Content Quality: Focused news piece on OpenAI's DevDay (Sept 29) subscription changes: new $500 Pro 500 plan with an Ultrafast tier, and the $200 Pro plan's Codex/Work allowance cut from 20x to 10x Plus from October 30. Clear structure, explicit 'What We Don't Know' section, and an attributed analysis section.
Source Verification: Snapshots: source-1.html.gz (thenextweb.com, HTTP 200, suspicious_patterns null) read from disk via gunzip. It confirms verbatim: 'A new Pro 500 plan costs $500 a month and is the only Pro plan with Ultrafast'; 'OpenAI said Ultrafast generates up to 300 tokens per second in Codex, up to eight times faster than standard'; 'From 30 October, Pro 200's included usage in ChatGPT Work and Codex falls from 20 times to 10 times the Plus allowance, OpenAI said'; GPT-6 Pro messages in chat fall from 200 to 100 a week; 'Existing subscribers keep their current limits until 29 October and get a one-time grant of usage credits worth $2,500, which expire at the end of the year'; 'OpenAI said it will not reintroduce Pro 200's five-hour limit'; 'Pro 200 is open to new subscribers again, but with a lower usage allowance than before'. TNW also confirms the Codex cloud, voice-steered CLI, Code Review and Codex Security Cloud items and that the Agents API supports computer use. FALLBACK USED: the Engadget and Unite.AI snapshots FAILED (manifest file null, HTTP 403 bot-blocked), so as a last resort the live URLs were read via WebFetch; these are NOT the archived provenance copies. Engadget (WebFetch) confirmed: Pro 500 'offers OpenAI's highest usage allowance and comes with access to its new Ultrafast feature - it also costs $500 per month'; $200 Pro usage decreases 'from 20x ... down to 10x'; GPT-6 Pro caps 200 to 100 per week; existing subscribers 'keep their current limits for a time, and will later receive a one-time credit' (no value or date given, as the article states); Sol usable by Plus, Pro, Business, Enterprise and Education users in Codex and Work; 'so it's possible other competitors will follow suit'. Unite.AI (WebFetch) confirmed: up to six times faster generation in the API; GPT-6 Astra Ultrafast available now on Pro 500 and Enterprise with GPT-6.1 Sol Ultrafast coming soon; Pro 500 at 25 times the ChatGPT Plus allowance and available now; gpt-6.1-sol at $2 per million input and $10 per million output tokens, one-fifth of Astra's standard prices; Codex on computer/phone/cloud, CLI voice steering and /agents view; Code Review; Codex Security Cloud; Agents API computer use on Pro 500 and Enterprise; OpenAI's note that own-model evaluations used its research environment or API while competitor evaluations came from public reports. WebFetch summaries were cross-checked against the TNW snapshot wherever they overlap; no discrepancies and no fabricated details found. Allowlist: engadget.com and thenextweb.com are allowlisted; unite.ai is NOT (chief:review warning). It is a specialist AI news outlet whose claims agreed with TNW and Engadget on all overlapping points, but it is the only source for the 25x figure, six-times API speed, Astra Ultrafast availability and Sol pricing, and its page could not be archived. Suspicious patterns: only the TNW snapshot was scanned (null); no snapshot exists for the other two sources, so no scan result applies to them.
Factual Accuracy: Headline and summary facts ($500 price, 20x to 10x, October 30, DevDay Sept 29) trace to TNW and Engadget. The 300 tokens per second figure is attributed in the body to OpenAI via TNW ('OpenAI said'). The headline states '300-Token-Per-Second' without attribution even though it is a vendor-stated maximum; this is a recoverable clarification. 'Halves' is arithmetically consistent with 20x to 10x. The $2,500 credit, Oct 29/30 dates and year-end expiry match TNW verbatim. The Agents API cross-reference /article/2026-09/16-openai-launches-agents-api-in-public-beta-opening-the-codex-agent-harness-to-developers resolves to a published article.
Overall Assessment: Substantively new, accurate and well-attributed coverage of OpenAI's Pro 500 launch and Pro 200 allowance cut, not a duplicate of the Sol article. Approved with a short clarification about the vendor-claimed speed in the headline.