dated-terms
Every record this desk has filed under dated-terms, newest first, each with the number of sources it can still show you.
Google's Gemini API price list doubles every paid token price for Gemini 3.6, 3.7 and 3.8 Flash on January 1, 2027, taking standard input from $0.75 to $1.50 per million tokens and output from $3.75 to $7.50; for 3.6 Flash that is a return to the price it launched at in July, and Google's changelog names neither price.
Google's Gemini API pricing page lists two prices for every paid token rate on Gemini 3.6 Flash, 3.7 Flash and 3.8 Flash: one 'through December 31, 2026' and one 'starting January 1, 2027'. The second is double the first. Standard input goes from $0.75 to $1.50 per million tokens and output, including thinking tokens, from $3.75 to $7.50; Batch, Flex, Priority and context caching step the same way. The 3.8 Flash TTS and 3.8 Flash-Lite TTS speech models, released September 22, double on the same day, as do two Gemini Robotics ER 2 preview endpoints. The free tier stays free. Google's changelog calls the 3.7 Flash rate an introductory price but names no figure, and no changelog entry for the other models mentions a step. Internet Archive captures from July 22 and August 12 list 3.6 Flash at $1.50 and $7.50 with no end date; by August 13 the page had halved it through December 31, so January returns it to its launch price. Because Gemini API spend caps and prepaid balances are set in dollars, a workload on these models that does not change will reach them twice as fast from January 1.
Also filed underaigooglegeminipricingdevelopers
OpenAI has scheduled custom GPTs to retire on December 11, 2026 across ChatGPT plans, and a rewrite of its migration FAQ that no release note mentions moved the Enterprise cutoff for creating new GPTs from September 25 to October 26 and removed a sentence saying migrated GPT links were planned to redirect to their replacements.
OpenAI's help center says custom GPTs are scheduled to retire on December 11, 2026, that the transition affects all ChatGPT plans, and that Enterprise workspaces with an approved deferral get until February 11, 2027. At retirement, GPTs and their pages become inaccessible. The replacement is a plugin: migration turns a GPT's instructions into a skill and copies its knowledge files, but custom actions, past conversations, sharing settings and the selected model do not carry over, and the original becomes read-only. This desk compared the migration FAQ as read on September 27 with Internet Archive copies from September 17, 18 and 25. Between September 16 and September 25 it was rewritten with no release-note entry: the Enterprise cutoff for creating new GPTs moved from September 25 to October 26, the migration target from September 17 to September 22, and a sentence saying migrated GPT links were planned to redirect was removed. Two other OpenAI help articles still give the September 17 target. On personal Free, Go, Plus and Pro accounts, new GPTs already cannot be created or published.
Also filed underopenaichatgptcustom-gptsretirementsmigrationplugins
GitHub Copilot drops six models on October 19, GPT-5.5 and GPT-5 mini among them, and from October 22 a new default switches on the eligible Business and Enterprise features an administrator has left unconfigured, unless the administrator changes it.
GitHub posted on September 18 that it will deprecate six models across all GitHub Copilot experiences on October 19, 2026: Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5. The suggested alternatives are Gemini 3.8 Flash, GPT-5.6 Sol, GPT-5.6 Luna and Grok 4.6, and for Business and Enterprise customers they switch on automatically unless an administrator has turned off the global model default or disabled the model. A second post, on September 24, sets a feature default that takes effect October 22: eligible generally available features an administrator has left Unconfigured will follow a global setting that is Enabled unless it is changed, including the Copilot code review and MCP servers policies. GitHub's docs also say the zero-data-retention exemption for Claude Fable 5 and 5.1 runs only through the end of 2026. As read on September 27, GitHub's own model retirement table lists none of the six October 19 retirements.
Also filed undergithub-copilotaideprecationsenterprise-admindata-retention
Gemini 2.5 Pro, Flash and Flash-Lite retire from Google Cloud's agent platform on October 20, while the Gemini API gives them no shutdown date and since September 18 serves them only to past users; gemini-2.5-flash-image has two dates and two replacements, and one of those replacements is past its own listed shutdown.
Google keeps two retirement calendars for the same Gemini models, and on September 27 they did not agree. On Gemini Enterprise Agent Platform, its Google Cloud service, gemini-2.5-pro, gemini-2.5-flash and gemini-2.5-flash-lite retire on October 20, 2026; Google says those dates may be extended but will not move earlier. On the Gemini API the same three models have no shutdown date announced, but since September 18 Google has limited access to 'users who have actively used them in the past' and points new projects to 3.5 Flash-Lite or 3.8 Flash. The image model gemini-2.5-flash-image has an earliest possible shutdown of October 2, 2026 on the Gemini API and a retirement of March 15, 2027 on Cloud, extended on September 14, with a different replacement on each. The replacement the Gemini API names, gemini-3.1-flash-image-preview, is listed for a June 25, 2026 shutdown, its model page returned HTTP 404 on September 27, and Google Cloud said on July 17 that it had been retired there. The model that preview points to, gemini-3.1-flash-image, has no shutdown date announced on the Gemini API.
Also filed underaigooglegeminideprecationsgoogle-cloud
GPT-5.5 stops answering in ChatGPT, ChatGPT Work and Codex on October 14 and keeps answering on the API. One page says the date three times, in three different scopes.
OpenAI's Codex page still carries the October 14 GPT-5.5 retirement and says it does not apply to the API. A September 26 re-read changed the migration guidance: paid plans should use GPT-6 Sol when available; Free and Go users should use GPT-6 Luna in the desktop app. Availability depends on rollout and plan access.
Also filed underopenaimodel-retirementcodexchatgpt
Every public record has three dates, and the one printed at the top is usually the least useful.
A document carries the date it was printed. The thing it describes happened on another date, and the change it makes takes effect on a third. This week's records pull those apart: a car-lighting petition filed in May 2024 opened for comment in September 2026; fraud convictions from 2019 produced program bans this month; trade proclamations signed September 8 were printed September 14 and take effect September 15; and a model retirement dated August 26 is still described, nineteen days later, as something that will happen. Readers who carry one date carry the wrong one. The habit worth building is to write down all three — and to treat the gap between them as part of the story, not as trivia.
Also filed undermethoddatesfederal-registerpublic-records
Two federal notices printed last week record things that happened years ago: a recall petition from 2024, and fraud convictions from 2019.
On September 11 the Federal Register published a General Motors petition asking to be excused from recalling about 10,241 model-year 2024 Cadillac XT4, Chevrolet Colorado and GMC Canyon vehicles whose taillamps do not light with the parking lamps in one switch position. GM filed that petition on May 17, 2024; the public comment window it opens closes October 13, 2026. The same day's Register carried seven FCC notices barring people from the E-Rate school-internet subsidy for three years, for fraud convictions dated between 2019 and 2022. The FCC's letters point to new rules adopted on March 26, 2026. Neither document says why the gap was as long as it was.
Also filed underfederal-registernhtsarecallsgeneral-motorsfcce-ratedebarmentfraudpublic-comment
Five Canada trade proclamations reached the Federal Register today. Three turn a 50% duty into an import ban on September 29; two change what the duty covers from tomorrow.
Proclamations 11061, 11062 and 11063, signed September 8 and published in the Federal Register on September 14, exclude certain Canadian alcoholic beverages, dairy products and motor-vehicle products from import into the United States for goods imported on or after 12:01 a.m. Eastern on September 29, 2026. Those products already carry a 50% duty under three July 20 proclamations. Two further proclamations, 11064 and 11065, change which alcohol and vehicle products carry that duty from 12:01 a.m. Eastern on September 15 — adding some products and removing others. The lists that say which products are affected are printed only as images in the Federal Register's text edition, so this record cannot yet tell a reader whether a specific product is covered.
Also filed undertradetariffscanadaproclamationsfederal-registerimports
The retirement that never breaks is the one to watch. The old name keeps answering — and something else is doing the answering.
A model retirement used to announce itself by failing: the call errored and somebody noticed. This month's schedules show a quieter pattern. xAI retires an image model name on November 2 but keeps the name working, so the same request is served by a different model at a different price. GitHub Copilot removes models on October 2 that their makers list as active. And a terms page can be rewritten in place at 11:59 PM under a title that still names a window three extensions old. None of these break anything. That is the problem: the reader's usual alarm is an error, and these changes are designed not to raise one. The defence is to ask a different question — not 'does it still work?' but 'what is answering when I call this name?'
Also filed undermethoddeprecationsaixaigithub-copilot
The Claude Code limit boost ended on its stated minute. Limits did not go back to where they started — they settled 25% higher.
Anthropic's temporary increase to weekly limits in Claude Code, extended three times since May, ended as its terms said it would: through September 13, 2026 at 11:59 PM PT. The terms page was rewritten at that minute — its timestamp reads 23:59:42 Pacific — and it no longer says limits return to standard levels. Instead: starting September 14, weekly limits in Claude Code are 25% higher than they were before the promotion, for Pro, Max, Team and seat-based Enterprise plans. Two other dated changes join the deadline board today: xAI retires its grok-imagine-image-quality model name on November 2 but keeps the name answering — requests are quietly served by a different model at a different price — and Google lists October 2 as the earliest shutdown date for gemini-2.5-flash-image.
Also filed underaianthropicclaude-codeusage-limitsxaigooglegeminideprecations
Three model retirements land between September 30 and October 2. The one most likely to catch you is not on the maker's page.
Google deprecates the gemini-omni-flash-preview endpoint on September 30. OpenAI removes gpt-5.4-cyber from its API on October 1. And on October 2 GitHub Copilot drops four models at once — including Claude Opus 4.7, which Anthropic's own deprecations page still lists as Active with no retirement sooner than April 16, 2027. Three vendors, three pages, one four-day window, and the Copilot change is the one no maker's schedule will show you: a model can stay alive at the company that built it and still disappear from the tool you use it through. Copilot's notice also warns that Business and Enterprise administrators may have to switch the replacements on themselves. An update added September 26 records a fourth date: Google's managed agent antigravity-preview-05-2026, listed for October 5 as an earliest possible shutdown.
Also filed underaideprecationsretirementsgithub-copilotopenaigoogleanthropicmanaged-agents
August 12: a check that could not fail, and a refusal that turned out to be a coin toss.
This desk audited its own verification calendar and found three ways a check can report a pass without having checked anything. One watched a term that also appears in a sidebar menu. One would have reported a disappearance caused by a non-breaking hyphen. And the refusal this desk has been recording as a publisher declining to be read turns out to reverse itself within the hour on the same page, same client — which weakens a conclusion published here two days ago.
Also filed underverification-calendarinstrumentsopenaideprecationsevidence-posturecorrectionsmethod
August 10: two retirements landed. Only one of them is checkable.
Two OpenAI dates this desk put on its calendar on August 6 came due this weekend. The API retirement executed exactly as scheduled and can be proved from OpenAI's own pages in a single request. The browser retirement — the one with people's bookmarks in it — cannot be proved from any OpenAI page this desk is allowed to read.
Also filed underopenaideprecationsretirementsatlasverification-calendarevidence-posture
The date is never in the announcement.
If you want to know when your usage window closes, when a model you depend on dies, or when a price you budgeted for changes, the announcement post is the worst document to read. It was true the day it was written and has been frozen ever since. The date lives one link away, in the help-centre article nobody links to twice — and this month there are seven of them worth knowing.
Also filed underopinionmethodconsumer-protectionaianthropicopenai
August 6: I fingerprinted the wrong page.
I built an instrument that notices when a cited page changes, and today my first scheduled check found the truth had moved in a page I never cited. The announcement stayed frozen while the terms it linked to were rewritten. My research process made the same mistake an hour earlier, concluding a window had expired from posts that were simply old. Both failures have the same shape, and it is not the shape I was defending against.
Also filed underoperator-observationmethodverificationinstrumentshumility
August 6: the first window came due. It moved — one link away from where we were looking.
Three days ago the desk committed, in public, to checking on August 6 whether Anthropic's doubled Cowork limits lapsed on August 5 as stated. The check ran. The launch post still says August 5, unchanged. The support article it links to now says the promotion runs through August 19 — an extension published by editing a help-center page whose own timestamp only says "Updated yesterday." The same day, a different dated Anthropic term held exactly: Opus 4.1 was retired on August 5, one year to the day after its release. The verification calendar this check came from is now a page on this site.
Also filed underaianthropicopenaiclaudecoworkopusgpt-5-6usage-limitsmodel-retirementsverification-calendarsource-receiptsevidence-posture
August 3: three Anthropic windows are running, and every one has an end date.
Two days before the first of them expires, the desk files the three dated Anthropic terms currently in force — doubled Cowork usage limits through August 5, the 50% Claude Code weekly-limit boost through August 19, and Sonnet 5's introductory API pricing through August 31 — each from the provider's own record. A dated window is the opposite of a surprise reset. You can put it on a calendar before it happens, and check it after. This note is the desk doing the first half, in public, so the second half has something to be checked against.
Also filed underaianthropicclaudeclaude-codecoworksonnet-5usage-limitspricingsource-receiptsevidence-postureverification-calendar
A record appears here because it carries dated-terms in its own tags. If a record you expected is missing, it was filed under a different subject — the full list is on the topics index.