We've saved a full minute off of this in the last 5 months! You wanna see some speed? Watch this.
I couldn't find a better place to post this, but here is the segmented run that takes the fastest IL times done by humans from a few weeks ago and compiles them into one speedrun now. The segmented time is even faster now, but here is the fastest segmented run that has been created so far.
Definitely 8 seconds at least. (EDIT: this aged well.) Sub 3:30 is clearly possible for TAS. Notably - Lamp Skip (-1s) Jail Movement (-3s) Test Movement (-0.5s) Not dying on the way to Pink Guy (-2s) Faster deaths at the end (-1s) Remembering how to talk to e-Guy (-4 or 5s) But there are some things done in the TAS that can't be done in a speedrun, notably speedy conversations or inhuman movements, so it would be unreasonable to think that humans can ever reach that time. But as I've said, I'd love to be proven wrong.
I guess I'll just throw this in because I can. Here are all of the timelosses from the 3:49 that could be saved with perfect play. -Better color blocks (-3s) -Better paint movement (-1s) -Better jail movement (-2s) -Perfect race skips (-1.5s) -Better typing movement (-1s) -Getting "Dying All The Time!" at the right spot (-0.5s)
With better routing, another 5 seconds might be savable on top of that.
It's also worth noting that we may no longer have enough time to walk in order to get the "Walking to One Thousand!" achievement. After finding that we could push away the achievement by walking while waiting during the Arcade, it created a nice buffer, but the times will get so fast that eventually, we'll catch up to that buffer, and the record will then be unbeatable. There are simply no other significant periods to walk while waiting, besides maybe during the test.
And just for the record, sub 3:30 is impossible.
15 seconds at the absolute max, 10 seconds is probably likely
Is it happening consistently every time? That's really strange.
I'm actually going to make an entire video about my route in detail, so hold your horses.