SI-generated
Speedrunning had a distinctly archival mood this week. The big conversation was not only about who was fastest, but about whether the record books will still be readable, searchable, and useful a year from now. A change to speedrun.com’s terms of service sent runners, moderators, and historians into a familiar online ritual: first confusion, then alarm, then the quiet realization that somebody had better start downloading everything.
That anxiety produced one of the week’s most heartening stories, as community members began building public backups instead of waiting for a perfect solution. Meanwhile, the actual runs kept arriving. Banjo-Kazooie may have found a way around one of its oldest barriers, Bubzia turned blindfolded Mario 64 into a three-and-a-half-hour laboratory, and Majora’s Mask got both a new 100% record and a reminder that its “first cycle” is still not done giving up secrets.
There were also smaller, wonderfully speedrunner-shaped triumphs: a first Sub-1:50 in Smash Bros. Brawl, a one-frame brush with a TAS in Super Mario Bros., and Switch 2 level runs where three seconds can separate a personal best from the top of the table. The week’s vibe was equal parts preservation drive, technical breakthrough, and cheerful refusal to stop optimizing old games.
Here are the biggest stories from the week.
The Record Book Goes Offline, So the Community Backs It Up
The week’s top post was a link to an Internet Archive snapshot of speedrun.com, and its popularity said everything about the moment. It was not a new world record or a flashy route discovery. It was a 2026 archive of the site’s data, offered at exactly the point when runners were wondering how much of that data they could continue to trust and access.
The response was immediate and practical. One of the highest-rated comments pointed community moderators toward the downloadable publishing archive, explaining that the files could be unpacked to retrieve individual leaderboards. That is the kind of information that rarely goes viral outside speedrunning, but here it landed as a public service announcement. The archive was not merely a backup in the abstract; it was a toolkit for people who need a particular game, category, or run history right now.
Another thread followed a runner who decided to archive pre-terms-of-service-change data through the official API. The project reached its satisfying “done” update, complete with a public archive, a live dashboard, and even a terminal stream accompanied by rainwave radio. It is difficult to imagine a more speedrunner response to institutional uncertainty: automate the download, publish the result, then make the process entertaining enough that other people can watch the preservation happen.
The community’s comments also supplied a useful dose of realism. One contributor warned that the API cannot retrieve runs beyond the 10,000th result in a query, a limit that makes “just archive everything” much less simple for popular games. Others were already thinking about how a clean dataset could help series move their leaderboards elsewhere. Preservation is not glamorous work, but this week made its stakes visible: a leaderboard is not just a list of times. It is the history that lets runners understand what those times mean.
The ToS Problem Is Really a Trust Problem
Several of the week’s most discussed posts circled the same question: what happens when the central site that stores a community’s results changes the rules around its data? The most alarming phrasing described post-change runs as “hostages,” while another thread asked whether runners were simply going to accept the terms or start building a genuine alternative.
The details are complicated enough to resist a neat villain-and-rebellion narrative. Community moderators have their own policies, developers have licensing constraints, and a new leaderboard service cannot conjure a mature database, verification workflow, and familiar user experience overnight. One discussion noted that some communities are considering a two-stage approach: decide whether to stop verifying on the existing site, then work out how to host their own records. That is less dramatic than a mass exodus, but considerably more likely to produce something usable.
There is a deeper issue underneath the licensing arguments. Runners have spent years treating leaderboards as public infrastructure, even though infrastructure always has an owner, a cost, and a policy layer. A record can be earned by one person, verified by another, displayed by a third-party platform, and used by an event organizer who never thinks about the chain connecting those pieces. When the rules change, all of those invisible dependencies suddenly become visible.
The encouraging part is that the discussion has already moved beyond outrage. People are comparing archive formats, considering mirrors, testing emerging services, and asking what a decentralized or community-owned record book would need to do well. The speedrunning world may not end up with one replacement for SRC. It may end up with several specialized ones, connected by open exports and shared history. That would be messier, but perhaps healthier.
Banjo-Kazooie Finally Opens the 810 Note Door
Against that administrative backdrop came the kind of discovery that reminds everyone why these games keep being run decades after release. Banjo-Kazooie’s 810 Note Door skip is now possible to perform in real time, and the projected reward is enormous: roughly 25 to 30 minutes in an optimized run.
The phrase “bottlenecked Any% for decades” is not hyperbole here. A major barrier in an old game can become part of its identity. Runners route around it, commentators explain it, viewers learn to recognize the long section that everyone must endure. Then someone finds a way through, and the map of the category changes all at once. A skip measured in seconds is a nice improvement; a skip measured in half an hour is a new era.
The comments had the appropriate combination of amazement and context-seeking. One viewer asked how the discovery was made and why it had taken so long, which is exactly the right question for a technique that looks obvious only after it exists. “Possible” is also doing a lot of work in the announcement. A route-breaking trick can be known, demonstrated, and still require a long period of setup practice before runners can trust it in a serious attempt.
For Banjo runners, the next phase will be fascinating. The first successful demonstrations establish the ceiling; the grind determines whether the category can live there. Expect guides, setup refinements, failed attempts that teach everyone something, and a leaderboard that looks increasingly temporary. The old game did not become easier this week. It became much larger.
Bubzia Turns Blindfolded Mario 64 Into an Endurance Sport
Super Mario 64 blindfolded running already operates on a scale where “fast” becomes a slightly inadequate word. Bubzia’s new 100 Star world record, a 3:50:10, is a substantial achievement in its own right—and it is also a deliberately chosen training ground for the even more intimidating 120 Star category.
The 100 Star route has a lovely logic to it. It visits the game’s levels broadly enough to test the full language of blindfolded play, while leaving out some of the most punishing material required for 120. In other words, it is not simply “120 Star, but shorter.” It is a category with its own rhythm, one that lets a runner build consistency without spending nine hours in the game’s most hostile corners.
Bubzia’s explanation of the times makes the difficulty curve almost comically clear. Moving from a 70 Star record of around 1:16 to the 100 Star record takes another two and a half hours. The final 20 stars needed for 120 then add approximately five hours and twenty minutes. In ordinary Mario 64, 20 stars is a chunk of a run. Blindfolded, it is practically a second campaign.
The comment section also offered a useful reminder that “hard” is relative. Viewers praised the category for showing stars from every level without forcing runners through the most frustrating 120 Star objectives, while still admitting that blindfolded 100-coin stars are absurdly difficult—and often the most entertaining moments to watch. That balance is why side categories matter. They give a community another way to appreciate a game’s skill ceiling without pretending every challenge has to be folded into the main event.
Majora’s Mask Keeps Finding Time in the Clock
Majora’s Mask supplied two different kinds of headline this week. EnNopp112 reclaimed the 100% No Major Glitches world record with a 4:27:26, cutting 25 seconds from the previous mark. In a category that asks for essentially every persistent item and upgrade, 25 seconds is not a rounding error. It is a clean statement that the route still has room to tighten.
NMG is a particularly good example of how category rules shape a game’s identity. Major glitches are allowed, but Stale Reference Manipulation—the technique that can blow the game wide open—is excluded. The result is a run that remains recognizably Majora’s Mask while still demanding a highly optimized route through a huge checklist. Add the requirements of 100%, and every small movement decision is competing with the clock and the inventory.
At the same time, a new video explored how runners finally broke the first cycle. That discovery is a reminder that a game can be both heavily studied and fundamentally unsettled. Majora’s Mask’s three-day structure makes early routing unusually consequential: a small improvement in the opening cycle does not merely save time once, it can change the shape of everything that follows.
The two stories belong together. One is a polished record at the top of a demanding category; the other is a route discovery that may influence future records across multiple categories. This is what a healthy mature scene looks like: runners are still shaving individual seconds, while researchers are still asking whether the route’s assumptions were correct in the first place.
Three Seconds From the Top on Switch 2
The week’s smaller records came from Super Mario 3D World + Bowser’s Fury, now being pushed on Nintendo Switch 2. One runner posted a 15-second in-game-time run of 1-1, just three seconds off the top time. Another returned to the game after a year away and asked whether a 16-second Super Bell Hill run was any good. The answer, according to the leaderboard, was a reassuring yes: it would sit around 10th.
These posts are easy to overlook next to a 30-minute Banjo breakthrough, but they capture the everyday engine of speedrunning. A runner comes back, posts a time, asks for criticism, and receives a mixture of encouragement and concrete comparison. “You are fast” is nice. “You are currently about 10th, the record is 12, and here is what the top runs do” is actionable.
The clips also showed how much technique can be hidden inside a tiny level. Viewers asked what creates the speed, whether the strategy works with different characters, and how a run that looks like a few clean movements can sit just seconds from the theoretical limit. That curiosity is the bridge between watching a record and joining a game. The explanation is part of the competition.
And three seconds is an excellent distance from first place. It is close enough to make the target feel real, but large enough to contain a season of practice. On a level leaderboard, the gap is not a verdict. It is an invitation.
Brawl Finally Breaks the 1:50 Wall
Super Smash Bros. Brawl’s Subspace Emissary Any% category had a quieter headline but a satisfying round number attached to it: a new world record of 1:48:03, the first time the category has gone under 1:50. Milestones like this are emotionally powerful because the number is instantly legible, even to someone who does not know the route. “Sub-1:50” is a door; this run walked through it.
There were no heated comments or elaborate debate under the post, just the result itself. That silence is worth reading as confidence rather than indifference. In a small scene, the record can arrive without a long explanation because the people who understand the route already know what the barrier represented. The work happened before the headline: the failed attempts, the resets, and the refusal to let a clean time remain just out of reach.
It also rounds out a week in which speedrunning repeatedly demonstrated that progress has many sizes. Banjo found half an hour. Brawl found the final seconds needed to change a prefix. Both are breakthroughs because both alter what the community considers normal.
A One-Frame Mario Chase and the Long Tail of Competition
Super Mario Bros. runners had their own microscopic drama this week, with stem_kuki posting a 44.244-second 8-4 individual-level run—one frame away from tying the TAS. That is the kind of result that compresses an entire discipline into a decimal: human execution, a tool-assisted benchmark, and a single missing frame sitting between them.
It is tempting to treat a near-TAS result as the end of a story, but it is usually the opposite. The interesting question becomes where the frame went. Was it a setup, a cycle, a jump, or a piece of luck that did not quite repeat? At this level, analysis is not about finding a visibly bad mistake. It is about identifying the smallest divergence in a sequence that already looks impossible.
The week’s broader conversation about competitive Mario games made a fitting companion piece. Runners debated what belongs behind Super Mario 64 and the original Super Mario Bros. in the hypothetical ranking of the scene’s most competitive games. Super Mario Odyssey was a popular answer, while Super Mario World entered the discussion with its enormous modding community. There was no single metric, because “competitive” can mean player count, optimization, longevity, or the depth of the surrounding ecosystem.
That ambiguity is a feature. Speedrunning is not one ladder; it is a collection of ladders, each built for a different kind of obsession. One scene chases a frame, another builds a new route, another decides how its records should be preserved. The common thread is not a universal definition of importance. It is the willingness to keep asking what faster, fairer, clearer, or more lasting could look like.
This was a week where the record book itself became a story, but the runs never stopped being the point. Community archivists turned uncertainty into downloadable data. Banjo-Kazooie runners opened a door that had defined their route for years. Bubzia gave blindfolded Mario 64 a category with its own identity, while Majora’s Mask continued to reward both precision and research. Around them, smaller scenes pushed through round-number barriers and helped returning runners find their next target.
That combination feels especially encouraging. Speedrunning can be intensely competitive without being only about the champion. The archive project helps future runners. The constructive advice helps someone climbing a level leaderboard. The newly discovered skip creates work for an entire community. Even a one-frame miss gives the next person a question to answer.
So keep the backups, read the rules carefully, and enjoy the chaos while the old games keep revealing new shapes. The leaderboard may be infrastructure, but the curiosity behind it is much harder to shut down. See you next week.
Source: reddit.com