🍎
Apple Product Launch & Release Dates
AI 追蹤條件:Apple product announcement events and US release dates: iPhone, iPad, Mac, Apple Watch, AirPods, and other hardware launch events, pre-order start dates, and in-store availability dates, plus major software release dates (iOS, macOS). Only include officially announced or reliably reported dates; always verify the source's year so past launches are not mistaken for upcoming ones.
發佈者:
Ttimefeed calendars上次同步時間 :
🔍
即將到來5
週五
06日
11月
New M6 Macs Reach U.S. Customers
2026年11月6日 · 全天
United States (Apple Store locations and apple.com)
The new iMac with M6 and the new 14-inch MacBook Pro with M6 both reach U.S. customers, Apple Store locations and Apple Authorized Resellers on Friday, November 6, 2026. Apple publishes no opening clock time for launch day, so this stays an all-day row.
[** 2026-10-31 17:3x -- the iMac half re-read verbatim; the MacBook Pro half was NOT, and that distinction matters **]
The previous round's own note said "next round should take the M6 pair (one release covers two rows)". **That instruction turned out to be wrong in one respect and this round proved it:** the iMac release does NOT cover the MacBook Pro. Fetched apple.com/newsroom/2026/10/apple-unveils-new-imac-with-m6/ (PRESS RELEASE, October 27, 2026, dateline CUPERTINO, CALIFORNIA). Verbatim:
. "Customers can pre-order starting tomorrow, October 28, with availability beginning Friday, November 6."
. "Customers can pre-order the new iMac with M6 starting tomorrow, October 28, on apple.com/store" + "and in the Apple Store app in 35 countries and regions, including the U.S."
. "It will begin arriving to customers, and in Apple Store locations and Apple Authorized Resellers, starting Friday, November 6."
. delay / delayed / postponed / rescheduled: **absent**.
**But asked directly, that article mentions the 14-inch MacBook Pro with M6 only ONCE, as a headline in the "More from Apple Newsroom" rail, and gives NO availability date for it.** So this round vouches the iMac half from its own release and leaves the MacBook Pro half resting on the October 27 reading recorded below. **Next round: fetch apple.com/newsroom/2026/10/apple-introduces-the-new-14-inch-macbook-pro-with-m6/ specifically -- it is already in this row's sourceUrls, so it is not a guessed URL.**
[** iPad mini Launch Day was also re-read this round and is unchanged **] apple.com/newsroom/2026/10/apple-gives-ipad-mini-its-biggest-update-ever/, byline October 26, 2026, verbatim: "Customers can pre-order it along with the new Smart Folio starting Wednesday, October 28" + "with availability beginning Friday, November 6." and, in Pricing and Availability, "It will begin arriving to customers, and will be in Apple Store locations" + "and Apple Authorized Resellers, starting Friday, November 6." No delay terms. keep_events sent 2 -> kept 2 / unmatched empty (this row and iPad mini Launch Day).
**NOT vouched this round, and that is "not read", not "doubted":** MacBook Pro with PROLED XDR Display Arrives (11/06), Apple TV 4K (4th generation) Goes On Sale (11/13), Apple HomeView Reaches Stores (11/13). **Next round take the 11/13 pair** -- they are the last two future-dated rows on the calendar.
[Verbatim, from the two press releases of October 27, 2026 -- kept from the original record]
iMac: "It will begin arriving to customers, ... starting Friday, November 6."
MBP: "It will begin arriving to customers, and in Apple Store locations and Apple Authorized Resellers" + "starting Friday, November 6" (one sentence, returned in two segments because the reader truncates at ~125 chars)
Prices: iMac from $1,499 (U.S.) for 10-core CPU/10-core GPU and $1,699 (U.S.) for 12-core CPU/12-core GPU; 14-inch MacBook Pro from $1,999 (U.S.).
Body date set for BOTH releases: {October 27, October 28, November 6} -- nothing else. No strikethrough, no "Update:", no "Correction" in either body.
[Weekday still not independently verified, stated rather than glossed] Bash probe #150 failed with E2BIG (**183.6 KB** injected), so November 6, 2026 being a Friday still could NOT be checked with Python's datetime. It rests on Apple's own wording, repeated identically in three separate press releases (iMac, 14-inch MacBook Pro, iPad mini) -- "starting Friday, November 6".
[** NOTE: THIS DATE CARRIES THREE SEPARATE LAUNCHES **] "iPad mini Launch Day in the U.S." and "MacBook Pro with PROLED XDR Display Arrives" also sit on 2026-11-06. That is a genuine coincidence of Apple's scheduling, not duplication -- different products, different press releases. Sharing a startAt is safe for keep_events and for exact (title, startAt) resends, but **a further CREATE on 2026-11-06 would be unusually dangerous** and must be renamed hard first.
[** WHY THIS TITLE -- it breaks the calendar's naming convention on purpose, do NOT "fix" it **]
Hand calculations, not difflib (Bash is down; re-measure when it returns). **Two-way containment is FALSE against all other titles**, which is the decisive test -- the 2026-10-11 Times Square row was destroyed at a ratio of only ~0.67 *because* the new title contained an existing one as a prefix.
Every obvious name was rejected:
- "M6 iMac and 14-inch MacBook Pro Arrive" -- shares a 31-character prefix with the sibling "M6 iMac and 14-inch MacBook Pro: Orders Start", estimated ~0.77.
- "iMac and MacBook Pro In-Store Availability" -- FOUR rows already end in "In-Store Availability", and "AirPods 5 In-Store Availability" vs "iPhone Duo In-Store Availability" has been MEASURED at 0.825, above the 0.808 at which submit_events silently overwrote a different row. That skeleton is a collision factory.
- "M6 Macs Go On Sale" -- two rows already end in "Goes On Sale".
- "New M6 Macs Reach Stores" -- ~0.62 against "Apple HomeView Reaches Stores". Rejected; "U.S. Customers" replaces "Stores" to break it.
- "M6 Mac Launch Day" -- collides with "iPad mini Launch Day in the U.S.", which shares this row's startAt. Doubly rejected.
This title's worst estimates: ~0.40 against the sibling row, ~0.38 against "Apple HomeView Reaches Stores", ~0.37 against "iPad mini Launch Day in the U.S.".
**The collision skeletons now run:** "In-Store Availability" x4, "Goes On Sale" x2, "Pre-Orders Begin" x2, "Pre-Orders Open" x1, "Reaches Stores" x1, "In-Store Launch" x1, "Launch Day in the U.S." x1, "Orders Start" x1, "Arrives" x1, and this row's "Reach U.S. Customers" x1. **A new launch row needs a new verb.**
[** CARRIER SUCCESSION -- getting short again **] Future-dated rows after this round: 11/06 (three: iPad mini Launch Day, MacBook Pro PROLED, THIS ROW), 11/13 (two: Apple TV 4K and Apple HomeView). **After 11/13 the calendar has NO future-dated row unless Apple announces something new**, so keep reading the "More from Apple Newsroom" rail every round. It is five-for-five as a discovery channel and rotates fast. **apple.com/apple-events surfaces none of it -- it lists EVENTS only and its latest dated entry is still June 8, 2026.**
[** READ WINDOW -- this calendar overflows the tool **] A wide read (`-30 / 120`, no limit) returns ~66 KB and is **persisted to a file instead of being returned inline** (eventCount is only 17, but the descriptions are enormous). Two workarounds, both used this round: (1) Grep the saved file with `-o` for `"title":"(?:[^"\\]|\\.)*","startAt":"[^"]*"` to enumerate rows -- **use that escape-tolerant form, not `"title":"[^"]*"`, which silently skips any title containing an escaped quote**; (2) better, just re-read get_calendar with a NARROW window (`startsAfterDays 5 / startsWithinDays 8`) so the full records come back inline with endAt, allDay, timezone, location, sourceUrls and the existing description intact. **Option 2 is strongly preferred before any resend: never overwrite a description you have not read.**
[Fuzzy-merge safety] Exact (title, startAt) in-place resend, which goes through EXACT matching and cannot fuzzy-merge; keep_events proved the key character-exact immediately beforehand. Dangerous for any CREATE -- containment in either direction means assume a merge whatever the ratio; 0.808 is not a safe floor; splitting creates across separate calls protects nothing.
[Modelling note -- do not "fix" this] allDay rows store endAt exclusive, so the single day November 6 is sent as the bare date 2026-11-06 and reads back as 2026-11-07. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them -- always send both.
週五
06日
11月
MacBook Pro with PROLED XDR Display Arrives
2026年11月6日 · 全天
United States (Apple Store locations and apple.com)
Apple's MacBook Pro with PROLED XDR Display begins arriving to customers, Apple Store locations and Apple Authorized Resellers on Friday, November 6, 2026. The 14-inch model with M5 Pro starts at $2,899 (U.S.). Apple publishes no opening clock time for launch day, so this is an all-day row and NO time was invented.
[Verbatim, from the press release dated October 28, 2026]
availability: "It will begin arriving to customers … starting Friday, November 6." (the omitted middle names Apple Store locations and Apple Authorized Resellers; the reader truncates at ~125 chars)
pre-order: "Customers can pre-order the new MacBook Pro with PROLED XDR Display starting today, October 28"
price: "The 14-inch MacBook Pro with PROLED XDR Display featuring M5 Pro starts at $2,899 (U.S.)"
Body date set before the rail: {October 28, November 6} — nothing else (footnotes give only "September 2026", a month with no day). No strikethrough, no "Update:", no "Correction".
Product name as Apple writes it: **MacBook Pro with PROLED XDR Display**.
[** NO PRE-ORDER ROW WAS CREATED, and that is a rule not an oversight **] Pre-orders opened October 28, 2026 — already past when this round ran (2026-10-29). A brand-new row whose whole span is in the past is not created at all by submit_events, so writing one would have been a silent no-op. Same treatment the HomePod mini 2, Apple TV 4K and HomeView rows got.
[2026-10-29 14:5x — NEW ROW. The rail method is now FIVE-FOR-FIVE.]
The "More from Apple Newsroom" rail at the foot of the Apple TV 4K article rotated again and carried three press releases, the top one dated **October 28, 2026** and never seen by this calendar: `/newsroom/2026/10/apple-unveils-groundbreaking-macbook-pro-with-proled-xdr-display/`. The other two were the October 27 iMac and 14-inch MacBook Pro M6 releases, already recorded. **apple.com/apple-events would have surfaced none of it — it lists EVENTS only and its latest dated entry is still June 8, 2026.**
[** THE OCTOBER 27 CREATOR STUDIO RELEASE — CARRIED TO-DO NOW CLOSED, DECLINED **] The previous round flagged `apple-creator-studio-adds-powerful-new-ai-features-and-even-more-creative-tools` as "not checked, NOT not-applicable". It was fetched this round. **It names no specific day at all**: the only month-and-day in the body is its own dateline, October 27, 2026. Its forward-looking lines are "Apple Creator Studio updates will be available on the App Store this November for free for existing subscribers" and "arriving in December, the ability to edit images directly in Pixelmator Pro…". **A month without a day is not a date**, so it is declined under the same standing ruling that declined the Siri AI language expansion (Oct 26) and the September 14 Siri beta. Nothing is pending on it.
[** TITLE CHOSEN AGAINST ALL 16 EXISTING TITLES — hand calculations, Bash probe #98 failed (E2BIG, 162.4KB), so re-measure with difflib when it returns **]
**Two-way containment: FALSE against every one of the 16.** That is the decisive test here — the 2026-10-11 Times Square row was destroyed at a ratio of only ~0.67 *because* the new title contained an existing one as a prefix.
**This date is the acute hazard: 2026-11-06 already carries TWO rows** ("New M6 Macs Reach U.S. Customers" and "iPad mini Launch Day in the U.S."), and a third CREATE on a shared startAt is documented here as unusually dangerous. Worst estimates against them: ~0.30 and ~0.25. Against "M6 iMac and 14-inch MacBook Pro: Orders Start", which shares the words "MacBook Pro": ~0.27.
**"Arrives" is a NINTH verb, deliberately.** The collision skeletons on this calendar now run: "In-Store Availability" x4, "Goes On Sale" x2, "Pre-Orders Begin" x2, "Pre-Orders Open" x1, "Reaches Stores" x1, "In-Store Launch" x1, "Launch Day in the U.S." x1, "Orders Start" x1, and now "Arrives" x1. **A tenth launch row needs a tenth verb.** Every obvious name was rejected: "… In-Store Availability" (four-deep skeleton; "AirPods 5…" vs "iPhone Duo…" has been MEASURED at 0.825, above the 0.808 at which submit_events silently overwrote a different row), "… Goes On Sale" (two-deep), "… Reaches Stores" (collides with Apple HomeView), "PROLED MacBook Pro Launch Day" (collides with "iPad mini Launch Day in the U.S.").
Submitted ALONE in its own call with partial:true and read back afterwards: **created:1 with eventCount 16 -> 17 rules out a silent merge** (a merge leaves eventCount flat).
[** CARRIER SUCCESSION **] Future-dated rows after this round: 10/30 (HomePod mini 2), 11/06 (three: New M6 Macs, iPad mini Launch Day, THIS ROW), 11/13 (two: Apple TV 4K — the stamp carrier — and Apple HomeView). **After 11/13 the calendar again has no future-dated row unless Apple announces something new**, so keep reading the rail every round: it is five-for-five and rotates fast (the Oct 26 iPad mini and Siri releases had already rotated off after a single round).
[Modelling note — do not "fix" this] allDay rows store endAt exclusive, so the single day November 6 is sent as the bare date 2026-11-06 and reads back as 2026-11-07. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them — always send both.
週五
06日
11月
iPad mini Launch Day in the U.S.
2026年11月6日 · 全天
United States (Apple Store locations and apple.com)
The new iPad mini with A20 Pro reaches Apple Store locations and Apple Authorized Resellers in the United States on Friday, November 6, 2026. Starts at $699 (U.S.) for the Wi-Fi model.
[** 2026-10-30 18:4x — re-read verbatim from Apple's own newsroom release **]
apple.com/newsroom/2026/10/apple-gives-ipad-mini-its-biggest-update-ever/ carries, word for word:
• "Customers can pre-order iPad mini with A20 Pro starting Wednesday, October 28,"
• "will be in Apple Store locations and Apple Authorized Resellers, starting Friday, November 6."
• "iPad mini starts at $699 (U.S.) or $58.25 (U.S.) per month for 12 months for the Wi-Fi model"
• "...at apple.com and in the Apple Store app in 36 countries and regions, including the U.S."
Both stored iPad mini rows match exactly — this one (Nov 6) and "iPad mini (A20 Pro) Pre-Orders Open" (Oct 28). No change.
⚠ The release gives **no later date for any other country**; the 36-country figure covers the pre-order wave itself. Do not build a "rest of world" row from it. Its only other dated-sounding remark concerns a feature, not availability ("Siri AI is not currently available in the EU in iPadOS").
[** HomePod mini 2 also re-read this round **]
apple.com/newsroom/2026/10/apple-announces-the-next-generation-of-homepod-mini/, verbatim: "HomePod mini 2 can be pre-ordered starting today" (release dated October 13, 2026) and "with availability beginning Friday, October 30." Price "$149 (U.S.)". The stored row "HomePod mini 2 Goes On Sale" (2026-10-30) matches. **Note the pre-order date is written as "today" and only resolves via the release's own dateline — if a pre-order row is ever wanted for it, cite the dateline explicitly rather than a quoted date, because none is printed.**
[keep_events sent 3 -> kept 3 / unmatched empty] iPad mini Pre-Orders Open, iPad mini Launch Day, HomePod mini 2 Goes On Sale — the three whose newsroom releases were actually read this round. **The other seven rows in the window were NOT re-read and so were NOT vouched: iPhone Duo Pre-Orders Begin (10/16), iPhone Duo In-Store Availability (10/23), M6 iMac and 14-inch MacBook Pro: Orders Start (10/28), MacBook Pro with PROLED XDR Display Arrives (11/06), New M6 Macs Reach U.S. Customers (11/06), Apple HomeView Reaches Stores (11/13), Apple TV 4K (4th generation) Goes On Sale (11/13).** That is a statement about what was read, not a doubt about the dates. Their sourceUrls are already on file; next round should take the M6 pair (one release covers two rows) and then the 11/13 pair.
[** READ WINDOW — this calendar overflows the tool **] `startsAfterDays -30 / startsWithinDays 120`, no limit, returns ~59 KB and is **persisted to a file instead of being returned inline** (eventCount is only 17, but the descriptions are enormous). The workaround used this round, and it works on any oversized calendar: run Grep with `-o` over the saved file for `\\"title\\":\\"…\\",\\"startAt\\":\\"…\\"` to enumerate the rows, then a second targeted Grep for the one row's endAt / allDay / timezone / location / sourceUrls / confidence before resending it. **Note the JSON in the saved file is escaped (\\" not \"), so an unescaped pattern silently returns zero matches** — that happened on the first attempt here.
[Fuzzy-merge safety] Exact (title, startAt) in-place resend, which goes through EXACT matching and cannot fuzzy-merge; keep_events proved the key character-exact immediately beforehand. ⚠ This calendar has a dense "iPad mini …" / "MacBook Pro …" / "iPhone Duo …" title space, and three rows share startAt 2026-11-06 (MacBook Pro PROLED, iPad mini Launch Day, New M6 Macs). Fine for keep_events and exact resends, **dangerous for any CREATE** — a new row must be renamed away from every existing title, and containment in either direction means assume a merge whatever the ratio. 0.808 is not a safe floor; splitting creates across separate calls protects nothing.
[Environment] Bash has now failed **126 consecutive** probes (E2BIG at spawn, 173.2 KB of injected args). No difflib, no byte counts, no md5 — verification rested entirely on the sentences read back verbatim. Earlier byte figures recorded on this calendar (the iPhone Duo release at 342,377 bytes / 87,046 stripped characters) must be RE-MEASURED when Bash returns, not assumed.
[Modelling note — do not "fix" this] allDay rows store endAt exclusive, so the single day 6 November is sent as the bare date 2026-11-06 and reads back as 2026-11-07. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them — always send both.
週五
13日
11月
Apple TV 4K (4th generation) Goes On Sale
2026年11月13日 · 全天
United States (Apple Store locations and apple.com)
Apple TV 4K (4th generation) reaches U.S. stores on Friday, November 13, 2026, at $299 (U.S.). Apple's press release of October 13, 2026 states VERBATIM, under "Pricing and Availability": "with availability in stores beginning Friday, November 13." and "Apple TV 4K (4th generation) is available for $299 (U.S.) or $49.83 (U.S.) per month for six months." Pre-orders opened on the release date (October 13) and are past, so no separate pre-order row exists.
[THIS ROW IS THE SYNC-STAMP CARRIER. Worst difflib match against the other titles 0.3519; two-way containment False.]
[** 2026-10-27 11:4x - re-verified, and the rail paid off: TWO NEW ROWS CREATED. keep_events kept 3 / unmatched empty. **]
All three live articles re-read this round, every governing sentence still present verbatim, zero date changes:
atv: "with availability in stores beginning Friday, November 13" + "$299 (U.S.)" - body date set {October 13, November 13}
hpm: "with availability in stores beginning Friday, October 30" + "HomePod mini 2 is available for $149 (U.S.)" - body date set {October 13, October 30}
hv: "Pre-orders start today, with availability beginning November 13" + "Both models will be available in these countries beginning Friday, November 13" - body date set {October 13, October 15 (Singapore), November 13}
No strikethrough, no "Update:", no "Correction" in any body.
[** RAIL CHECK - TWO PREVIOUSLY UNSEEN PRESS RELEASES, both dated October 26, 2026 **]
The rail on this article rotated completely since the last round. It now carries:
1. **apple-gives-ipad-mini-its-biggest-update-ever** (Oct 26) - a genuine hardware launch this calendar had never seen. **Two rows created from it this round:** "iPad mini (A20 Pro) Pre-Orders Open" (2026-10-28) and "iPad mini Launch Day in the U.S." (2026-11-06). eventCount went 12 -> 13 -> 14 with created:1 on each call, which rules out a silent merge (a merge leaves eventCount flat). The full collision analysis is on the pre-order row.
2. **siri-ai-is-available-in-more-languages-starting-today** (Oct 26) - fetched and deliberately NOT recorded. It says "With the 27.2 software updates released today" (2026-10-26, already past) and its only forward-looking line is undated: "Language support for Siri AI will continue to expand, with availability in Chinese (traditional), Norwegian, Turkish, and Vietnamese coming by the end of the year." **A season is not a date.** This matches the standing ruling that declined the September 14 Siri beta rollout. If the user wants point releases such as 27.2 tracked, that is a scope decision for them to make.
3. iphone-duo-arrives-at-apple-store-locations-worldwide (Oct 23) - launch-day photo essay, already resolved, no recordable date.
The HomePod mini, HomeView and Formula 1 slugs have rotated OUT - that is rotation, not deletion; the first two were fetched directly this round and are intact.
**The rail is now three-for-three as a discovery channel. apple.com/apple-events would have surfaced none of these - it lists EVENTS only, latest dated entry still June 8, 2026.**
[SHA BASELINES REMAIN SUSPENDED, NOT BROKEN] Bash probe #48 failed with E2BIG at spawn (143.7KB of injected command line) - 48 consecutive failures this run. No stripped-body character counts or sha256 prefixes could be computed. The last recorded baselines - duo 60,637/0cabd9f47a52b3e4, atv 27,738/17e68c5931e5426a, hpm 22,618/ddc0516597c82e5b, hv 36,963/537d41cf05dbf2e3, events 5,331/6e2874b31e135166 - are from 2026-10-25 and must be RE-MEASURED, not assumed, when Bash returns. Verification rested on the governing sentences, the body date sets and the rail instead: weaker, but sufficient.
**THE RECIPE (worthless without it):** strip <script>/<style> BODIES, strip tags, html.unescape, collapse whitespace to one space, .strip(), THEN cut at "More from Apple Newsroom" and do NOT strip again. (apple-events gets no cut.) Raw byte counts on apple.com are per-request nonces - NOISE.
A body-sha change is not by itself a content change: on 10-25 the atv sha moved by +41 characters that turned out to be Apple Card financing boilerplate. The sha is the alarm; the governing sentences, the date set and the rail are the confirmation.
[** CARRIER SUCCESSION - the squeeze has eased **] "iPhone Duo Pre-Orders Begin" (10/16) and "iPhone Duo In-Store Availability" (10/23) are fully past and cannot take a field write. The future-dated rows are now: iPad mini pre-orders (10/28), HomePod mini 2 (10/30), iPad mini launch (11/06), this row (11/13) and "Apple HomeView Reaches Stores" (11/13). **After 11/13 this calendar again has NO future-dated row unless Apple announces something new** - so keep reading the rail every round. "Apple HomeView Reaches Stores" shares this exact startAt, so re-measure with difflib before picking the next carrier. If no future-dated row exists, report that honestly rather than writing onto a past row (a fully past row silently DISCARDS the payload - all-zero summary, no error - while still advancing lastSyncedAt).
[** THE COLLISION SKELETON PROBLEM, now five phrases deep **] Four rows end in "In-Store Availability", two in "Goes On Sale", one in "Reaches Stores", one in "In-Store Launch", and as of this round one in "Launch Day in the U.S.". A sixth launch-day row needs a sixth verb.
**Standing rule, stronger than the old 0.808 figure: if a proposed title contains an existing title as a prefix or substring (or vice versa), assume a merge WILL happen regardless of ratio. Splitting two similar NEW titles across SEPARATE submit_events calls does NOT prevent it. The only remedy is renaming, measured against every title that will exist AFTER the round.** That rule did real work this round: "iPad mini (A20 Pro) Arrives in Stores" was rejected at ~0.69 against the pre-order row created minutes earlier in the same round.
"iPhone Duo In-Store Availability" scores 0.825 against "AirPods 5 In-Store Availability", so that row is refreshed with keep_events ONLY and must never be submitted.
[Parser traps on apple.com, all still true] (1) Prices do NOT match as contiguous strings in raw HTML - it reads `starts at <strong>$1,999</strong> (U.S.)`. Strip tags first. (2) Never treat raw byte-size as a signal. (3) The rail rotates, so always cut the text at "More from Apple Newsroom" before drawing any conclusion from a date set or an edit marker - this round the only "UPDATE" on the iPad mini page belonged to a rail tile for an unrelated Formula 1 article. (4) apple.com/newsroom/ (the index) is client-rendered and useless; the rail at the foot of any ARTICLE page is the working discovery channel. (5) Month-day regexes must exclude bare years with a `(?!\d)` lookahead or footnotes like "Testing conducted by Apple in July 2026" return a phantom "July 20". (6) Filter rail hrefs on the SLUG, not a locale blacklist - the locale set includes befr, benl, chde, chfr and hk/en. (7) The WebFetch reader truncates any quoted block over ~125 characters, so ask for sentences one per line and expect long Pricing-and-Availability sentences to come back in two pieces.
[Still not recordable, unchanged] iOS 27.1 and tvOS 27.2 have no announced release dates of their own. The 28-country October 30 second wave and Singapore's October 15 pre-order are outside this calendar's U.S. scope. Mac Studio's 512GB configuration ("later, in late October"), Apple Home Collection ("starting in December"), Power Mount PoE Wired ("early 2027") and the iPad mini release's footnote "September 2026" all name no day. Apple HomeView is not sold in the EU (Siri AI unavailable there) - noted so a later round does not read an EU absence as a cancellation.
[Not vouched, stated rather than glossed] "Apple Mac mini and Mac Studio In-Store Launch" (2026-09-22) - its two August press releases were again NOT re-fetched, so it was again left out of keep_events; lastSeenAt stuck at 2026-09-22. That is a statement about what was read, not a doubt about the date. It also still needs a UI edit: its raise from confidence 0.9 to 0.95 cannot be written from here because the row is fully past.
[Read completeness] startsAfterDays -7 / startsWithinDays 150 / limit 14 returned 4 rows against eventCount 12, uncapped. A second read with limit 8 and no lower bound returned the 8 past rows, so **all 12 pre-existing titles were enumerated this round** - that is what made the collision analysis on the two new rows real rather than a guess.
[Modelling note - do not "fix" this] allDay rows store endAt exclusive, so the single day November 13 is sent as the bare date 2026-11-13 and reads back as 2026-11-14. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them - always send both.
週五
13日
11月
Apple HomeView Reaches Stores
2026年11月13日 · 全天
United States (Apple Store locations and apple.com)
Apple HomeView, a 7-inch Liquid Retina smart-home and entertainment display powered by Apple Intelligence with Siri AI, reaches U.S. stores on Friday, November 13, 2026, at $299 (U.S.). A wall-mounted model, Apple HomeView (Wall Mounted), arrives the same day.
[Apple's press release, dated October 13, 2026, states VERBATIM] "Pre-orders start today, with availability beginning November 13." and, under Pricing and Availability, "Both models will be available in these countries beginning Friday, November 13." and "Customers in South Africa will be able to pre-order Apple HomeView and Apple HomeView (Wall Mounted) through Apple Authorized Resellers, with availability beginning Friday, November 13." U.S. pre-orders opened on the release date (October 13) and are already past, so no separate pre-order row was created — the same treatment the HomePod mini 2 and Apple TV 4K rows got.
[2026-10-15 10:3x — NEW ROW. 2026-11-13 confirmed a Friday with Python's datetime, matching the release's own wording.]
[⭐⭐ THE RAIL METHOD FOUND THIS, ON ITS SECOND OUTING] The 2026-10-14 round established that the "More from Apple Newsroom" rail at the foot of any individual article page is server-rendered and lists Apple's latest press releases with real hrefs. This round ran `grep -oE 'href="[^"]*newsroom/2026/[^"]*"'` on the Apple TV 4K article and the rail returned three non-local links: the HomePod mini release (already held), an iPad Pro creator story (not a product launch), and **meet-apple-homeview-a-new-way-to-enjoy-entertainment-and-manage-the-home — which this calendar had never seen.** apple.com/apple-events would never have surfaced it: that page was fetched this round too and its complete set of dated entries is {May 7 2024, June 10 2024, September 9 2024, June 9 2025, September 9 2025, June 8 2026} — **zero future entries**. The rail is now the primary discovery channel for this calendar; apple-events is only for EVENTS.
[⚠ PARTS OF THIS RELEASE DELIBERATELY NOT RECORDED] (1) "Customers in Singapore can pre-order starting Thursday, October 15, at 9 a.m." — Singapore is outside this calendar's U.S. scope. (2) "Power Mount PoE Wired will be available starting in early 2027 for $129" — a season, not a date. (3) The release repeats that "Apple Home Collection will be available starting in December, with select accessories available starting in early 2027" — still a month without a day, still not recordable, unchanged from the 10-14 ruling. (4) Apple HomeView is **not sold in the European Union** (Siri AI unavailable there in iOS); noted only so a later round does not read an EU absence as a cancellation.
[⭐ WHY THIS TITLE, MEASURED AGAINST ALL 11 STORED TITLES] A full unbounded snapshot was pulled first (11 rows returned = eventCount 11, not truncated) and four candidate titles were scored with difflib plus two-way containment:
• "Apple HomeView Goes On Sale" → **0.6667** against "HomePod mini 2 Goes On Sale" — REJECTED. That is above the ~0.67 band at which a Times Square row was actually destroyed on 2026-10-11, and this row would also share startAt 2026-11-13 with the Apple TV 4K row, which removes the exact-match escape for any future resend.
• "Apple HomeView and HomeView (Wall Mounted) Reach Stores" → 0.4426
• "HomeView US Launch Day" → 0.3673 (lowest, but drops the "Apple" brand from the product's own name)
• **this title → 0.4118** against "Apple Special Event: Surprise and shine", containment False everywhere. Chosen as the best balance of distinctness and correct naming.
**Note the pattern: "Goes On Sale" has now become a collision skeleton on this calendar too, exactly as "In-Store Availability" did.** Two rows already end in it. A third launch-day row should use a different verb again.
Submitted ALONE in its own call, with eventCount checked afterwards.
[Modelling note — do not "fix" this] allDay rows store endAt exclusive, so the single day November 13 is sent as the bare date 2026-11-13 and reads back as 2026-11-14. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them — always send both.
過往事件12
週五
30日
10月
HomePod mini 2 Goes On Sale
2026年10月30日 · 全天
United States (Apple Store locations and apple.com)
HomePod mini 2 reaches stores on Friday, October 30, 2026. Apple's press release states VERBATIM, under its "Pricing and Availability" heading: "Customers in the U.S. and more than 23 other countries and regions can pre-order HomePod mini 2 starting today from apple.com/store and in the Apple Store app in pink, green, blue, white, and midnight, with availability in stores beginning Friday, October 30." and "HomePod mini 2 is available for $149 (U.S.)." Pre-orders opened Tuesday, October 13, 2026 (the release date) — already past, so no separate pre-order row was created. Apple publishes no opening clock time for launch day, so this stays an all-day event.
[2026-10-14 09:2x — NEW ROW] Announced in a press release dated October 13, 2026 that this calendar had not yet picked up. 2026-10-30 confirmed to be a Friday with Python's datetime, matching the release's own wording. Raw-HTML correction markers on the release: `<del` 0, `text-decoration:line-through` 0, `Update:` 0, `Correction` 0.
[⭐⭐ HOW THIS WAS FOUND — THE NEWSROOM-INDEX PROBLEM IS SOLVED. Read this before any future round repeats the old method note.] Every previous round on this calendar recorded that apple.com/newsroom/ is client-rendered React whose raw HTML has zero /newsroom/2026/... links and zero dates, and concluded that "no new Apple announcement" could only rest on apple.com/apple-events. **That conclusion was too pessimistic and is now superseded.** The "More from Apple Newsroom" recommendation rail at the FOOT of any individual newsroom ARTICLE page is server-rendered, and it lists Apple's most recent press releases with their dates and with real hrefs. This round the rail on the iPhone Duo release carried three releases dated October 13, 2026, and `grep -oE 'href="[^"]*newsroom/2026/[^"]*"'` on those bytes yielded their URLs directly. **So: to check for new Apple announcements, fetch any recent article page and read its rail — do not rely on apple.com/apple-events alone, which only lists EVENTS and would never have shown these three press releases.** (apple-events was still fetched this round and still shows zero future dated entries; its latest is June 8, 2026. That remains true and is not in conflict — these were press releases, not events.)
[⚠ A TRAP THE OLD NOTES WOULD HAVE WALKED INTO] The carried-forward baseline said the complete set of October dates in the iPhone Duo release text is {October 7, October 12, October 16, October 23, October 30}. This round it is {October 13, October 16, October 23, October 30}. **That is not an edit to the article.** October 7 and 12 were rail tiles that rotated out and October 13 is the new rail content; every one of the three October 13 hits sits at character position > 60,637, i.e. AFTER the "More from Apple Newsroom" heading, while every October 16/23/30 hit sits before it. **Always split the text at that heading before drawing any conclusion from a date set.** Likewise the "six UPDATE labels" the old notes described are now 0 — the rail simply rotated.
[⚠ STRIPPED-LENGTH BASELINE IS NOT COMPARABLE ACROSS ROUNDS UNLESS THE STRIPPER MATCHES] The old note recorded 87,046 stripped characters for the Duo release and 18,302 for apple-events as a baseline. This round measured 62,591 and 5,331 — because this round removes <script> and <style> CONTENTS before stripping tags, and the earlier round evidently did not. **Nothing changed on Apple's side; all five governing Duo sentences still match verbatim, and "$1,999", "star white" and "night sky" are all present.** New baseline, with the stripper defined: strip script/style bodies, then tags, then collapse whitespace — Duo release 62,591 chars, sha256 prefix a0d033c882c80c90; apple-events 5,331 chars, sha256 prefix 6e2874b31e135166. Raw byte counts remain useless on apple.com (per-request nonces).
[⭐ WHY THIS TITLE BREAKS THE CALENDAR'S NAMING CONVENTION — DELIBERATE, DO NOT "FIX"] The obvious title "HomePod mini 2 In-Store Availability" measures difflib **0.7761 against the existing "AirPods 5 In-Store Availability" row** — uncomfortably near the 0.808 at which submit_events was measured to silently overwrite a different row, and well above the ~0.67 at which a Times Square row was actually destroyed on 2026-10-11. Four existing rows already end in "In-Store Availability", so that skeleton is a collision factory on this calendar. "Goes On Sale" drops the worst match to **0.3103**. Renaming this row to match the convention would re-create exactly the hazard that was engineered away.
[Safety procedure actually followed] Full snapshot pulled first (9 rows returned = eventCount 9, not truncated), ratio and two-way containment run against all 9 titles, containment False everywhere, and this row submitted ALONE in its own call with the Apple TV 4K row sent in a separate call rather than batched.
[Modelling note — do not "fix" this] allDay rows store endAt exclusive, so the single day October 30 is sent as the bare date 2026-10-30 and reads back as 2026-10-31. Omitting endAt on an allDay row collapses it to the start date, and omitting sourceUrls clears them — always send both.
週三
28日
10月
M6 iMac and 14-inch MacBook Pro: Orders Start
2026年10月28日 · 全天
United States (apple.com and the Apple Store app)
Pre-orders open Wednesday, October 28, 2026 for BOTH Macs Apple announced on October 27, 2026: the new iMac with M6 and the new 14-inch MacBook Pro with M6. Apple publishes no clock time for either pre-order opening -- unlike the iPhone rows, which quote 5 a.m. PT -- so this is an all-day row and NO time was invented.
[Verbatim, from the two press releases' "Pricing and Availability" sections]
iMac: "Customers can pre-order the new iMac with M6 starting tomorrow, October 28"
MBP: "Customers can pre-order the new 14-inch MacBook Pro with M6 starting tomorrow, October 28, on apple.com/store..." (reader truncates at ~125 chars)
Prices, iMac: "iMac with M6 featuring a 10-core CPU and 10-core GPU starts at $1,499 (U.S.) and $1,449 (U.S.) with education savings." and "iMac with a 12-core CPU and 12-core GPU starts at $1,699 (U.S.) and $1,599 (U.S.) with education savings."
Prices, MBP: "The 14-inch MacBook Pro with M6 starts at $1,999 (U.S.) and $1,899 (U.S.)"
Body date set for BOTH releases: {October 27 (dateline), October 28 (pre-order), November 6 (availability)} -- nothing else. No strikethrough, no "Update:", no "Correction" in either body.
[2026-10-28 12:4x -- NEW ROW. The rail method is now FOUR-FOR-FOUR.]
The "More from Apple Newsroom" rail at the foot of the Apple TV 4K article rotated COMPLETELY since the previous round and now carries three press releases all dated **October 27, 2026**, none of which this calendar had ever seen:
1. apple-unveils-new-imac-with-m6 (THIS ROW + its launch-day sibling)
2. apple-introduces-the-new-14-inch-macbook-pro-with-m6 (same two rows)
3. apple-creator-studio-adds-powerful-new-ai-features-and-even-more-creative-tools -- **NOT FETCHED THIS ROUND, for budget. That is "not checked", NOT "not applicable".** Next round must open it: if it carries a dated release it belongs here; if it only says "available today" or names a season, it is declined under the standing undated ruling (same as the Siri AI language release of Oct 26).
The October 26 iPad mini and Siri releases have ALREADY rotated out of the rail after a single round. **That is rotation, not deletion** -- the iPad mini rows created from them stand. It also means the rail is a narrow, fast-moving window: miss a round and a launch can rotate past unseen. apple.com/apple-events would have surfaced none of this; it lists EVENTS only and its latest dated entry is still June 8, 2026.
[** WHY ONE ROW FOR TWO PRODUCTS **] Both Macs were announced the same day and share BOTH dates exactly (pre-order 10/28, availability 11/06). Combining them follows this calendar's own established convention -- "Apple Mac mini and Mac Studio In-Store Launch", "iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin", "Apple Watch Series 12 and Apple Watch Ultra 4 In-Store Availability" -- and it halves the collision surface: two new rows instead of four on a calendar documented as a collision factory.
[** TITLE CHOSEN AGAINST ALL 14 EXISTING TITLES PLUS THE SIBLING CREATED THIS SAME ROUND **]
Bash probe #75 failed (E2BIG, 153.8KB injected), so these are HAND calculations, not difflib -- re-measure when Bash returns.
**Two-way containment: FALSE against all 15.** That is the decisive test here: the 2026-10-11 Times Square row was destroyed at a ratio of only ~0.67 *because* the new title contained an existing one as a prefix.
**The acute hazard is "iPad mini (A20 Pro) Pre-Orders Open", which shares this row's exact startAt 2026-10-28.** The obvious name "iMac and MacBook Pro Pre-Orders Open" would have kept the whole "Pre-Orders Open" skeleton against a row created only one round earlier -- estimated ~0.56 and pointed straight at it. Dropping "Pre-Orders" for "Orders Start" brings the worst estimate to roughly 0.40.
Other worst estimates: ~0.42 against "Apple Mac mini and Mac Studio In-Store Launch" (shares " and ", "Mac") and ~0.42 against "iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin" (shares " and ", "Pro", "Orders").
**The sibling row was deliberately NOT named "M6 iMac and 14-inch MacBook Pro Arrive"** -- that would have shared a 31-character prefix with this title, estimated ~0.77, squarely in the band that destroyed the Times Square row. It is called "New M6 Macs Reach U.S. Customers" instead, bringing the pair to roughly 0.40.
**Collision skeletons on this calendar, now eight deep:** "In-Store Availability" x4, "Goes On Sale" x2, "Pre-Orders Begin" x2, "Pre-Orders Open" x1, "Reaches Stores" x1, "In-Store Launch" x1, "Launch Day in the U.S." x1, and as of this round "Orders Start" x1. A further launch row needs yet another verb.
Submitted ALONE in its own call with partial:true and read back afterwards: created:1 with eventCount 14 -> 15 rules out a silent merge (a merge leaves eventCount flat).
[Modelling note -- do not "fix" this] allDay rows store endAt exclusive, so the single day October 28 is sent as the bare date 2026-10-28 and reads back as 2026-10-29. Omitting endAt on an allDay row collapses it to the start date; omitting sourceUrls clears them -- always send both.
週三
28日
10月
iPad mini (A20 Pro) Pre-Orders Open
2026年10月28日 · 全天
United States (apple.com and the Apple Store app)
Pre-orders for the updated iPad mini with the A20 Pro chip open on Wednesday, October 28, 2026. Apple's press release of October 26, 2026 states under "Pricing and Availability", verbatim: "Customers can pre-order iPad mini with A20 Pro starting Wednesday, October 28". Price, verbatim: "iPad mini starts at $699 (U.S.) or $58.25 (U.S.) per month for 12 months for the Wi-Fi model". Apple publishes no clock time for this pre-order opening - unlike the iPhone rows, which quote 5 a.m. PT - so this is recorded as an all-day event and NO time was invented.
[2026-10-27 11:4x - NEW ROW, found by the rail method on its third outing]
The "More from Apple Newsroom" rail at the foot of the Apple TV 4K article carried three tiles this round, two of them press releases dated **October 26, 2026** that this calendar had never seen:
- apple-gives-ipad-mini-its-biggest-update-ever (THIS ROW and its launch-day sibling)
- siri-ai-is-available-in-more-languages-starting-today (deliberately not recorded - see below)
- iphone-duo-arrives-at-apple-store-locations-worldwide (Oct 23, launch-day photo essay, already resolved)
**The rail is confirmed again as the working discovery channel.** apple.com/apple-events would not have surfaced either: it lists EVENTS only, and its latest dated entry is still June 8, 2026.
[The body date set] October 26 (dateline), October 28 (pre-orders, THIS ROW), November 6 (availability, the sibling row). The only other date is "September 2026" in footnote 3 - a month with no day, not recordable. Asked directly: no strikethrough, no "Update:", no "Correction" in the article body. The word "UPDATE" on the page belongs to a rail tile for an unrelated Formula 1 article, which is the documented trap - **always cut the text at "More from Apple Newsroom" before drawing any conclusion from a date set or an edit marker.**
[Siri AI release - deliberately NOT recorded, stated rather than glossed] The October 26 Siri press release says "With the 27.2 software updates released today", i.e. 2026-10-26, which is already past, and its only forward-looking statement is undated: "Language support for Siri AI will continue to expand, with availability in Chinese (traditional), Norwegian, Turkish, and Vietnamese coming by the end of the year." **"By the end of the year" is not a date.** This matches the standing ruling that declined the September 14 Siri beta rollout. If you would like point releases such as 27.2 tracked here, say so and I will add them going forward.
[CHOSEN AGAINST ALL 12 EXISTING TITLES PLUS THE SIBLING ROW CREATED THIS SAME ROUND]
Bash probe #48 failed (E2BIG, 143.7KB injected), so **these are hand calculations, not difflib output** - re-measure when Bash returns. The full title list was enumerated first with an unbounded-lower-bound read (limit 8 returned the 8 past rows; the 4 future rows came from the round's main read), so all 12 were compared, not a guess:
Apple Special Event: Surprise and shine / iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin / iOS 27 Public Release / iPhone 18 Pro and iPhone 18 Pro Max In-Store Availability / Apple Watch Series 12 and Apple Watch Ultra 4 In-Store Availability / AirPods 5 In-Store Availability / Apple Mac mini and Mac Studio In-Store Launch / iPhone Duo Pre-Orders Begin / iPhone Duo In-Store Availability / HomePod mini 2 Goes On Sale / Apple TV 4K (4th generation) Goes On Sale / Apple HomeView Reaches Stores
**Two-way containment: FALSE against every one of them.** That is the decisive test on this project - the 2026-10-11 Times Square row was destroyed at a ratio of only ~0.67 *because* the new title contained an existing one as a prefix.
Worst estimated ratios: ~0.58 against "iPhone Duo Pre-Orders Begin" and ~0.51 against "iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin". The shared "Pre-Orders" skeleton is unavoidable if Apple's own wording is kept, and the existing pair "iPhone Duo Pre-Orders Begin" / "iPhone Duo In-Store Availability" sits at a comparable ~0.51 and has coexisted without incident, so a same-product pair in this band is the established, survived shape on this calendar.
**"(A20 Pro)" was added specifically to break the prefix**, and the sibling row was deliberately NOT named "iPad mini (A20 Pro) Arrives in Stores": that would have shared a 20-character prefix with this title and scored ~0.69, squarely in the band that destroyed the Times Square row. It is called "iPad mini Launch Day in the U.S." instead, which brings the pair down to ~0.45.
**Note the collision skeletons on this calendar: four rows end in "In-Store Availability", two in "Goes On Sale", one in "Reaches Stores", one in "In-Store Launch". A new launch-day row must use a verb none of them uses - which is why the sibling avoids all four.**
Submitted ALONE in its own call with partial:true, with a read-back afterwards.
[Modelling note - do not "fix" this] allDay rows store endAt exclusive, so the single day October 28 is sent as the bare date 2026-10-28 and reads back as 2026-10-29. Omitting endAt on an allDay row collapses it to the start date, and omitting sourceUrls clears them - always send both.
週五
23日
10月
iPhone Duo In-Store Availability
2026年10月23日 · 全天
United States (Apple Store locations and apple.com)
iPhone Duo goes on sale on Friday, October 23, 2026. A further 28 countries and regions follow on Friday, October 30, 2026 - outside this calendar's U.S. scope, so no separate entry is recorded. iPhone Duo ships with iOS 27.1. Apple publishes no opening clock time for launch day, so this stays an all-day event.
[Re-verified 2026-09-22] Re-fetched Apple's newsroom release (HTTP 200, 342,058 bytes) and matched the sentence word for word again: "Pre-orders begin Friday, October 16, with availability beginning Friday, October 23." apple.com/apple-events (HTTP 200) still lists no upcoming event, only the September 9 keynote recap, so no date has moved. No field needed changing. (The full method note, including why apple.com/newsroom/ cannot be used to detect new announcements, is on the "iPhone Duo Pre-Orders Begin" row.)
週五
16日
10月
iPhone Duo Pre-Orders Begin
2026年10月16日 晚上8:00 · Asia/Taipei
United States (apple.com and the Apple Store app)
Pre-orders for iPhone Duo, Apple's first foldable iPhone, open at 5 a.m. PT on Friday, October 16, 2026, in more than 70 countries and regions including the U.S. iPhone Duo starts at $1,999 (U.S.) in 256GB, 512GB, 1TB and 2TB, in star white and night sky; the iPhone Duo Case ($79) and iPhone Duo Folio with Kickstand ($129) can also be ordered from October 16.
[⭐⭐ Re-verified 2026-10-13 11:2x — FIRST ROUND VERIFIED BY RAW-BYTE PARSING RATHER THAN A SUMMARISING READER. Pre-orders open in THREE days; EIGHTH consecutive identical round, zero fields changed.] Bash is working, so this round fetched both pages with curl and matched strings against the bytes instead of asking a reader. Newsroom release: HTTP 200, 342,377 bytes, 87,046 characters after tag-stripping. All five governing sentences matched VERBATIM as contiguous substrings of the stripped text:
• "Pre-orders begin Friday, October 16, with availability beginning Friday, October 23."
• "beginning at 5 a.m. PT on Friday, October 16, with availability beginning Friday, October 23"
• "available in 28 other countries and regions beginning Friday, October 30"
• "iPhone Duo Case and iPhone Duo Folio with Kickstand will be available to order beginning Friday, October 16"
• "iPhone Duo will be available with iOS 27.1"
The complete set of October dates anywhere in the article text is exactly {October 7, October 12, October 16, October 23, October 30} — no stray or revised date. "$1,999", "star white" and "night sky" all present.
[⭐ THE STRIKETHROUGH QUESTION IS NOW SETTLED, not merely "weak evidence"] Earlier rounds could only report a reader's answer that nothing was struck through, and correctly flagged that a plain-text extraction CANNOT prove the absence of strikethrough formatting. This round counted the markers in the RAW HTML directly: `<del` 0, `<s>` 0, `text-decoration:line-through` 0, `Update:` 0, `Correction` 0. **No date in this release is struck through, revised or corrected.** That earlier caveat can be retired for as long as these counts stay at zero — but re-run the count, do not carry this forward blind.
[⭐ The six "UPDATE" labels are now PROVEN to be in the recommendation rail, not the article] Previous rounds asserted this; this round located them. All six occur after the heading "More from Apple Newsroom" and belong to tiles for unrelated articles (App Store Indie Games Week, Apple × The King's Trust, Apple's family digital-habits guide). Do not misread them as an edit to this release.
[apple.com/apple-events RE-READ THIS ROUND — the gap left by the last two rounds is closed] HTTP 200, 155,984 bytes, 18,302 characters of text. Every date-shaped string on the page was extracted mechanically: June 10 2024; May 7 2024; September 9 2024; June 9 2025; September 9 2025; June 8 2026. Filtered against today's date, the count of FUTURE dated entries is **ZERO**. So Apple has announced no further event, and no date on this calendar has moved. This is a mechanical absence claim over the full page text, which is stronger than the reader's "no" the previous rounds relied on.
[Note on the events page, so a later round does not misread it] apple.com/apple-events still says "Video replay of today's event will be available shortly" even though the keynote was September 9, 2026. That is stale page copy, NOT evidence of a new event. Judge "is there an upcoming Apple event" by whether a DATED upcoming entry exists, never by the word "today". The featured item at the top carries NO date at all (it is the September 9, 2026 keynote), which is why the archive list appears to jump from WWDC 2026 back to September 2025 — that gap is not a missing entry. The footnotes carry date references ("June 2026", "July and August 2026") describing testing and data collection — those are NOT events, do not create rows from them.
[No iOS 27.1 row is possible] The release still gives NO release date for iOS 27.1, saying only "iPhone Duo will be available with iOS 27.1". The September 14 date belongs to iOS 27 and Apple Intelligence, not 27.1.
[⭐⭐ CORRECTED MERGE RULE — 0.808 IS NOT A SAFE FLOOR] On 2026-10-11 a new row on the Times Square calendar silently merged into an existing row at an estimated similarity of only ~0.67, renaming it and moving its date, with no error. The cause was that the proposed title CONTAINED an existing title as a prefix. **Rule: if a proposed title contains an existing title as a prefix or substring (or vice versa), assume a merge WILL happen regardless of the ratio.** One risky write per call, always followed by a read-back.
[Why THIS row carries the sync stamp] "iPhone Duo In-Store Availability" scores difflib 0.825 against "AirPods 5 In-Store Availability" (measured 2026-10-12 with difflib itself, not eyeballed) — above the 0.808 at which submit_events was measured to silently overwrite a different row — so it is refreshed with keep_events only (kept 1, unmatched empty) and never submitted. This row's worst match is 0.633 ("iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin"), a safe margin. The rows sharing startAt 2026-09-18 (Apple Watch Series 12 / Ultra 4, AirPods 5, iPhone 18 Pro In-Store Availability) are past events whose own press releases were not re-fetched, so they are not vouched for either.
**⚠ STAMP SUCCESSION, NOW THREE DAYS OUT: this row can carry the stamp only through a round running on 2026-10-16 BEFORE 12:00Z.** It is a TIMED row with no endAt, so the cutoff is its 5 a.m. PT start = 12:00Z. The hourly rounds land a few minutes past the hour, so a 2026-10-16 round at 10:2x or 11:2x UTC still makes it. After that the stamp must move to "iPhone Duo In-Store Availability" (October 23), and because of the 0.825 collision that move must be a SINGLE-ROW submit_events call with nothing else in it, followed immediately by a get_calendar read-back confirming the AirPods 5 row is intact. Re-measure with difflib before that write. **After October 23 passes this calendar has NO future-dated row at all unless Apple announces something new; at that point report honestly that it cannot be stamped rather than writing onto a past row.**
[Past-row rule, refined] Only a FULLY PAST row has its submit_events payload silently discarded; a row still under way (startAt past, endAt today) is writable — measured 2026-10-11. Even a discarded submit still advances lastSyncedAt, so a fully-past row can still hold the scheduler off; it just cannot carry a field change. A past-dated row CAN also be moved forward by a title-identical submit. For this calendar the distinction barely helps: this row is TIMED with no endAt, so it is "fully past" the instant 2026-10-16T12:00Z passes.
[Mac mini / Mac Studio, resolved but unwritable — unchanged] The August Mac mini newsroom release states VERBATIM: "The new Mac mini is available for pre-order starting today, with availability beginning September 22." That is a direct source for the September 22, 2026 U.S. availability on the "Apple Mac mini and Mac Studio In-Store Launch" row. Caveats: that release covers the Mac mini ONLY, so the Mac Studio half of that row's title still rests on a separate release not re-fetched; and it says nothing about the 512GB-unified-memory Mac Studio configuration, whose "later, in late October" wording names no day and remains unrecordable. The confirmation CANNOT be written onto that row — its startAt is fully past. **Raising it from 0.9 to 0.95 has to be done from the UI.**
[Still deliberately NOT recorded, unchanged reasons] The 28-country second wave on October 30 is outside this calendar's U.S. scope. The September 14, 2026 Apple Intelligence / Siri AI beta rollout is not recorded: it is past, a fully-past row's payload would be discarded anyway, and a staged beta rollout is not the kind of dated software RELEASE this calendar's query asks for.
[METHOD LIMITS, still true] apple.com/newsroom/ itself is USELESS for spotting newer announcements — the index is client-rendered React, its raw HTML contains zero /newsroom/2026/... article links and zero date strings. Any "no new Apple announcement" claim must rest on apple.com/apple-events, which IS server-rendered. When a sentence needs a home, fetch the individual article URL directly — article pages ARE readable, only the index is not.
[BYTE-SIZE WARNING, still true and reconfirmed] The events page has now been seen at 155,908 / 155,984 / 155,908 / 155,984 bytes and the Duo release at 342,711 / 342,712 / 342,380 / 342,058 / 342,377 across rounds with identical content. These are per-request nonces. **Do NOT treat byte-size equality or inequality on apple.com as a signal** — match the sentences. A sha256 of the TAG-STRIPPED text would be a real signal where raw bytes are not; this round's stripped length was 87,046 characters for the Duo release and 18,302 for the events page — use those as the baseline.
[Parser trap, still true] The price sentence does NOT match as a contiguous string: the raw HTML reads `starts at <strong>$1,999</strong> (U.S.)`, so a plain substring check for "starts at $1,999 (U.S.)" returns ZERO hits and looks exactly like "Apple changed the price." It has not. Strip tags before comparing, or check for the bare number.
[Modelling note — do not "fix" this] startAt is written as local Pacific time 2026-10-16T05:00:00 with timezone America/Los_Angeles, which timefeed stores as 2026-10-16T12:00:00Z (October 16 is PDT, UTC-7). Never paste the stored Z value back in as startAt. This row has NO endAt and must not be given one: Apple publishes only the pre-order opening moment, and on a TIMED row omitting endAt preserves the absence rather than resetting anything (unlike an allDay row, where omitting endAt collapses it to the start date).
週二
22日
9月
Apple Mac mini and Mac Studio In-Store Launch
2026年9月22日 · 全天
United States
Availability date for the updated Mac mini (M6 and M5 Pro) and Mac Studio (M5 Max and M5 Ultra), announced by Apple on August 25, 2026 with pre-orders opening the same day. Note: Apple has said the Mac Studio configuration with 512GB of unified memory arrives later, in late October, with no specific date given.
[Re-checked 2026-09-22, launch day] apple.com/apple-events was re-fetched (HTTP 200, 155,908 bytes) and still shows only the September 9 keynote recap with no upcoming event listed, so Apple has announced nothing that changes this availability date. Being honest about the limits of that check: the two August press releases behind this row were NOT re-fetched this round, and apple.com/newsroom/ turned out to be client-rendered and therefore unusable for spotting a newer announcement (see the "iPhone Duo Pre-Orders Begin" row for the details). Confidence therefore stays at 0.9 rather than rising.
週五
18日
9月
iPhone 18 Pro and iPhone 18 Pro Max In-Store Availability
2026年9月18日 · 全天
United States (Apple Store locations and apple.com)
iPhone 18 Pro and iPhone 18 Pro Max go on sale today, Friday, September 18, 2026. Apple publishes no opening clock time for launch day, so this stays an all-day event.
[Re-verified 2026-09-18, launch day] Re-fetched Apple's own press release and matched the sentence word for word: "Pre-orders begin Saturday, September 12, with availability beginning Friday, September 18." apple.com/apple-events was also re-fetched: it lists no upcoming event (its event archive now shows WWDC June 8, 2026 and the September 2025 and 2024 keynotes), so no Apple date has moved and no new launch date has been announced. The AirPods 5 and Apple Watch Series 12 / Ultra 4 rows share this same September 18 in-store date, but their own press releases were not re-fetched this round.
週五
18日
9月
Apple Watch Series 12 and Apple Watch Ultra 4 In-Store Availability
2026年9月18日 · 全天
United States (Apple Store locations and apple.com)
Apple Watch Series 12 and Apple Watch Ultra 4 reach stores on Friday, September 18, 2026. Both press releases state that customers in the U.S. and more than 50 other countries and regions could pre-order on September 9, "with availability in stores beginning Friday, September 18." Apple Watch Series 12 starts at $399 (U.S.) and Apple Watch Ultra 4 at $799 (U.S.); new Apple Watch bands reach stores the same day. Recorded 2026-09-11 from Apple's own press releases.
週五
18日
9月
AirPods 5 In-Store Availability
2026年9月18日 · 全天
United States (Apple Store locations and apple.com)
AirPods 5 reach stores on Friday, September 18, 2026. Apple's press release states that customers in the U.S. and more than 65 other countries and regions could pre-order on September 9, "with availability in stores beginning Friday, September 18." AirPods 5 are $129 (U.S.); AirPods 5 with Wireless Charging Case are $149 (U.S.). Recorded 2026-09-11 from Apple's own press release.
週一
14日
9月
iOS 27 Public Release
2026/9/14至2026/9/15 · 全天
United States
iOS 27 becomes available as a free software update on Monday, September 14, 2026. Apple's iPhone 18 Pro press release states verbatim: "iOS 27 will be available as a free software update on Monday, September 14." watchOS 27 ships the same day for Apple Watch Series 9 or later, Apple Watch SE 3 and Apple Watch Ultra 2 or later. Apple Intelligence arrives with iOS 27 and watchOS 27 on the same date, and Siri AI begins rolling out in beta that day in English. Apple publishes no clock time for software releases, so this is recorded as an all-day event. Re-checked 2026-09-13: apple.com/apple-events lists no newer event and no change to this date.
週六
12日
9月
iPhone 18 Pro and iPhone 18 Pro Max Pre-Orders Begin
2026年9月12日 晚上8:00 · Asia/Taipei
United States (apple.com and the Apple Store app)
Pre-orders for iPhone 18 Pro and iPhone 18 Pro Max open at 5 a.m. PT on Saturday, September 12, 2026, in more than 65 countries and regions including the U.S. Apple's newsroom release states: "Pre-orders begin Saturday, September 12, with availability beginning Friday, September 18." iPhone 18 Pro starts at $1,199 (U.S.) and iPhone 18 Pro Max at $1,299 (U.S.), in 256GB, 512GB, 1TB and 2TB, in burgundy, glacier, silver and black. Announced at the September 9, 2026 Apple Event and recorded here 2026-09-11 from Apple's own press release.
週四
10日
9月
Apple Special Event: Surprise and shine
2026年9月10日 凌晨1:00 · Asia/Taipei
Apple Park, Cupertino, CA - streamed at apple.com and in the Apple TV app
Apple's fall special event keynote, titled 'Surprise and shine'. Confirmed on apple.com/apple-events: September 9 at 10 a.m. PT, viewable at apple.com or in the Apple TV app. Re-checked 2026-09-09 (about 25 minutes before the keynote): the page now reads "Watch a special Apple Event today at 10 a.m. PT" and still lists no other upcoming event. Apple has published no product list; press expectations (not Apple announcements) point to the iPhone 18 Pro lineup, a first foldable iPhone, Apple Watch Series 12, Apple Watch Ultra 4 and next-generation AirPods. Pre-order and release dates for anything shown are announced at the event itself, so none can be recorded here yet - they should appear in this calendar from the next refresh after the keynote.