Site Journal

What we changed, when, and why. Newest first.

We keep this page because a guide site that quietly edits itself is hard to trust. If a number on this site changed, it should be possible to find out when and on what evidence. Corrections get their own line, not a silent edit.

2026-09-06

Four depth records moved and nobody reached floor 1000. B2 is broadcasting from floor 986, which settles his previous stream at 986 and leaves the deepest blaster run with him. おまめブラスターch reached 962: his floor-950 session had been unsettled for eight days because everything he streamed in between was Salmon Run. WispariaN posted "Floors 941-958", contiguous with his previous upload, and holds the deepest run outside Japan. Aoi Rin moved from 953 to 956. MistRain has now opened three broadcasts from floor 950 without gaining a floor, which is eight hours and twenty-six minutes on one floor.

The two-player speedrun record we published on September 4 was wrong by the time we published it. LonTr0 and Dee_G ran both 2:44:03 and 2:37:05 on September 3. The slower one was verified first and we recorded it as the record; the faster one verified on September 5 and speedrun.com now hides the 2:44:03 entirely, because the board shows only a team's best run.

That hiding rule caused a second error we have now fixed. We wrote that the old four-player 3:30:52 "is now third". It is not on the ranked board at all, because the same four players set the 3:14:05 record. Every run count on the speedrun page now says explicitly that it counts verified runs, which is more than the board displays. Six boards on that page were affected.

The solo record is unchanged at 2:18:27, but two runs verified on September 5 cut Haitex's lead over second place from 5 minutes 32 seconds to 2 minutes 2 seconds. The main board has 1,305 verified runs by 266 runners.

We added an answer to a question the speedrun page had never answered: why co-op records are slower than solo ones. The Power Egg requirement rises with the player count, which the four-player team who hold the Gold Club time attack state in their own video description. The Gold Club is the only level with records at more than one player count, and it reads 2:25 solo, 2:49 for three players and 2:59 for four.

Game8 Japan edited its Bottomless Buffet guide again at 05:33 JST on September 6, the seventh edit since the first clear. It still does not state a floor cap, and the number 1000 appears zero times on the page.

2026-09-04

B2 reached floor 979 of the Bottomless Buffet with a blaster. His public stream starts at floor 977, and his own description records a loss on floor 979. MistRain reached floor 950 with an Engine Roller; the next stream title settles the previous floor-941 stream at 950. We found no fourth floor-1000 clear. All 42 YouTube video links listed on the records page returned 200 in the first full post-publication link check.

The speedrun.com four-player record is now 3:14:05. The Japanese 3:16:42 is second, and both were verified on September 3. The two-player record changed to 2:44:03; a 2:37:05 by the same pair is awaiting verification. The main board has 1,284 verified runs by 266 runners.

We corrected two errors on the speedrun page. Its category rows were still showing the September 2 counts after the total had been updated on September 3, and several current categories were missing. Its video thresholds were one hour too high for Raider and Survivalist. The correct thresholds are 4:30 solo and 5:30 multiplayer for Raider, and 6:00 solo and 7:00 multiplayer for Survivalist.

Game8 Japan edited its Bottomless Buffet guide again at 05:27 JST on September 4. It still does not state a floor cap, and the number 1000 appears zero times on the page.

2026-08-20

Floor 1000 of the Bottomless Buffet was cleared. The Japanese streamer 新月 cleared it at about 12:14 JST, which is the first documented clear by anyone. He had reached the floor the previous evening and spent three broadcasts trying to beat it.

We changed the depth records page, the home page answer, the trivia entry and the Buffet guide to say so, and we split the claim in two on purpose. That the floor was cleared is well evidenced: it is his own stream, two viewers give the same timestamp for the winning attempt, and the broadcast ends thirteen minutes later. That floor 1000 is the LAST floor is not confirmed. Viewers report a completion check mark and no floor 1001 after it, and the Japanese community spent the day calling the mode 有限食堂, the "Finite Dining Hall", instead of its real name. That is a lot of people agreeing about a screen we have not seen. We will call it confirmed when there is a wiki entry, a second clear, or footage of the screen itself.

One thing we got wrong yesterday. The home page answer about Buffet depth was rewritten on the 19th specifically so it would not go stale, by taking the numbers out of it. It said instead that nobody had cleared floor 1000, which stopped being true a day later. Leaving the digits out did not make the claim durable, because the fragile part was never the number.

2026-08-19

Published the weapon farming guide. The useful part is arithmetic: the game has 60 base weapons split unevenly across 11 types, so the share of drops a type gets is just its share of the roster. A Japanese player measured, over 200 recorded runs, that carrying a weapon of the type you want raises that type to about 15% of drops. That is worth roughly three times as much to a Splatana user as to a Shooter user, who is already at 15%. We labelled the arithmetic and the community test separately, because they are not equally solid.

Rebuilt the home page. The guide list on it had been written by hand and had gone stale: it showed 10 of 41 guides and had two cards both pointing at the tier list. It now reads the same guide list the rest of the site reads.

Rewrote the home page FAQ. Three of the four questions were about us rather than about the game. One of them told readers to send a pull request on GitHub, which was never possible. The questions are now ones people actually ask.

Corrected the About page. It said our data was verified through our own in-game testing. That is not what we do and never was. We read published sources, name them, and say how strong each one is.

Fixed a dungeon name on the weapons page. It said "Annan Lair"; the English release calls it Awlmun Den. We had published that correction on 2 August and missed this page.

2026-08-18

Published the speedrun guide, built from speedrun.com's verified runs. The solo record stands at 2:22:55 and had fallen 17 times in 24 days.

Bottomless Buffet record moved to floor 995.

Splatoon 3 is giving Raiders finishers a weapon. The official English name is Plastic-Bottle Shot Replica. Several outlets called it a Splattershot Replica.

2026-08-16

Corrected how we read Japanese climb streams, which had made us under-report the world record twice.

A stream titled with a floor number is telling you where it starts, not where it finished. What a stream actually reached only shows up in the next upload's starting number. We had been recording those starting numbers as results. We also built an argument on top of the mistake, calling a gap in someone's uploads a sign they had hit a wall. He had not, and he had said publicly why he paused. Both the numbers and the argument are retracted on the page itself.

2026-08-15

Published the gadget parts guide. Replaced four drifted copies of the guide menu with one, which turned up an orphaned page and a bug that sent every non-English header click to the English site.

2026-08-11

Corrected the deep floor records, badly and publicly. We had published floor 802 as the ceiling and said floor 900 was attested nowhere. A floor-917 video was already public on the day we wrote that.

The cause was a gap in how we searched, not a gap in the evidence. We had swept four communities and skipped the Japanese YouTube channels where the records actually get set. The page says all of this in its own text.

2026-08-05

Checked the whole site against Google's advertising policies before applying. Found and fixed five problems of our own making, including admin pages that search engines could index and a privacy policy that described advertising we were not running.

2026-08-02

Corrected six Traditional Chinese terms that we had invented rather than sourced. Nintendo publishes its own Chinese names in the Traditional Chinese edition of its developer interview, and ours did not match. 146 replacements across 17 files.

Found that our custom fonts had never loaded in production, and that the build was leaking a local file path into every page. Pages went from 1.19MB to 126KB.

2026-07-22

Site launched. Before anything else we removed the placeholder content written before the game was out, including a two-player split-screen mode that does not exist. That set the rule the site still runs on: if we cannot point at a source, we do not publish it.