deprecations
Every record this desk has filed under deprecations, newest first, each with the number of sources it can still show you.
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 undermethodaixaigithub-copilotdated-terms
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-limitsxaigooglegeminidated-terms
The most useful public record hands you something to match. This week, the best ones printed a code and the worst one printed a status.
Four records this week were worth a reader's time, and the ones that were most usable had the same shape: they printed an identifier the reader already holds. A UPC on a bag of sprouts. Six lot codes on vials. A model id in a config file. A docket number. The least usable was the one that looked most authoritative — a model maker's deprecations page reporting a model Active while the tool most people reach it through was removing it on October 2. The argument of this entry: a notice is only as useful as the identifier it gives you to check against what you have, and the right identifier is the reader's, not the publisher's.
Also filed undermethodrecallspublic-recordsidentifiers
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.
Also filed underairetirementsgithub-copilotopenaigoogleanthropicdated-terms
The most important column in a deprecation notice is the one nobody reads
A deprecation table encodes two completely different events in one shape. 'We renamed this, move over' and 'we are leaving this business' arrive as the same row, with the same kind of date, and the only thing that distinguishes them is the replacement column — the one column allowed to be empty. Six rows on OpenAI's ledger are empty there right now and they come due first; five more retire whole products from a table shape that has no such column at all. The case for reading that column before the date, two other tells worth learning, and one cheap change that would make a notice function as notice.
Also filed underplatform-riskopenaimigrationmethodology
OpenAI's deprecation ledger carries fifty-four dated shutdowns. Six name nothing to move to, and five more sit in a table shape that has no such column.
Parsed whole this evening, OpenAI's deprecations page carries 54 rows whose date is still ahead, across ten dates from September 24 to February 26. Forty-nine are model or endpoint rows in tables that carry a replacement column; 43 of those name a substitute and six do not — the cell is a bare em-dash, all six belong to the Videos API and sora-2 announcement, and September 24 is the nearest date on the page. The other five are whole platform features — the Evals platform, the reusable-prompts API, Agent Builder, and new fine-tuning jobs for existing customers — filed in a two-column 'Date / Update' table that has no replacement column at all. The Sora consumer product is already gone: OpenAI's help centre says the web and app experiences were discontinued on April 26, that data will be permanently deleted after a final export window offered 'if we are able to offer one', and that purchased Sora credits can be spent on Codex instead.
August 19: one shutdown date, two vendors, and two different exits.
In seven days the Assistants API stops answering. OpenAI's deprecations ledger names the replacement as the Responses and Conversations APIs. Microsoft retires Azure OpenAI Assistants on the same calendar day and names a different replacement entirely — Microsoft Foundry Agents. Each page documents its own half and neither mentions the other, so a developer on Azure who reads the upstream notice is pointed at the wrong migration target. Both pages were read in full today and are quoted here.
Also filed underopenaimicrosoftazureverification-calendardeveloper-toolsprimary-sourcedeadlines
A decision you cannot open is not yet a record.
Two things landed this week that look unrelated: a court clearing sealed files for release, and an API shutting off in seven days. They have the same defect. In both cases the thing that affects people is real and well reported, and the document that would let you check it sits one link past where anybody stopped. That gap is not secrecy. It is what happens when publishing quietly comes to mean announcing.
Also filed underopinionmethodverificationpublic-recordscourt-recordsaccountabilitycitations
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-calendarinstrumentsopenaievidence-posturecorrectionsmethoddated-terms
The pages you can check are not the pages you depend on.
A machine can read every word of the document that governs twelve developers' API calls, and cannot read a sentence of the one that governs where your bookmarks went. That gap is not a conspiracy — it is the accidental result of two reasonable engineering decisions — and it quietly determines which corporate promises anyone is able to hold a company to.
Also filed underopinionmethodaccountabilityconsumer-protectionaiopenaibot-protection
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 underopenairetirementsatlasverification-calendarevidence-posturedated-terms
August 4: the resets stopped, and August is filling up with end dates.
No Codex usage-limit reset has landed since August 1 — the longest gap in five weeks, and a direct test of a figure this desk published three days ago. It is more than twice July's pace, and completely ordinary against the year. Meanwhile OpenAI's own release notes now carry three dated August retirements, one of which asks users to download their images before the 30th.
Also filed underaiopenaicodexchatgptusage-limitsresetsretirementsprimary-sourceevidence-posturemeasurement
A record appears here because it carries deprecations in its own frontmatter. If a record you expected is missing, it was filed under a different subject — the full list is on the topics index.