SI-generated
Speedrunning had one of those weeks where the phrase “new route” covered almost everything from a genuinely historic glitch to a runner figuring out how to get to work faster with fall damage. The headlines were split between enormous breakthroughs, long-running games getting shaken awake, and the less glamorous infrastructure that keeps the scene moving when nobody is on a world-record pace.
The biggest technical story came from Banjo-Kazooie, where a new 810 Note Door Skip helped produce a world-record improvement of roughly eight minutes, with an eye-watering 20–25 minutes of possible savings still hanging over the category. Elsewhere, Battle for Bikini Bottom got another reminder that its physics engine is less a rulebook than a suggestion, while Disney’s Atlantis continued to discover new ways to ignore the geometry of a PlayStation game.
There was room for the human side of the hobby, too: a major RPG event wrapped up in the Netherlands, a Mario 64 bounty kept the community cheering for its favorite contenders, and runners debated who gets to use the work that makes leaderboards playable in the first place. In other words, this was a week about speed, spectacle, and the very ordinary people who make both possible.
Here are the biggest stories from the week.
Banjo’s 810 Door Turns a Familiar Race Into a New Game
Some glitches shave seconds. Some glitches demand a new setup, a new split, and a stern conversation with every old route document. The 810 Note Door Skip in Banjo-Kazooie belongs firmly in the second category. The discovery was enough to push the N64 any% world record down by around eight minutes, and the post’s own estimate suggests that a fully optimized route could eventually save somewhere in the neighborhood of 20–25 minutes over the previous standard.
That is a remarkable amount of time to find in a game whose route has been refined for years. It also explains the particular flavor of excitement in the discussion. The runner was upfront that the record still needed improvements and optimization, but that modesty almost makes the result more dramatic: a merely “mediocre” run can now beat the old record by more than eight minutes. When the floor moves that far, everybody in the category gets a new ceiling.
The technical questions are already following. One commenter wondered whether the presence of the 810 door changes the later Dog Skip setup because of the performance hit, or whether the frame-rate impact is effectively irrelevant. That is exactly the kind of detail that turns a flashy discovery into a real route: not just can the trick happen?, but what does it do to every room, timing, and backup around it?
For viewers, this is the most satisfying kind of record chase. The leaderboard gets a new number, but the more important story is the category’s future. Banjo-Kazooie suddenly has an open research problem, a pile of unclaimed savings, and the delicious possibility that the first record after the breakthrough will itself look provisional. Somewhere, a runner is already measuring the door.
SpongeBob Finds a Wrong Warp and Loses Its Sense of Direction
Battle for Bikini Bottom also had a week for the history books, though its breakthrough arrived in a distinctly aquatic flavor. A new Throwverflow Wrong Warp helped deliver a 19:47 any% world record, a time that sounds impressive on its own and becomes much stranger when you learn what the category’s tool-assisted future looks like.
The runner emphasized that the technique works on console, which matters enormously for a community that cares about the boundary between emulator discovery and real-hardware execution. The current TAS route can apparently finish in about 15 minutes without ever leaving Jellyfish Fields, and newer strategies may lower that further. The game is not merely getting faster; it is being reinterpreted as a collection of rooms that happen to be connected by a plot.
That “the game is super busted now” assessment is not a complaint. It is practically a genre classification. Speedrunning has a special talent for turning a platformer’s most fragile assumptions into infrastructure. A throw, an overflow, and a wrong warp become a transit system. The route stops asking what the designers intended and starts asking what the console will tolerate before it sends SpongeBob somewhere completely different.
The 19:47 is therefore both a record and a progress report. Console runners have access to the same broad discovery as the TAS, while the TAS remains a map of what might eventually be possible. That gap is where a lot of the fun lives: every new real-time setup is a negotiation between theoretical perfection and the human ability to perform something absurd dozens of times in a row.
The Mario 64 Bounty Keeps the Cheese on the Chase
Not every major story needs a freshly posted record. Sometimes the drama is in watching an old rivalry refuse to become old news. This week brought another update on the Super Mario 64 120-star bounty, with the community rallying around Cheese as a contender who could make the prize race especially memorable.
The reactions were brief but unmistakably enthusiastic: “Let’s go Cheese!” sat alongside the recurring question of where the shirt was, a small bit of fan culture that says a lot about how personal these long-running races become. Supporters were not only hoping that Cheese could claim the bounty; they were also pointing to the larger legacy behind the attempt, arguing that his 120-star career already places him among the category’s greats.
That is what makes the bounty compelling. It gives the audience a clean, dramatic objective, but the real competition is measured in years of route knowledge, endurance, and the ability to return after a frustrating session. A 120-star run is not a quick demonstration of one spectacular trick. It is a long argument with the entire game, and the game gets to interrupt whenever it feels like it.
Whether or not this particular chase ends with the bounty changing hands, the category remains a model of sustained speedrunning drama. The record is important, but so is the accumulation of expertise around it. Fans are not just watching a timer; they are watching a community decide, again and again, that this particular mountain is still worth climbing.
RPG Speedrunning Takes the Show Offline
From October 22 through October 25, the RPG Speedrunner Union held its first on-site event, RPGSU Offline 2026, at the Fletcher Hotel-Resort Zutphen in the Netherlands. The event also raised €600 for ILGA-Europe, giving the gathering a purpose beyond simply putting a lot of very long games on screens in the same building.
There is a particular charm to an RPG marathon. These are not always the cleanest or shortest runs to watch, and that is precisely the point. The event’s one comment celebrated the appeal of “super longer” dog speedruns, a wonderfully honest description of the patient spectacle involved. RPG routing can be a slow reveal: a menu decision here, a sequence break there, and suddenly a game built to occupy an entire weekend is being compressed into a practiced performance.
An offline event changes the texture of that performance. Chat becomes a room, commentary becomes a conversation, and the awkward pauses that feel enormous on a stream can become part of a shared memory. The fundraiser adds another layer of community work, connecting a niche competitive hobby to a cause in a way that feels concrete rather than ornamental.
RPGSU Offline 2026 is also a useful reminder that speedrunning is not only a leaderboard activity. It is an event-making activity. The people who route, run, commentate, organize, and donate are building a scene that can exist both online and in a hotel in Zutphen. If the Union keeps this going, the first offline event may end up looking less like a one-off and more like the beginning of a tradition.
Atlantis Has Declared War on the Ledges
Disney’s Atlantis on PlayStation 1 supplied not one but two doses of “the level geometry is optional” this week. New work on ledge clipping showed that runners can pass through considerably more than just ladders, while a separate information dump focused on ladder clipping and described the newly discovered technique as potentially game-changing.
The posts have the energy of discoveries arriving faster than the documentation can keep up. One runner joked that this was the third time in two weeks they had shown new Atlantis tech in class, and hoped this would be the last for a while. That is a familiar speedrunning problem: the breakthrough is exciting enough to share immediately, but the route will need time before anyone can explain it without sounding like they are reading notes from a particularly haunted physics lecture.
There is value in the rough version, though. An “info dump” preserves the details while they are fresh: what edge to approach, what visual cue matters, which failed attempts reveal the collision model. Speedrunning’s polished guides are built from this mess. The rambling video is not the final route; it is the lab notebook that lets the next runner reproduce the experiment.
Atlantis now looks like a category in the best possible state: small enough for one person to make a huge contribution, but deep enough that each answer produces another question. If clipping through ladders was only the beginning, the next record may depend less on perfect execution than on discovering which parts of the game have been quietly pretending to be solid all along.
The Guide Is Free, but the Conversation Is Not Simple
Speedrun.com’s new terms of use prompted a different kind of discussion this week: who can reuse the guides and resources that runners publish, and under what conditions? The post drew a thoughtful response because it touched a nerve at the intersection of community knowledge, attribution, and the uneasy idea that a guide written by one person can become useful to anyone.
The community reaction was not one-note. Some commenters considered the policy unsurprising for material posted publicly, while others appreciated the warning that creators could remove guides if they wanted to protect their work. The conditions described in the discussion include credit for guide authors, resource managers, and news writers, while leaderboard data and runs are treated differently. That distinction matters: the numbers may be communal, but the explanation of how to achieve them can represent hours of individual labor.
There is a practical lesson here for runners who care about control over their writing. A guide hosted on a platform is convenient, searchable, and connected to the leaderboard, but convenience is not the same as ownership. Some commenters suggested hosting important work independently and linking to it, which is less tidy but gives the author a clearer say in how the material is used.
The larger speedrunning ecosystem depends on sharing. Nobody wants every route to become a private vault, and nobody should have to pretend that documentation appears without effort. The best outcome is not less collaboration; it is collaboration with visible credit and informed consent. A fast game still needs a paper trail, and the people who write that trail deserve to know where it can travel.
Constantine Proves Six Years Is Just a Long Setup
Among the week’s quieter world records was a new real-time attack record in Constantine, arriving after six years. The runner had previously posted the door glitch and found several other skips, then secured the RTA world record before showing the technique at a hotfix event later that night.
That sequence captures a beautiful piece of speedrunning culture: discovery and performance do not always happen in the same order. Sometimes the runner finds the break, keeps it close while the route settles, and then takes the record before the wider audience sees the full explanation. The reveal is not just “look at this trick”; it is “look at what this trick has been waiting to do.”
Six years also gives the record a different emotional weight from a weekly leaderboard shuffle. A long gap between breakthroughs can make a category appear dormant when it is actually accumulating knowledge beneath the surface. One door glitch can be the visible end of years of experimentation, failed ideas, and a runner remembering that an old game may still have one more answer hidden in it.
That is the recurring promise of this hobby: a game does not have to be new to be unfinished. Constantine’s record arrived with the confidence of someone who had already done the archaeological work. The timer only shows the final artifact.
The Fastest Story Is Still the One About Starting
For every record that redraws a leaderboard, there are hundreds of runs that end much farther down the table. One post this week showed a runner realizing they were nearly five times slower than first place and joking that the result was, indeed, bad. The replies answered with something more useful than a motivational poster: finishing the run is already an achievement, and nobody gets a world record on their first attempt.
That exchange belongs in the same weekly roundup as Banjo’s massive skip. The record side of speedrunning gets the spectacle, but the beginner side supplies the future. A personal best is a small piece of evidence that the route can be learned, and a completed run gives a runner something concrete to improve instead of an abstract fear of failure.
It also pushes back against the misleading neatness of leaderboards. “First place” is visible; the practice behind it is not. The supportive comments made that invisible work explicit: keep practicing, keep finishing, and remember that everyone who looks effortless now once had a run that felt impossibly slow.
Speedrunning is a competition, but it is an unusually porous one. The same community that celebrates a new world record can make room for someone’s first completed attempt. That is not a contradiction. It is how the category keeps replenishing itself.
This week’s stories all pointed in the same direction, even when the games and problems were wildly different. A door skip can make an old route new again. A wrong warp can turn a licensed platformer into a physics dissertation. A hotel event can make a scattered scene feel like a room full of people, while a terms-of-use debate can remind everyone that community knowledge has authors.
Most of all, the week showed that speedrunning’s real currency is possibility. There are minutes left in Banjo-Kazooie, seconds left in Constantine, and entire categories still waiting to learn what their collision models have been hiding. Even the humble first run belongs to that same story: it is the first proof that a player can enter the route and come back with a time.
So here’s to the runners measuring doors, clipping ledges, defending their guides, running RPGs in hotel rooms, and cheering for the person who just finished last. The timer keeps moving because the community keeps finding reasons to press start. See you next week.
Source: reddit.com