This thread will serve to list segmented, or otherwise notable. runs, which don't fit in the normal leaderboard format.

Jumpless Any% in 26:34.967
By apple1417.

This run started with an innocent enough question in the discord one day: "Is it possible to do a no jump run?" at the time the answer was no, but it got people thinking. A few days, and several terribly inconsistent strats later, this run was born.

Gehenna Jumpless Any% in 19:47.267
By apple1417.

Continuing on from the Jumpless Any%, the question of if Gehenna Jumpless was possible was quickly asked. Strats were found and the run was started soon after, only to die to a Goliath strat too inconsistent for even segmented. The run was restarted a year later when an alt was discovered, which still today is one of the best combination of glitches ever used in a route.

A commentated version is linked in the video description.

Auto Test Bot in 4:23:54
By Bisqwit.

The game contains a bot that automatically solves all puzzles, used for regression testing during development. This video shows it playing through the full game, showcasing all the intended solutions.

Quote


Played on public beta version [418338]. This version has slightly different behavior in auto test bot compared to previous versions: It spends longer time staring at the messenger screens, reads fewer QR codes, and ignores Alexander’s messages completely. It still reads every file and interacts with Milton.

As usual, it failed to solve some puzzles. The puzzles it failed to solve are NT2, Messenger Island 1, and ∗∗13. It cheats the solution to NT2, but forfeits the island, ∗∗13, and also ∗∗24. It also got itself killed in NJ1, but cheated itself back into life.

This time, the bot was not assisted by the player at all. After the two cheats (enable cheats, enable autotestbot) were entered, the player shut down the monitor and commuted to work and left the bot to its own devices. The recording was terminated remotely when the log indicated that the game was over.

Useless
By apple1417, EDB, Gelly, HarryPoster, hernanjnx, Ninja and Nozu

So if you've ever seen any other Talos run, you know it plays nothing like the simple puzzle game it is casually. But for almost as long as we've been breaking the game apart, we've also been asking another question: How can we do all that, but less usefully? Basically our version of SM64's A Button Challenge, in Useless we try to collect all sigils, across both the main and DLC campaigns, while pressing the Use buttons as few times as possible. And just like the ABC, it leads to a dramatically different run. Just how many presses does it take? Well you'll just have to watch the video to find out

Now Useless is segmented out of necessity. Even ignoring how any one mistake can lead to an extra press invalidating the entire run, there are dozens of different extremely difficult tricks, which each could eat up hours of attempts by themselves. Because of this, our primary focus was simply getting a presentable run, not grinding out every possible frame. We still focus on speed of course, but in a similar vein to the various Zelda LOTADs, this might be better called a Low Optimization Segmented Speedrun.