labs and developer platforms rows 1–9 | ||||||| | 1 | Anthropic · Computer use tool computer toolset 20260801 | | nxcode.io, “Claude’s Agent Stack Is GA” https://www.nxcode.io/resources/news/claude-platform-agent-tools-ga-browser-skills-files-2026 · c. 2026-08-19 browser toolset 20260801 , a new tool /v1/skills releasebot.io https://releasebot.io/updates/anthropic aggregator · 2026-08-20 openai.com, “Offering Zero Data Retention for frontier models” https://openai.com/index/offering-zero-data-retention-for-frontier-models/ · 2026-08-19 reader-mode fallback after a direct 403 Security Boulevard https://securityboulevard.com/2026/08/openai-unveils-private-safety-processing-to-detect-ai-misuse-without-storing-enterprise-data/ and Cyberpress https://cyberpress.org/openai-unveils-private-safety-processing/ · by 2026-08-22 learn.chatgpt.com changelog https://learn.chatgpt.com/docs/changelog · 2026-08-19 cursor.com/changelog/08-19-26 https://cursor.com/changelog/08-19-26 · 2026-08-19 re-fetched independently of the anchor document’s own fetch github.blog changelog https://github.blog/changelog/2026-08-21-the-new-github-copilot-experience-in-slack/ · 2026-08-21 github.blog changelog https://github.blog/changelog/2026-08-21-shared-agentic-work-with-github-copilot-in-microsoft-teams/ · 2026-08-21 github.blog changelog https://github.blog/changelog/2026-08-18-enterprise-managed-settings-in-github-copilot-for-jetbrains/ · 2026-08-18 blog.jetbrains.com https://blog.jetbrains.com/dotnet/2026/08/19/rider-resharper-2026-2-1/ · 2026-08-19 dotnetramblings.com https://www.dotnetramblings.com/post/19 08 2026/19 08 2026 13/ · 2026-08-19 antigravity.google/changelog https://antigravity.google/changelog · 2026-08-20 GIGAZINE https://gigazine.net/gsc news/en/20260822-antigravity-remote-control · 2026-08-22 GitHub Releases, tag v0.57.0-preview.0 https://github.com/google-gemini/gemini-cli/releases · 2026-08-19 GitHub Releases, eve@0.44.0 https://github.com/vercel/eve/releases/tag/eve%400.44.0 · 2026-08-21 tracePolicy retains only conversations classified as public ; private and unknown remain fail-closed.” vercel.com/changelog https://vercel.com/changelog/gpt-5-6-sol-is-now-50-percent-off-a-lower-price · 2026-08-21 vercel.com/changelog https://vercel.com/changelog/connect-v0-apps-to-slack-google-and-100-other-services · 2026-08-21 vercel.com/changelog https://vercel.com/changelog/deepseek-v4-flash-with-vision-now-available-on-ai-gateway · 2026-08-21 api-docs.deepseek.com/news/news260821 https://api-docs.deepseek.com/news/news260821/ · 2026-08-21 Bloomberg https://www.bloomberg.com/news/articles/2026-08-21/deepseek-unveils-test-model-to-rival-anthropic-s-opus-4-8 · 2026-08-21 x.ai/news/grok-bot-more-plans https://x.ai/news/grok-bot-more-plans · 2026-08-21 9to5Mac https://9to5mac.com/2026/08/21/grok-bot-is-an-all-new-iphone-and-mac-app-from-spacexai-and-cursor/ · 2026-08-21 also Memeburn z.ai/blog/glm-5.3 https://z.ai/blog/glm-5.3 “GLM-5.3: Frontier Coding with Emergent Cyber Capabilities” · blog dated 2026-08-14; direct API live 2026-08-18 VentureBeat https://venturebeat.com/technology/glm-5-3-hits-the-api-at-1-4-4-4-per-million-tokens · 2026-08-18 z-ai/glm-latest floating alias expiration date: "2098-12-31" is OpenRouter’s no-expiry placeholder stealth/ox-alpha openrouter.ai/stealth/ox-alpha https://openrouter.ai/stealth/ox-alpha · listed 2026-08-20 20:04 UTC knowledge.hubspot.com KB article https://knowledge.hubspot.com/reports/use-company-surge-signals-powered-by-bombora · page dated 2026-08-11 Orange Marketing, “HubSpot Updates – Monday Recap” https://blog.orangemarketing.com/hubspot-updates-monday-recap-august-17-2026 · 2026-08-17 Meta for Business newsroom https://facebook.com/business/news/new-meta-ai-features-for-small-business · 2026-08-19 afaqs.com https://www.afaqs.com/news/digital/meta-rolls-out-new-ai-features-for-small-businesses-12398288 · 2026-08-20 Google Ads API release notes https://developers.google.com/google-ads/api/docs/release-notes · 2026-08-19 blog.google https://blog.google/products/ads-commerce/ai-max-testing-planning-tools/ · 2026-08-20 Search Engine Land https://searchengineland.com/google-adds-new-ai-max-testing-and-planning-tools-485566 · 2026-08-20 Search Central updates https://developers.google.com/search/updates · 2026-08-20 Search Engine Land https://searchengineland.com/google-makes-the-preferred-source-button-more-seamless-485548 · 2026-08-20 LinkedIn Marketing API recent changes https://learn.microsoft.com/en-us/linkedin/marketing/integrations/recent-changes?view=li-lms-2026-08 · “August 2026 - Version 202608” shopify.dev changelog https://shopify.dev/changelog/hydrogen-developer-preview-update-august-18-2026 · 2026-08-18 changelog.shopify.com https://changelog.shopify.com/posts/managed-markets-is-ending-delivered-duty-unpaid-ddu-support · announced 2026-08-11, effective 2026-08-24 ftc.gov press release https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-seeks-comment-enforcement-policy-statement-regarding-personalized-pricing · 2026-08-19 reader-mode fallback after a direct 403 Quartz https://qz.com/ftc-personalized-pricing-disclosure-warning-retailers-081926 · 2026-08-19 NVIDIA newsroom https://nvidianews.nvidia.com/news/nvidia-guarantees-sb-energy-s-ports-pike-technology-campus-in-ohio-to-exclusively-host-nvidia-ai-compute and NVIDIA 8-K https://www.sec.gov/ix?doc=/Archives/edgar/data/0001045810/000104581026000069/nvda-20260817.htm · 2026-08-17 CNBC https://www.cnbc.com/2026/08/17/nvidia-financing-open-ai-data-center-ohio.html · 2026-08-17 Marvell 8-K https://www.sec.gov/Archives/edgar/data/0001835632/000119312526356217/d412696d8k.htm · filed 2026-08-18 CNBC via Yahoo Finance https://finance.yahoo.com/technology/ai/articles/marvell-shares-jump-giving-google-132501905.html · 2026-08-19 04 — ScoreboardFive mismatches in 26, all in one direction . Counting the Match column gives 21 rows scored same, 3 upgraded, 2 not in vendor copy, 0 downgraded and 10 with no independent coverage word located. Restricting to the 26 rows where coverage was actually found, 21 of 26 81%, n=26 agree and 5 of 26 19%, n=26 do not. Every one of the five moved toward the more definite word: toward “GA”, “unveils”, “rolls out”, “beta” or “warns”. No sampled outlet hedged below its vendor. The chart is drawn directly from that column; no derived arithmetic sits between the table and the bars. Row 9 Cursor is the row that sets the denominator. Its vendor page carries no maturity word, and the firmer characterisation circulates in aggregator summaries and secondary paraphrase — but no single outlet headline carrying it was fetched and confirmed, so the row is recorded as coverage not located rather than scored as an upgrade. Scoring it as an upgrade instead would read as six mismatches in 27; the direction of the other five is unchanged either way. Where coverage was found 21 of the 26 rows with an independent coverage word sampled scored same 81%, n=26 . Most are either a fair synonym “out of beta” reported as “GA” or two parties declining to label a routine release at all. Toward the firmer word All five mismatches moved from a hedged or absent vendor word to a more definite coverage word. Zero moved the other way. The two sharpest OpenAI, FTC both involve a vendor deliberately describing something as not yet final. No independent word located More than a quarter of dated items drew no sampled outlet using its own status word. Nine of the ten are developer-tooling, gateway or model-router entries; the tenth is a Shopify deprecation notice. An honest absence, recorded rather than filled. Two interpretive points follow. First, the mismatches cluster where the vendor left a gap: no maturity word at all Anthropic’s brand-new browser tool, the HubSpot feature page , a progressive verb “starting to introduce” , or a deliberately non-final posture “previewing”, “proposed” . Where a vendor used a crisp status word “public preview”, “developer preview”, “Rolling out in September”, “Generally available” , sampled coverage repeated it. The pattern in this sample is not that outlets overwrite clear labels; it is that absent or soft labels get resolved to something firmer on the way to a headline. Second, the direction is uniform in a 26-row sample, which is suggestive rather than conclusive. A larger census, or a second window, could surface downgrades an outlet calling a GA feature “beta” out of caution, as the Grok Bot row nearly does . Until then the defensible claim is narrow: in this window, when the word changed between vendor page and coverage, it changed upward. 05 — Seed RowsThe three rows that started this, re-checked . This table began with three suspected mismatches from the week’s anchor work. Each was re-fetched independently for the ledger rather than carried forward, because a mismatch that rests on one fetch is a claim, and one that rests on two is a finding. Cursor row 9 : the word that is not on the page The 19 August changelog for always-on cloud agents contains neither “GA” nor “general availability”, confirmed on two independent fetches. The closest sentence is a continuing-work statement, quoted in full in the table. The “GA” framing is real in the sense that it circulates, but no single outlet headline carrying the literal word was fetched and confirmed, so the row is recorded as coverage not located rather than attributed to a named desk. The feature itself is covered in our always-on agents guide /blog/cursor-automations-always-on-agentic-coding-agents-guide and in the Cursor and Codex cloud-agents piece /blog/cloud-agents-outlive-the-session-cursor-codex-gitlab . HubSpot row 25 : a label with no confirmed author The knowledge-base article for Company Surge, dated 11 August, carries no “beta” in the raw HTML body; the only occurrence is an unrelated i18n dictionary string. A separate rendered fetch showed a “Beta” token adjacent to the page title that could not be resolved as a genuine badge or as an artefact of the HTML-to-text conversion, and an in-product badge sits behind login. The row therefore attributes “now in beta” to the third-party trackers that used it and makes no claim about what HubSpot said. Feature detail is in our Company Surge note /blog/hubspot-company-surge-bombora-intent-crm-record . OpenAI row 7 : previewing versus unveiling OpenAI’s own page says “Today, we’re previewing Private Safety Processing” and that it “is currently being tested with early customers”, with broader rollout and a technical white paper planned for September 2026. Two outlets led with “Unveils”. This is the cleanest upgrade in the table: a limited early-customer test described by the vendor as a preview, reported as a completed launch. Background on the mechanism is in our Private Safety Processing preview note /blog/openai-private-safety-processing-zdr-preview . “We’re continuing to improve cloud agents and the Cursor harness so always-on agents can operate as a system, building and shipping software on their own without the need for intervention at each loop.” No status word appears anywhere on the page. 06 — No Status WordRows where the vendor said nothing about maturity. Fourteen of the 36 vendor pages rows 2, 6, 9, 12–14, 16–18, 21–23, 27 and 30 carry no maturity word of any kind. These are the rows most exposed to a later upgrade, because a reader, a tracker or a summariser has to supply one to write a sentence about them. Four patterns appear in the table. GitHub Copilot enterprise settings for JetBrains row 12 No preview, beta or GA word anywhere in the entry, three days before two sibling entries that both say “This feature is in public preview and subject to change.” The only way to call this GA is to infer it from the missing disclaimer, which the table does, and labels as an inference. Anthropic Python SDK v1.0 row 6 · JetBrains 2026.2.1 row 13 A major version or a dot-release number stands in for a maturity word. Both vendor and coverage treat v1.0 as self-evidently stable and 2026.2.1 as self-evidently routine. The semver reading is convention, not a statement either vendor made. eve 0.44.0 row 16 · Sol price cut row 17 · Google Ads API v25.1 row 27 A default-behaviour flip, a price notice and a minor API version. None of these is a preview-beta-GA event, and forcing a status word onto them would misrepresent the source. The Sol page in particular is a near-miss: the model is GA, but this page never says so. Cursor row 9 · v0 Connect row 18 · Antigravity row 14 · Preferred Source row 30 A feature ships with descriptive, present-tense copy and no maturity tier. Two of the four Cursor, v0 Connect acquired a “GA” characterisation somewhere downstream; two Antigravity, Preferred Source were reported as descriptively as they were announced. The contrast between rows 10–12 is the sharpest illustration in the set. GitHub published three Copilot changelog entries three days apart. Two carry an explicit public-preview disclaimer and a pointer to terms that define what that means; one carries nothing. The sandbox-billing sentence appears in the Teams entry only, which is the kind of detail our Copilot in Slack and Teams note /blog/github-copilot-slack-teams-public-preview-sandbox-billing covers. Same vendor, same week, three different disclosure postures. 07 — Control RowsControl rows and the findings that cut against the pattern. Three rows form the control group, rows 34–36: wording that is definitive by construction, so that the hedged rows have something to be measured against. The NVIDIA 8-K row 35 and the Marvell 8-K row 36 are legal instruments with a single verb each “guarantees”, “issued” . Coverage of both filings matched the filings: CNBC’s “backing” and “giving” are fair paraphrases, and both outlets avoided “investment”, a word the NVIDIA filing does not support since the $105B figure is a financing ceiling. The regulator row is the one that failed. The FTC press release row 34 is as hedged as a regulator can be: “proposed”, “draft statement”, “seeking public comment”, plus an explicit disclaimer of authority to ban the practice outright. Quartz’s headline was “FTC warns retailers on personalized pricing disclosure rules”. There is no rule, and the release does not warn; it asks. This row and the HubSpot row are the table’s two “not in vendor copy” scores: here the word is simply absent from the release; for HubSpot the visible page copy does not confirm it, and an unresolved rendered token near the title keeps the question open. "The FTC does not have the legal authority to ban personalized pricing in all circumstances."— Andrew Ferguson, FTC Chairman, press release, August 19, 2026 Two rows cut against the “coverage inflates” reading and are kept for that reason. Grok Bot row 21 : xAI’s expansion notice says “now included” with no maturity label, while 9to5Mac’s framing retained “in beta”, the label the product launched with ten days earlier. Coverage was, if anything, more conservative than the vendor page. GLM-5.3 row 22 : VentureBeat’s “hits the API” is a plain factual headline for a plain factual milestone, and neither side reached for a tier word. The alias that sits on top of that model, z-ai/glm-latest row 23 , drew no independent coverage at all, which is its own small finding given that it is the row most likely to surprise a developer later; our floating-alias note /blog/glm-latest-floating-alias-model-pinning-risk explains why. The five Anthropic rows are also a negative finding of a kind. Four of the five were scored same even though the vendor said “out of beta” and the headline said “GA”, because the two describe the same operational state. The fifth browser use, row 2 is scored upgraded only because a blanket “GA” headline swept in a brand-new tool that was never in beta. The distinction is fine, and the table keeps it rather than rounding all five up to a mismatch. Detail on all five is in our Claude Platform betas-to-GA note /blog/claude-platform-betas-ga-computer-use-files-skills . 08 — Reading the WordHow to read a vendor’s status word before you rely on it. The table supports four habits for anyone who has to turn a vendor announcement into a decision. None of them requires trusting this ledger; each requires only the vendor page. Read the page, not the headline. In five of the 26 rows where coverage was located, the coverage word was firmer than the vendor’s; in a sixth Cursor, row 9 the firmer word circulates but no headline carrying it could be pinned at all. The vendor page is one click further, and it is the only wording the vendor itself published. Look for the definition, then assume its absence. Only GitHub’s rows link to published terms that say what “preview” costs you as-is, may be withdrawn without notice . For every other vendor in the table, the status word means what the sentence around it says and nothing more. Treat “no word” as a question, not an answer. Fourteen vendor pages carry no maturity word. A missing preview disclaimer is evidence of nothing in particular until the vendor is asked; the JetBrains settings row shows how a GA reading can rest entirely on silence. Separate the status from the event. The Sol price cut is not a GA announcement, the Grok Bot expansion is not a launch, and the FTC statement is not a rule. In each case the underlying thing is real and the status word being attached to it is not on the page. Our procurement questions checklist /blog/ai-procurement-questions-before-you-sign-2026 carries the contractual side of this: what to ask for in writing when the status word is undefined. Teams that want the same read-the-primary discipline applied to a vendor shortlist before a commitment can find it under our AI transformation work /services/ai-transformation . 09 — ConclusionA ledger, not a verdict . When the word changed between vendor page and coverage, it changed upward, every time, in this sample. Thirty-six dated announcements, two words each. Twenty-one of the 26 rows with coverage located agree with their vendor. Five do not, and all five resolved a hedged or absent vendor word into something firmer: “GA”, “unveils”, “rolls out”, “beta”, “warns”. Ten rows drew no independent coverage word at all, most of them developer-tooling changelog entries that only the vendor wrote about. The three seed rows held on independent re-fetch: two scored as mismatches, and the Cursor row recorded as coverage not located because no outlet headline carrying the firmer word could be pinned. The HubSpot row carries an explicit unresolved token rather than a claim in either direction. The contractual-meaning column is mostly empty, and that emptiness is the durable finding: outside GitHub’s published pre-release terms, almost no vendor in this window defines the status word it uses, which is what leaves the gap that coverage fills. The table is maintained at this URL and will be refreshed in place, re-dated through its modified time. Every row links to the page it was read from; anyone can re-run the comparison. Where a live page has changed since 22 August 2026, the change is the news, and the as-of date is the thing to cite with it.