Neuro ARG

[safe monolith]

Quick update (ARG Monitoring):

• https://neuro-arg.github.io/arg-monitoring/?fromDate=2024-01-23&toDate=2024-01-24 is a bugfix swapping the order of Numbers 2 and Study to their correct field positions. If you've subscribed to the feed and received an update today, you can safely disregard it
• I have uploaded the """new""" Study video to the docs if anyone wants to do any form of additional analysis
• For historical reasons, https://neuro-arg.github.io/arg-monitoring/?fromDate=2024-01-22&toDate=2024-01-23 is when we first discovered the difference (here, Numbers 2 is mistakenly swapped with Study)

[onyx sparrow]

Hello. I am new here. Want to know what the ARG resolution process looks like? So I understand, for example, if I find a clue or solve a puzzle, then this must be written down in the appropriate document. Or is everything discussed here first and then documented? And things that are unlikely to help. Like, I found a couple of funny numbers today, but I'm not sure they're useful

[wispy thistle]

read the google docs first

then people talk in this channel and write down the progress and potential leads in the google doc

[onyx sparrow]

Ok, thanks

[mint sonnet]

whenever i find something intresting idk what to do with i just put it next to where i found it from is in the google doc

[abstract horizon]

You write it here

And if it's fine to the Google doc

But just if Is fine and if it works for the arg

[safe monolith]

putting it here is the best way to get people to think about stuff

shadow those into the docs for preservation, especially if it sounds like a promising lead / you spent too much time on it and need some consolation that your effort wasn't wasted

it might become the key to solving the next part, or something to consider for someone else

[uncut cargo]

care to share those?

[onyx sparrow]

Probably, why not. Need some time

[safe burrow]

[onyx sparrow]

But it's about Study, which, by docs, is solved. First number is frame count - 49969. Looks suspicious for me probably due combination of 9 and 6.
Next is bit schizo. Because of "Is anyone watching?" I started to watch too much. And noticed that pixel noise on I don't want to be an engineer isn't consistent, and have some... Splashes or something, I don't know how call it.. So I made a video to check it (sample - https://youtu.be/pR1K5NYPqaM). Then, exported brightness of each frame to excel, where counted number of frames between splashes. And got the graph - https://imgur.com/a/AIpHyVg

[uncut cargo]

...wow

[wispy thistle]

oh that graph looks interesting, good job

[onyx sparrow]

Oh, forgot to say - top graph for first I don't want to be an engineer, bottom for second

[wispy thistle]

did you use 60fps?

[safe burrow]

Yes, great job

[wispy thistle]

let me make brightness graphs for every video just in case

[onyx sparrow]

I don't remember, probably not, because original video doesn't have 60fps. On the sample, red pixels mean those, which became brighter, blue - darker

[onyx sparrow]

Ok. But tomorrow, cause now I should sleep 🙂

[wispy thistle]

well i'm the one making it...

[onyx sparrow]

Thanks

[wispy thistle]

doing meaning of life rn to check whether a 16gb black&white bitmap can be compressed to decent sized png

(1 horizontal pixel = 1 frame, vertically the video's whole range will be linearly covered in 2048 pixels)

if that wont work out, guess i'll reduce it to 1024

wait no not 16gb im stupid

it's only 50mb we good

[wispy thistle]

[Filtered] may contain a brightness graph message now that i think of it

anyway it's generating at a rate of approximately 1 video second per real second, so you'll have to wait for like an hour for me to generate everything

[wispy thistle]

heres i dont wanna be an engineer frame-by-frame brightness

there are dots at the bottom, those are the outliers

i should increase the dot size i guess

thicker dots

you cant actually see these dots if you zoom out to the entire video

heres [filtered]

heres brightness at the end of hello world! (during the black screen)

...and during the "hello world!" text

public static void, static 1

[abstract horizon]

This Is a CLUE!

[rugged plank]

Where it all began.

[onyx sparrow]

It seems to me that I did not express myself quite clearly. This brightness graph (excel) is not from the original video, but from the red-blue video like in the sample. And since this video shows the change in frame relative to the previous one, the peaks on the graph mean how many pixels have changed. So, for example, on the left side the line goes very high, because at that moment the Study girl image appeared.
I'll try to do the same with another video.
I can publish Excel, where this graph was drawn (or a Google spreadsheet, but I haven’t figured it out yet, so I’m not sure), but it seems that I have a problem in the spreadsheet - the frame indexes do not match the video.

A few notes. Since the original video has 23.98fps, blender used this value when rendering. And somehow it turns out that there are 49969 frames only at this fps. If I change fps in blender, the frame count of the source video changes. I still don't understand why it works this way.
And I also saw that on my excel graph not only I don't want to be an engineer. I didn't realize this before because Study girl doesn't disappear/blink in the video until ???. So yes, the graph shows data from the start of I don't want to be an engineer to the start of static, ???

What's interesting to me about this graph is why do these numbers (mostly 141 and 94) appear here? Like, I can assume that the creator of the video could have some segment (141 frames), which he then repeats. But no, if you look at 2 frames with a difference of 140 frames between them (the 141st will be the one we are comparing with), you can see that they are different. If this is just a picture on which an effect is applied, then how likely is it that it is only because of it that such a graph is observed? Or, for some reason, the creator of the video did it this way...
Also, a strange thought about “Is anyone watching?”. It seems that everything that was found from the video appeared simultaneously with some kind of audio event - a change in tone or BPM of the song. So it turns out that watching is not necessary.

[drowsy dagger]

Wow, how much observation...

[safe monolith]

There are 49969 frames in the video. If you increase / decrease the frame rate in any editor, some kind of interpolation has to take place - point is, you will get more/less frames, albeit automatically generated.

As for whether this effect is intentional, either by interpreting the value, interval, or the fact that it exists at all; I had an undocumented crackpot theory about CRTs at some point.

It was something like: you are viewing Study from a CRT screen (some giveaway sounds of a CRT turning on and off in Study and PSV), and there is also what looks like a CRT in the computer room found in Meaning of Life. CRTs are known to flicker and produce static-y images when some sort of "connection" is loose.

[safe burrow]

[wispy thistle]

you can still see it on the graph

[wispy thistle]

here, look below the white line

there are individual dots spaced 141 pixels apart

thanks for the clarification though, i'll see if that makes it clearer

[wispy thistle]

if i use raw frame difference the picture is about the same

xor looks a bit more interesting

thicker dots again

public static void

filtered (denoised and THEN passed thru xor)

psv vs filtered static

[thorn kelp]

But what’s up with background on the soundcloud?

Does marker on left side means something?

[wispy thistle]

what marker?

[thorn kelp]

Oh

I can’t send photos

Can I send you in DM?

[wispy thistle]

yeah embed perms requires either subbing or talking for a while, and sure, you can

[safe monolith]

x-post it when you receive it pls. given the feed didn't update, such a marker should have been there ages ago

[wispy thistle]

imo its jpeg artifacts from cropping

(look at the original bg, it had white lines to the left and to the right)

[safe monolith]

Yeah probably an artifact

[thorn kelp]

Thx

Can someone reply or resend link to google docs pls?

I’m from phone and can’t find them

[cloud crystal]

https://docs.google.com/document/d/17KrbPUAVu4wDTbj3sBtWTnjRsjDTCdh89F_5L1LvX0A/edit

[thorn kelp]

Thx

[wispy thistle]

meaning of life xored graph

the only interesting thing i see is how brightness increase at the start has some dots above

maybe the artifacts from study are the same effect

heres full study

[safe monolith]

If you can confirm this, you'll be solving one mystery we've had for a while

[wispy thistle]

anyway most of study has patterns like this

where the same static repeats, sometimes in long fragments, sometimes in short fragments

[digital radish]

alot of chairs

[safe monolith]

Which is whether all static is sampled from the same source

[wispy thistle]

well xor wise it's totally different

ex

[digital radish]

alot of beds

[wispy thistle]

but the same idea here - sometimes static fragment plays to the end, sometimes it's cut short

so this can be transcribed to 0 and 1

theres a loooot of 1s tho

so maybe it's tap code or something

this is just xor, so the actual static may still be the same everywhere

(the original static is probably additive)

[wispy thistle]

for example, see that high-amplitude xor pattern with dots?

here's a low-amplitude xor pattern with dots

it may well be the same static

[wispy thistle]

Study ends with very even static

11111111110 repeats 3 times

the longest 1 sequence i found was 13 1s

the shortest was 0 1s

Rough comparison of first and second i dont wanna be an engineer static:

111111110111111111011111111111001001111001110101111111011111111001111001111110111111100111
101111111011111111111110111111100111111101111111011101100111011111111111110010111010111111

I'm not like the other girls vs I'm fine (same audio):

111111101111111111111011111110111110
110111111111101111111111011111111110

[wispy thistle]

well, i have no idea whether this means anything

but i guess it's worth putting in the google doc

also converting gaps between 0s to numbers

(this isn't tap code btw)

[onyx sparrow]

Wooow... It looks like you did a really great job? @wispy thistle. It will probably take me a while to figure out what's coming from. If you don't mind, I'll ask you a couple of questions 🙂
I hope I understood correctly that the dots on the graphs reflect its brightness - the higher the dot = brighter the frame.
In these interesting XOR graphs, what is processed by this operator? Color/brightness of individual pixels? I now think of it like this - the grayscaled pixel matrix of the first frame is XORed onto the grayscaled matrix of the second frame, and the average brightness (value) of the resulting frame is recorded in the graph. Is it correct?
high-amplitude and low-amplitude are just different parts of the graph, right? Or are they processed in some other way?

[wispy thistle]

I'm simply calculating the average xor of every byte in the frame bitmap and the matching byte in the previous frame's bitmap

[onyx sparrow]

Oh, really simple

This is how it was done by me - https://imgur.com/a/mZd7vgS

About the meaning of life. Has anyone noticed this "eye"? https://imgur.com/a/eqLQrBE
It can be seen at start after Hello world text, and after music ends

[wispy thistle]

yeah

imo it's simply meant to signify a transition from a picture on 1 side to 360° surroundings

[onyx sparrow]

What is the sound starts at 1:02? Sounds like some sort of ASMR, maybe breathing or whispering

[wispy thistle]

the breathing might've been part of the first loop or something

either way breathing is a recurring theme

most prominently it's connected to anxiety attacks

[onyx sparrow]

I can hear it on original track too btw

[wispy thistle]

keep in mind the method of audio isolation means parts of audio may still be heard faintly

the audio does have what sounds like breathing/crying

but what you're referring to seems to simply be leftovers of the loop

[wispy thistle]

in that case it's intended breathing i guess

I don't want to be human / Is anyone watching? static comparison

11111110111011111111011111111011011111111011                            
11111101111110111101111111100111111011111111

wait, on the "supernatural human" sequence, these dots may correspond with the stripes

like this

could it be that all the static has stripes, but it's usually not visible?

@onyx sparrow I've added this info here https://docs.google.com/document/d/1RBvEINbc--fCZ_pVhPVEIbrbN2CJtPo60sxx8cx85HA/edit

@hollow garnet something i noticed for your filtered algorithm is the xor with previous frame always has a black bar at the right

also, one frame (2053) has this xor

[hollow garnet]

neat

[thorn kelp]

It’s look strange

[onyx sparrow]

Rorschach test
https://imgur.com/a/97JuRhW

[thorn kelp]

Ahhahaha

[crimson tendon]

You have unlocked new role

[thorn kelp]

Wow, now I can post images

[wispy thistle]

i mean thats actually what the image has

[thorn kelp]

Vedal?

[wispy thistle]

? this is hiyori

[thorn kelp]

It’s just a joke

But I can’t recognise hiyori on this image

It always happened to me irl

Ohhh

I sawed it

I took another picture with hiyori and sawed it

Thx

[onyx sparrow]

I understand. Just a joke)

Btw, since I don't know how to xor images, I want ask you to make xor on frame with same one, but shifted at 1 px by X

[safe burrow]

I'm just checking

[wispy thistle]

what video?

and do you want me to create a graph or actual xorred frames?

[timid swift]

xor a frame with itself

[onyx sparrow]

Just one frame, where hiyori visible

[onyx sparrow]

Yes, but second one shifted at 1px by X

[drowsy dagger]

Wow, a little creepy

[wispy thistle]

https://cdn.discordapp.com/attachments/1112082183235047506/1178354031421313024/out_scaled.mp4

[timid swift]

is this the one bred made, or is it a new approach?

[wispy thistle]

its the same one

i linked the same video he sent

[timid swift]

okok, just checking (:

[safe monolith]

this is the dark magic window version right

[timid swift]

yea

[safe burrow]

The sounds here are a piano, as you remember, I think I already talked about this

[wispy thistle]

yeah i'm slowly looking into analyzing the audio, just dont have enough time

[safe monolith]

time

[thorn kelp]

Real life

[timid swift]

imagine having time

and imagine having energy when you do have time

[uncut cargo]

oh hello didn't see you before, you're new?

[timid swift]

you're gonna scare new people away

[uncut cargo]

Like I wasn't doing that already in NN

[timid swift]

Now they have nowhere to go!

[wispy thistle]

yeah a bunch of russian people joined

[timid swift]

Makes sense with the jofi vid

[uncut cargo]

The schizo club is growing

[safe burrow]

My plan is working

All we have to do is wait for reactions and then we will have more people muaahaha

[timid swift]

Static playing the long con

[safe burrow]

No I play the piano

[timid swift]

GACE

[safe burrow]

And a little more about the nickname that I received from you.

Static is a keyword in C++ used to give an element special characteristics. For static elements, memory is allocated only once and these elements exist until the end of the programme. All these elements are stored not in heap and not on stack, but in special memory segments called .data and .bss (depending on whether the static data is initialised or not).

Yes, other terms of my nickname are also suitable

It looks like I need to drink chamomile tea again, otherwise I don’t know what I’m I do...

[tulip minnow]

Called it

Russian bros coming in with observations

Anyways, clue incoming

Just a week left

[timid swift]

February first?

[drowsy dagger]

Surprisingly, there's actually a lot of us here

More than a couple people

[tulip minnow]

Probably second or third

[timid swift]

Why?

[tulip minnow]

The pattern is that it has some connection with 3

Not sure tho cause time zones

I’ll have to check it with utc

[timid swift]

least crazy arg member

[tulip minnow]

The pattern

[onyx sparrow]

Is something about to happen?
And what is this pattern you are talking about? (I might have missed it in the files)

[wispy thistle]

a new video may drop soon

or soundcloud track

[tulip minnow]

it’s just that some guesses of mine

[digital radish]

btw hes never right

[wispy thistle]

because usually new videos drop on start/end of month

[timid swift]

Nah, it's just the voices getting out of hand

[tulip minnow]

Nah I was right both times

[digital radish]

both times you were on vacation

please go on vacation

now

[timid swift]

oh, i thought that was bred

[digital radish]

oh

[onyx sparrow]

And then it will be necessary to look for the keys to two tracks?)

[digital radish]

yeah it was bred

[wispy thistle]

well it may be a hint akin to Public Static Void

[tulip minnow]

It should be a video

[wispy thistle]

either way we're given hints when the puppetmaster thinks we're too slow

[digital radish]

we're too dumb please give us another clue

[tulip minnow]

They’ve been cooking, i can literally smell it

[strange parcel]

im new

[wispy thistle]

anyway I've done everything I could think of with bb ce and stuff and I'm not touching it until we get new clues

[wispy thistle]

welcome! here's the google docs https://drive.google.com/drive/folders/1qm6GSRlstTmjbyOFCuH1OhSmPK4pi4Aa (I recommend starting with overview), and neuro-arg.github.io has short writeups about the puzzle solutions and a toolbox with some ARG-specific ciphers

[strange parcel]

thanks

[wispy thistle]

I ran out of disk space, decided to check the ARG directory, and turns out splitting meaning of life to png frames took up 120GB lmao

(I guess we haven't converted the camera bobbing to data yet, but I don't wanna do it as it sounds too schizo)

[drowsy dagger]

Lmao

Powerfully inflated

[wispy thistle]

If there's a soundcloud link somewhere, it has a /s- in it
47, 115, 45
101111, 1110011, 101101
2f, 73, 2d

oh yeah, at 107.5 bpm, [Filtered] has around 150 notes

a private soundcloud track link would normally have around 50-60 characters if it has a full https prefix

[strange parcel]

https://www.youtube.com/watch?v=PyJSpRsWfls does anyone already watched this?

just asking

[wispy thistle]

yes, this is pretty much thought to be a bug

at least vedal alluded to it

[strange parcel]

ok

who is hiyori btw?

[wispy thistle]

there is an analysis on google docs, but it's in the dead ends or closed folder

[strange parcel]

alr thx

[wispy thistle]

hiyori is the sample live2d model, neuro sama used the hiyori model before she got her own one

[strange parcel]

ohh thx for telling me

[wispy thistle]

= hiyori

[strange parcel]

your good at explaining

i understand

[uncut cargo]

thus =hiyori

[wispy thistle]

proof by induction

[safe burrow]

https://media.discordapp.net/attachments/1072697081443131476/1199268003603296326/20240122_225417.gif

[strange parcel]

ahh my brain

[safe burrow]

[strange parcel]

i just need some rest

i will try again later

[solid aspen]

that bro got tired too fast.

[safe monolith]

one step at a time, there is alot of information, as you may know

[strange parcel]

i agree

and i didn't got sleep

[solid aspen]

yeah, it's kind of hard to start from zero to understand information that's been accumulated over months.

[drowsy dagger]

For those who are not really swimming in such information, in numbers, it is even more complicated

[solid aspen]

man, it took me 5 days to figure it out, of course I understand how hard it is.

[safe monolith]

i apologize for the crazy wall , hope you didn't spend too long deciphering that mess

[drowsy dagger]

Yeah I felt a little silly when I watched your video XD

[solid aspen]

heh, it was not that difficult.

[timid swift]

Yeah.. it's a bit of a mess..

[drowsy dagger]

It's actually interesting to dig into this

[solid aspen]

It was more difficult to translate it into russian in a way that others could understand.

[drowsy dagger]

But it still worked, thanks

[solid aspen]

did you watch?

[drowsy dagger]

Of course)

[timid swift]

Quite a few new faces since the video )

[drowsy dagger]

That was the last point that made me want to come in here. Even though I don't know much english(

[onyx sparrow]

I also thank @solid aspen for the video. And although this is the second video I watched (or listened to) about ARG, after it I decided to come here)

[solid aspen]

now you can upgrade your english language skills

right here

[onyx sparrow]

Yes, i really need it

[drowsy dagger]

I agree, good opportunity)

[safe monolith]

we're glad there are new people around the block. Please give us any new perspectives

[cobalt axle]

new people are always welcome, someone could have a schizo moment that solves the whole ARG

[safe monolith]

randomizer

[onyx sparrow]

It's was easy. For reliability, I made 2 tracks

1. https://docs.google.com/spreadsheets/d/1NVaLRey2na17xkR2r_tv-Ag4XMvK76niPF4fvZtmsJ0/edit?usp=sharing
2. https://docs.google.com/spreadsheets/d/1ysbf-joOIDamzaUHXHEM1PQuq0ZdHGsQPOfTcu-ekH8/edit?usp=sharing

Honestly, I don't see any patterns here - https://imgur.com/a/jlIDZOL

[safe monolith]

When bred first made the graph, I figured it was too chaotic to mean anything

[onyx sparrow]

Oh didn't know there was such graph

[safe monolith]

Assuming the curves in that bobbing are continuous, the derivative will also be chaotic due to the sudden direction changes

[safe monolith]

Oh don't worry about this, it wasn't documented

It's good to have someone verify it though

[onyx sparrow]

ok

About XORing an image with a 1 pixel shift.
Here is my answer: https://youtu.be/IpcQAJvwQJs
For simplicity, I will say that the pixels here are the boundaries of horizontal line segments.
||Yes, this is not XOR, but a simple difference between neighboring pixels||
Question - is it just me or does the sound somehow correlate with what is happening in the video?

[safe monolith]

Mm, the connection between video and audio is worth a shot. Thought about perhaps somehow applying the same rotation to the sound, but I know nuts about audio to begin reinterpreting the rotatation algo

[timid swift]

If there's a connection between the sound and the video scramble, I can't see it with the naked eye. It's a really neat idea though

[onyx sparrow]

I'm confident only in one sound/note - one which appears with hiyori. Have idea of watching video side by side with spectrogramm

Probably it should be done as single video

[timid swift]

yes, that should clarify it a bit more i think

[onyx sparrow]

Btw, am I only one who doesn't like sound from Filtered?

[timid swift]

It's not a very comfortable sound, no

[solid aspen]

no

[safe burrow]

[hollow garnet]

😔 i really need to find a better way to generate videos from code the ffmpeg combining a bunch of pngs

[drowsy dagger]

A lot of sounds give me goosebumps in general

[hollow garnet]

this is not speed

esp with 25000 fames

[safe burrow]

[hollow garnet]

i wonder if this would be faster if i piped bitmaps through a fifo

[hollow garnet]

i took your track data from the second first google drive link and made a video out of it. the head of the line is the posistion of the tracker. the whole thing is centred around where it started

i doubt it does anything useful but its a cool graph eitherway

also i made your track into a "drawing" where i summed up the changes from the track (i think this is integrating it?)
though coordinates might be completely useless here

also did the same thing for one of my tracks

wait im not sure if that works at all, i don't know if its supposed to be moving like that

oh well

code is here https://paste.gg/p/anonymous/6c8fab0f028a4f79bf974badf1976e6a

[hollow garnet]

this is the one for the actual second link because i forgot which was which

(these are suprisingly fast to generate, the whole thing is done in <130ms in release builds w/o multithreading of any kind)

[onyx sparrow]

In one moment I thought that there is something wrong with my track, because it's movement differs from your. But now I think it can be not simple movements, but also tilting

[hollow garnet]

i was also tracking a different part (i think?)

i believe you were on the left track and i was on the right one

(though that's all going off of your screenshot, idk where you actually tracked )

[onyx sparrow]

For some reason this video doesn't plays on my phone. Will check tomorrow on pc.

[onyx sparrow]

One of them was on the left, the second on the right. This should be obvious from the X in the csv. But I don’t remember which one from where exactly

[hollow garnet]

that's fine, itss not too important anyways, though interesting that they're pretty different in how they look

[onyx sparrow]

Agree

[hollow garnet]

honestly the fact that its always moving down and to the right probably means i picked a bad first frame or something

[onyx sparrow]

You mean, moving down and to the right in big scale? 'Cause it moves on opposite directions, but on small scale

[hollow garnet]

ill try and make some better code to explain it

[onyx sparrow]

Are you sure, you should spend time on this thing?

[hollow garnet]

i've already spent plenty of time on random arg stuff, a little more can't hurt

[onyx sparrow]

Ok, I guess ¯_(ツ)_/¯

[hollow garnet]

well

[hollow garnet]

ok i made this video to try and explain what i mean, basically being on a bad starting frame means its constantly "moving" in a direction, whereas a properly centred one will only "walk" if its away from the centre

ok yeah this produces much more consistent graphs

[hollow garnet]

i think i've integrated it now

[safe monolith]

this makes sense actually, although I'm not sure where the best tracking point would be

[hollow garnet]

I made it find the one that "walked" the least

which means it's staying in sorta the same place and thus probably not drifting

[safe monolith]

Makes sense. Sorta like getting a centered function

[safe monolith]

Still looks chaotic af to me, but looks a lot better

[hollow garnet]

anyways if you see anything of meaning in this I'd be delighted because it looks like squiggle

fair enough lol

surely I need to integrate again

[safe monolith]

do you have the csv for this? might be useful for some kind of future analysis

[hollow garnet]

I used these

[safe monolith]

ah i see

dunno if we'll get anything from this, but the fact that you've got it centered is impressive on its own

[hollow garnet]

one minute lemme upload the code again

[safe monolith]

it looks continuous, so I reckon it is possible to mathematically model this with beizer curves

[hollow garnet]

https://paste.gg/p/owobred/463507a2f6ee44e1b1947e1919e71008

[safe monolith]

or we could also use the tracking data to stablize the camera movement and then do frame by frame similarity to check for any frame changes

could have already done it before tho, so not a new idea

[hollow garnet]

¯_(ツ)_/¯

[safe monolith]

inb4 elliptic cryprography with this as the curve

[glossy sphinx]

[mint sonnet]

yall still working on bb ce?

[wispy thistle]

it's fine for me, but it's much more unsettling in combination with the video

[wispy thistle]

btw if some of you didnt know this is a monoalphabetic substitution cipher solver https://quipqiup.com/

[onyx sparrow]

Is there any certainty that this can be solved with the help of monoalphabetic substitution? Like, what words are there that consist of two identical letters? (about the "bb" at the beginning)
Or, if we ignore spaces, which words start with two identical letters

[wispy thistle]

no, bb ce definitely isn't a monoalphabetic cipher

or it may be, but it doesn't represent english words

[timid swift]

so is it something like vigenere then?

[hollow garnet]

@safe monolith i think i may have messed up with this double integration shenanegans

ngl i don't think my algorithm makes any sense at the moment

[wispy thistle]

it may or may not be something like vigenere

as I mentioned in the doc, I think it should decode to something like numbers as it has a low word length, also, it has a relatively low entropy assuming a monoalphabetic cipher, but with a polyalphabetic cipher it may be even lower (or, well, higher)

[timid swift]

Isn't it a little small to do entrophy analysis on?

[wispy thistle]

it has 15 different chars out of 33 letters

which is kinda small

but indeed doesnt say anything conclusive

[timid swift]

I don't know enough about language and probability to say if that's likely or not

[wispy thistle]

well the sentence "but indeed doesnt say anything conclusive" has 37 letters and 16 different letters so i guess it may be english, but the word lengths still don't match

[timid swift]

so it's at least in the realm of the possible

[wispy thistle]

it's possible we need to first scramble the letters in the correct order and then solve a substitution cipher

there are 7 letters that only appear once each, but e.g. b appears 5 times

which may serve as a hint or something idk

i dont actually remember the most frequent letters in english

i only know that о is the most frequent letter in russian

https://en.wikipedia.org/wiki/Letter_frequencies

[timid swift]

right, but with a word of the same letter twice, it wouldn't be a simple substitution cipher, or at least not in English

hence the question about vigenere or similar

[wispy thistle]

it might mean that we need to scramble the groups in the correct order and concatenate them

[hollow garnet]

damn i thought this would've gotten hotter but i guess im probably memory bound or something silly

[wispy thistle]

the most common doubled letter in English are LL EE SS OO TT FF RR NN PP CC btw

[timid swift]

That's fair, we may need to invent an algorithm, but I think it'd be wrong to dismiss preexisting ones

[wispy thistle]

considering B is also the most frequent letter in the sequence, it wouldn't be too huge of an assumption to say B = E

[wispy thistle]

i'm not saying that this isn't vigenere, it's just that i've tried both vigenere and shifted vigenere from my toolbox with some keys i thought up

[timid swift]

Ignoring the spaces altogether is a valid thought

[wispy thistle]

no, not ignoring

[timid swift]

alright have fun

[wispy thistle]

the letter groups are clearly relevant

spaces shouldn't be ignored

but

it doesn't mean these are the final words

for example, maybe the real word is formed by concatenating 2 groups

i mean just ignoring groups is an option too but i've tried that already (with mono substitution ciphers, and with vigenere it wouldn't matter)

[hollow garnet]

it might be a good idea to do this properly instead of just brute forcing it

[timid swift]

Still tracing the bobbing?

[hollow garnet]

yeah but im doing it twice

actually on the count of doing the math this will take a ridiculous amount of time

perhaps i should rewrite this

[timid swift]

Can video editing software do point tracking?

[hollow garnet]

this isn't about doing the tracking, its trying to draw from the track

afaik all the tracks are from blender

[timid swift]

like draw a line from the pattern of the bobbing?

[hollow garnet]

yeah

except the line's posistion is being integrated for lack of a better way to put it

[timid swift]

what, subtract the previous change from the current one?

[hollow garnet]

adding them instead

ok it finished running

is this a clue (double integrated)

[timid swift]

ah yes, this means something

[hollow garnet]

i don't even know if this is integrating the best one or not, my code is broken as hell

[timid swift]

I don't even know what you mean by integrated :p

[hollow garnet]

idk if this helps at all

[onyx sparrow]

i'm too. Explain, please

[hollow garnet]

look its another mysterious bit of bred code that doesn't make sense

ill try and cook up an explanation while im eating

[timid swift]

ah yes, the mystery luck pot. Keep at it and we'll have another unexplainable but functional piece of evidence like the envelop rotation thing

[hollow garnet]

so I guess the first question to ask is if you understand what integration means in a math context

because it's kinda doing that for each axis (x and y) on their own, then drawing them after they're added together

[timid swift]

Integration is summing the area under a graph, no? Infinitesimally small slices and all that

[hollow garnet]

yeah, if you have a continuous graph that's what you do

[timid swift]

Right, we have points in time - frames

[hollow garnet]

this is discrete though, so it's just adding the heights instead because it's effectively the same

[timid swift]

So what determines the slice's height?

[hollow garnet]

it's vector from where it was on frame 1

[timid swift]

oh, so it's just where each frame is located in 2D space compared to the first frame?

[hollow garnet]

yeah, then summed together like a running total

if I can figure out Godot I'll try and explain it better

[timid swift]

I get it now, but I can see why you can't find an easy way to explain it :D

[hollow garnet]

then this is the result of doing it twice

[timid swift]

Doing it twice meaning running the output through the same algorithm again?

[hollow garnet]

the reason it takes so long is because I go through every frame offset to find the one that produces the smallest shape

yeah

[timid swift]

repeat enough times and it'll be a straight line

[hollow garnet]

actually this method would probably just create chaos or something

differentiating might get you a straight line

[timid swift]

Ye, it reminded me of differentiating (specially with all the math talk)

[hollow garnet]

we haven't used the differentiate integrate line yet have we

[timid swift]

that could change

[onyx sparrow]

I'm not sure I understood this. The message about the area under the graph seems to remind me of something.
To check, I drew https://imgur.com/a/3dhXwqD
Do I understand correctly that you are calculating the area (yellow)? Or, do you calculate it for each frame (separated by orange)?
And then you use the resulting values to build a graph (or for next iteration), right?

[hollow garnet]

basically, that yeah

one minute im cooking something in godot

[onyx sparrow]

I'm not in a hurry) It's time to sleep anyway, get up in about 5 hours. And I'll be gone for a couple of days.

[hollow garnet]

ok i have made quite possibly the worst godot app ever

so the cat in the centre is where the track started, the amongus is the posistion at a frame and the neurooper is where the line would be connected to for the corresponding frame

[drowsy dagger]

Arg is more interesting and informative than the whole school and university

[hollow garnet]

basically all the code is doing is this

except mine has some silly stuff to choose a "better" starting frame that means the neurooper doesn't wander too far astray

[onyx sparrow]

Today tried to create a draw.io diagram, with all clues and related stuff. Just for better understanding what is going on (and I love restructurize information when diving into something new). And I get wondered - did it required 2 more videos to decrypt description from Numbers I?

Beside 2 that already was on channel

[wispy thistle]

Numbers is self-contained, but indeed Public Static Void made it easier

[timid swift]

Numbers I could've been decrypted without further evidence, yes. But we didn't

[onyx sparrow]

By the way, about PSV. When I heard about "go back to your studies" in jofi video, knowing about lost on translation, I thought that it can be clue referring to Study. But subtitles refused that idea

Now I think that name can be a meta humor about Numbers I. Like, public - video is public; static - it was there since YT chanel was discovered. Or wait.. Most likely it's about description, which for some observers it's just a noise, like what calls "static" in videos; void - because description didn't produse results

[solid aspen]

what's lost on translation?

[onyx sparrow]

I found a program, where you can change spectrogram by drawing on it, and then play result. However, it's paid, and free demo crashes when I turn lossless mode on.
https://photosounder.com/ maybe there is some alternatives, but I didn't find

[onyx sparrow]

One moment, generating explanation

[solid aspen]

[safe burrow]

[timid swift]

I'm lost in translation

[solid aspen]

[onyx sparrow]

Maybe it's not "lost in translations", idk how to name it properly. I'm about https://youtu.be/J4w3UUPQpsE?t=22m14s. In that moment I thought that in original it can be something like "go back to study". To study video. But in original it's "..to your studies". In translation I can't get original form of word. And, now I see that it was "go back to YOUR..."

Probably, it's me, who don't know English so well

[safe burrow]

What? English only

[onyx sparrow]

I'm sorry, I know. Yes, I should insert a link with timecode. May I correct it?

[safe burrow]

[safe burrow]

Okay, continue.

[timid swift]

Surely it's fine. The original phrasing was relavent to the discussion (:

[fast path]

I am so confused
the numbers ciphers seem so easy now

[timid swift]

how so? o:

[onyx sparrow]

As I see, this is not strictly prohibited. Either way, please accept my apologies.

[fast path]

for numbers it was just following the instructions given in the lyrics to change a number
now we have to shift video frames and stuff to find the clues
maybe that's not actually hard, I just personally lack the knowledge

[solid aspen]

i don't think that was "lost on translation", because how else could I translate that sentence?

but ok

[safe burrow]

[onyx sparrow]

Yes, I understand it)

[solid aspen]

[safe burrow]

Now it's even more fun

[safe monolith]

[safe monolith]

Chances are, the solution will always be retrospectively simple

We just do many things because we dunno what is the right way

sometimes angel numbers

other times dial up sounds

[mint sonnet]

i dont know how, but i got something readable in english with it

[south sierra]

Did you guys solve it yet?

Holy I missed a lot of stuff lol

[timid swift]

There was a bit of a leap of faith after doing the number change thing. We had 1759032572147924 early (in a jumble of many other numbers), and I think that several people actually had it. The change to 1bad0fcabc1ebdce, the real decryption key, was what we didn't get without the hint in public static void.
It could be similar here. All the groundwork already done, but nothing to show for it because we're still missing just one step

[strange parcel]

did you guys found anything on hello world?

[timid swift]

There's a document for each video (:
https://docs.google.com/document/d/171dJEDRxJRUDYic2vboz4VKaOh_AC-VgVYtnoAbeuuo/edit
but in short, no

[strange parcel]

thanks

[tulip minnow]

Lol hiyori is dead

[unique mason]

soon

||by which i mean many months||

[safe monolith]

you're getting there

[unique mason]

someday

i think ill start posting updates sometime when i think part 1 is done

[safe burrow]

Hiyori never die

[solid aspen]

already 😭

[hollow garnet]

just found https://github.com/alesanfra/iterframes/blob/master/src/decoder.rs which would mean my silly rust code doesn't need to have 30gb of pngs to work on a video, might consider using this if its able to work fast

still need to find a nice way to make videos without saving a bunch of pngs but that should hopefully be possible

[timid swift]

eh, you always gotta make an inefficient prototype first to ask the right questions!

[hollow garnet]

just need to spend ages making a good library for this

[wispy thistle]

I've tried ffmpeg and it was quite slow for decoding (slower than using ffmpeg binary) but then again I did it in the first way I found

as for encoding... yeah I tried the av1 encoder however it's called (rav1e or something) but couldn't figure it out

and encoding using ffmpeg was even more annoying

biggest problem is colorspace though, virtually no video format uses rgb

[safe monolith]

hmm, in theory converting from yuv to rgb should only require a math formula. small performance hit, but surely better than having a bajillion frames as an intermediary

gpu programming for extra speed

[wispy thistle]

yeah I mean especially for grayscale it's pretty easy

it's just that yuv is... weird, I still have no idea what the memory layout is

[safe burrow]

[dapper wraith]

time to decode

[wispy thistle]

sure

it looks like a valid, albeit hand drawn, qr code

im guessing it was originally this

[safe monolith]

fill it up with a buncha whites and hope there is enough error correction. I think almost all of the ecc bytes are there

surely important clue

[safe burrow]

[hollow garnet]

where is this from?

[wispy thistle]

#1201523099267502110 message

time to look into how qr ecc works because this isnt decoding

for the record
111111100101101111111100000101101001000001101110101100101011101101110100101001011101101110101000101011101100000101001101000001111111101010101111111000000001111100000000110100110110011110110101011010100100000000011011101010100000000010001011011100000000001110101100000000000000000001010000000000111111101010000000000100000100100000000000101110100100000000000101110101110000000000101110100110000000000100000101110000000000111111101100000000010

[digital radish]

thats the default windows error qr code

wait no

theres many

guh

its different for each code

you need to find the correspondant code

[wispy thistle]

well in this case the size doesnt match though

the windows qr code encodes a link

while this qr code is too small to encode a link

well not too small but it's still quite small

plus look at the author

[digital radish]

yeah

theres nothing special in the qr code tbh

plus we'd have to bruteforce it

[wispy thistle]

we dont

theres ecc

[delicate nebula]

Aren't the markers in the corners color inverted, or does that not matter

[wispy thistle]

it was just easier to recreate it that way

of course i inverted it

[delicate nebula]

It looks like a real qr code anyway

[timid swift]

Have we tried drawing in the little QR-eye and see if it scans?

I think all of the error correction data should be there

[wispy thistle]

i have drawn it into

(btw you cant just draw it, you have to adjust it depending on the xor mask)

[wispy thistle]

wait bruh this has the lowest ec level

7%

https://www.thonky.com/qr-code-tutorial/

[safe monolith]

this pixel shouldn't be possible

we have enough bytes to brute force the data bytes to check if it matches our error correction bytes

[timid swift]

really? i thought the eyes were the same regardless of mask

[safe monolith]

this size shouldn't have smol eyes

it's a 21x21

[timid swift]

ah, right

[safe monolith]

at bred do you still have the qr bruteforcing thingy

oh i may as well @hollow garnet

[hollow garnet]

uuuuuuh

i do but its currently hardcoded for 25x25 qr codes

and also a different qr code to this one

[safe monolith]

okie no problem

[timid swift]

why shouldnt that be possible?

[hollow garnet]

https://paste.gg/p/owobred/4f105bab8e2440ff95b98c014a560fbb

its not particularly amazing code

[safe monolith]

transcribing the same QR but on qrazybox doesn't let me add that additional pixel. If it's low error correction on mask 7, which it should be, then that pixel, being part of the metadata, cannot be there

ignore the bottom right, i took the liberty to guess it might be byte encoded

[hollow garnet]

to be fair im not totally sure that pixel is visible

[safe monolith]

we're only missing 15 bytes, ezpz

[hollow garnet]

just one aes key

[timid swift]

ye id say that one isnt there

[wispy thistle]

? check again

[safe monolith]

https://merri.cx/qrazybox/ crazybox literally doesn't let me add that pixel

[hollow garnet]

i realized this was missing two files so i've updated it

[wispy thistle]

oh right i'm dumb

forgot about the 1010101 patterns

[safe monolith]

if i have some time tonight (matter of if ), i'll mod bred's bruteforcing thingy and check if we can get something out of this

[timid swift]

suppose there's not enough data to recover the full qr and automatic tools wont work. how hard is it to manually decode in part? it sounded like you were confident about the mask

[safe monolith]

(guessing this is byte encoded)

[wispy thistle]

the mask is encoded near the squares

[hollow garnet]

fwiw qr.txt and mask.txt takes qrazybox inputs

but there's a lot of 25s that will need to become 21s

[wispy thistle]

in this case, look below the top left or to the right of the bottom squares

bbwbwwb....
bb = lowest level of error correction
wbw = specifies the mask
the rest = idk

[safe monolith]

the rest is ecc for the metadata

[wispy thistle]

let me proofread your qr in a bit

[safe monolith]

thank u

qr codes are stupidly sturdy because of how much ecc is going on

[hollow garnet]

also I just wanna say my chaotic code is not to be trusted lol

[safe monolith]

a sketch for another potential way to go about this:
i'm supposing we'll be getting ascii + a limited set of symbols, probably -,@. In that case, we really only have to check 26*26*3*16 possible combinations, and do a similarity comparison for ECC. Can be slightly off even if we don't trust our transcribing since we have the mask and everything

[wispy thistle]

should be white white white in the middle

[abstract horizon]

where does that qr code come from?

[wispy thistle]

#1201523099267502110 message

[abstract horizon]

Oh

[safe monolith]

so qr code generator library -> check pixels, return similarity score -> get lowest set of scores. algo should only take a few minutes to run unless these assumptions are all wrong

now the question is do i drop all my responsibilities for this

[wispy thistle]

may be safe to add a white here but im not sure

[hollow garnet]

actually that's something I should check

how much text can you fit in a qr code before it gets big

[wispy thistle]

this has lowest ec so up to ~40 chars or something

[safe monolith]

yup, but the grey squares (unknowns) limit us to like 16 unknowns max I think

like if we're not going to transform the remaining bits (which could be a combination of some data + all of error) that is

if alphanumeric encoding could be a bit higher

[wispy thistle]

anyway the data we do have doesnt have any ascii

so i guess its all EC even if it's set to 7% or something? idk how that works

[safe monolith]

EC is kinda merged with the data a little

but only towards the end of data

[wispy thistle]

i see

we arent sure about this being white

hmm

i guess it does tint blue a little

but if you expand

this kinda looks like a shadow

so imo this should be unset

anyway the rest looks fine

this part is also very weird

it can be
BW
BW
WB
WB
or it can be
BW
BB
WB
WB

im learning towards the latter, your version has the former (as has mine lol)

[hollow garnet]

@safe monolith just thought i'd give you a heads up that im reworking my thing to understand smaller qrs so we don't do the same work twice

[hollow garnet]

https://paste.gg/p/owobred/748187d17bf543b38c1b27c6fbea8c40

(fwiw im hitting about 19k/s)

(i have also never checked if this actually detects if they are valid )

[wispy thistle]

nice const stuff

this is rust we're talking about, const stuff is pain and suffering (unless you use some random nightly flags idk whether there are some that help)

mostly because of no const impl trait i think

[hollow garnet]

yeah but is it too much to ask to be able to square a const generic number

[wispy thistle]

ah, i guess it is lmao

there's some crate for type-level const nums, guess you could use that

[hollow garnet]

meh, ill just bodge it for now and manually calculate the numbers

gaming

[timid swift]

how many possibilities are there?

[hollow garnet]

there are 110 unknown bits in the qr im using

so 2^110

[wispy thistle]

1298074214633706907132624082305024

[timid swift]

...

I have no words

[hollow garnet]

don't worry it slightly less stupid than just brute force

we might get lucky if it decides to error correct it

[strange parcel]

ok

[timid swift]

How so? It detects near matches, or?

[hollow garnet]

it relies on the qr code's built in error correction

[timid swift]

hmm, but that could trigger on false positives too, no?

else the QR could be scanned

[hollow garnet]

i dunno

[timid swift]

Hmm

[hollow garnet]

well i let it run though 50 million and it found nothing so i've had enough of my pc being slow

[timid swift]

we legit may have to learn how to read the error correction for a partial recovery..

[wispy thistle]

as I said, the issue is it's set to the lowest EC level

but yeah I'll try to understand the EC math

maybe there's a way to approximate the data

[safe monolith]

ec for qr uses reed solomon, which i only barely understand in theory. In essence, it uses galois field math, which relies on some properties of XOR

The issue is that AFAIK these operations rely on both the ec and source data to correctly compute a desired result. I think the irreducible polynomial or something

So there isn't a way to go from error code to actual data because the operations are lossy by default. Verification is possible

At least that was how i interpreted it

I need to re-read all that stuff though

[wispy thistle]

https://en.wikipedia.org/wiki/Chien_search may be related idk

anyway this is tons of complex math

[safe monolith]

It's highkey harder than AES in terms of math

[timid swift]

Right, it's hyper optimized with a lot of tricks. What a pain

[safe monolith]

https://en.m.wikiversity.org/wiki/Reed–Solomon_codes_for_coders

Polynomial long division

[wispy thistle]

thx

[safe monolith]

Ah ok after a quick glance through. Tl;dr: to verify, we want to get 0s after some operations. To recover, we use smart math to find the possible polynomial that would give us 0 after those same operations

It's not reversible, but guessable

[wispy thistle]

so chien search may help?

it's used for finding roots of polynomials

[safe monolith]

Hmm maybe? There is a irreducible polynomial in the mix though

[wispy thistle]

ah right

[safe monolith]

Ok no, it must be able to help:

def rs_calc_syndromes(msg, nsym):
    '''Given the received codeword msg and the number of error correcting symbols (nsym), computes the syndromes polynomial.
    Mathematically, it's essentially equivalent to a Fourrier Transform (Chien search being the inverse).
    '''

This is for the syndrome polynomial though, hmm

[wispy thistle]

sadecsss when he keysmashes into a qr encoder and erases 50% of the qr code, knowing full well we will waste hours of our life on it:

[safe monolith]

decoding Reed–Solomon consists of a Fourier transform (syndrome computer), followed by a spectral analysis (Berlekamp-Massey or Euclidian algorithm), followed by an inverse Fourier transform (Chien search)

those are words

[hollow garnet]

fourier transforms my beloved

[wispy thistle]

fourier 🤝 fourier
my beloved

[delicate nebula]

how much of a coincidence is it if a qr code has the exact same pattern in the top middle?

like this

[safe monolith]

ecc is sensitive to both itself and data

So as long as the ecc bytes don't fully match, quite unlikely they represent the same thing

So in that image you've given, the left side deviates too much

[gentle locust]

Mmmh? What qr code is this?

[delicate nebula]

this is what you get if you put "neurosama" in some random online qr generator

[gentle locust]

Eh? Why would you do that :o?

[safe monolith]

Oh shit, padding bytes. @dapper wraith sorry for the ping, but do you have any insights about padding?

[safe monolith]

In particular, between this QR and the one we are trying to crack

[delicate nebula]

figured it was worth a shot to try some likely phrases and see if the qr looked similar

[safe monolith]

If the middle section has uncanny matches, it might be padding

[gentle locust]

Ah I see

But which QR code are we trying to fill? Where does it come from :o?

[safe monolith]

The bottom of the QR code looks really similar

[delicate nebula]

#1201523099267502110 message

[safe monolith]

yeah ok i'm pretty much convinced the last few bytes are padding / share the same last alphabets. lemme emperically check

[wispy thistle]

turns out i already know some finite field math from competitive programming lol

[safe monolith]

if this is right it means all of grey is going to be pure data

[safe monolith]

galois field math is basically XOR but with polynomials

[wispy thistle]

do we create a doc like we did with enigma machine and stuff?

or does this go into stubs?

[delicate nebula]

If practically all we have is padding bytes, is this even possible to decode lol

[safe monolith]

what website did you use here btw

[gentle locust]

Oooh thank u

[wispy thistle]

theres a lot of error correction data

which doesnt necessarily mean it's possible, but it may be possible

[delicate nebula]

https://www.qr-code-generator.com/

I tried a few others and they basically all gave different results for the same string, one of them had the same top middle bit though

This is the other one I tried that produces the same pattern at the top https://keremerkan.net/qr-code-and-2d-code-generator/

[safe monolith]

https://www.nayuki.io/page/qr-code-generator-library <-- replicated with this. mask = 7, don't boost ecc

can practically confirm that has to be padding

empty should look something like this. top pattern matches. i'm ignoring the bottom circle because that's in error territory and i'm scared of it

[vapid token]

Interesting decision, to mix the old blue screen look with new one.

[wispy thistle]

btw i somehow guessed it correctly

it's exactly 40 chars

(the limit)

but padding means the string is shorter than 40 chars

[safe monolith]

by pure experimentation 13 chars is the limit before the padding dies

[wispy thistle]

it's fine for me with 012345678901234567890123456789

for some definition of fine

[safe monolith]

probs alphanumeric?

wait lemme check

[wispy thistle]

ah yeah its numeric

[safe monolith]

surely it must be ascii right

knowing the unknown is pure data is actually quite useful. this means we don't have to worry about the data + error correction bits of the ecc

dictionary attack?

[hollow garnet]

wait is my code generating qrs with the wrong ec level

[safe monolith]

is it?

.ecl(fast_qr::ECL::M)

[hollow garnet]

https://paste.gg/p/owobred/748187d17bf543b38c1b27c6fbea8c40#lib-rs-l165

I don't set anything in the random one

so I have no clue

but probably not

[safe monolith]

if it's auto it can choose anything

[safe monolith]

arbitrarily

[wispy thistle]

ah yes, simple

(i know some of those words)

well tbh it is quite simple if you know the basics of finite fields... mhm yeah

[dapper wraith]

Sorry. I missed it. All I know is that the padding is just an alternating sequence of 236 and 17 and is not subject to any masking.

[safe monolith]

thank you that info is huge

this explains why sometimes i see a giraffe instead

[wispy thistle]

btw we can reduce quite a lot of possibilities if we consider it to be valid ascii

(surely it isn't numbers)

[safe monolith]

I agree
(echoing the sentiment that it isn't just numbers)

[wispy thistle]

we don't know the exact length tho, so while the beginning is ascii we don't really know how much padding there is... I think?

wait 237 17

236

[safe monolith]

yep so we have to search ascii + padding

[wispy thistle]

.

it ends with 3!

ec 11 ec 11 is padding

no way we decoded a character!!!

we do get the entire error correction sequence though

that's nice

[safe monolith]

pure guess, 572943 but byte encoding?

[wispy thistle]

too short but maybe, why not

[safe monolith]

need a qr generator that also lets us select encoding

[wispy thistle]

I just parsed the qr code manually btw which is the only reason I was able to get the 3 lol

[safe monolith]

makes sense

[wispy thistle]

btw I was only able to decode the last 95 bits from the image, so the 0 in the beginning is speculative, if it's 1 then isn't ascii tho

I wonder how padding works with other encodings

oh right it works more or less the same

anyway better to assume it's ascii

for our sanity

[dapper wraith]

The padding is always the same. Only difference is that you might need to add 0-bits to the end of the data to get it to a byte boundary

[wispy thistle]

I see, thanks

there's a total of 17 chars with that version and ecc level

so 4 bytes padding means the string is 13 chars

we now have the presumable mode and the length field

less data to guess

https://research.swtch.com/field another article

[strange parcel]

ohh math

[wispy thistle]

yeah this article breaks it down nicely though imo, compared to the articles oriented towards... people who know math

[safe burrow]

[wispy thistle]

it looks like decoding is complex and somewhat of a research topic, while encoding is simple enough

does anyone here want to do a phd on reed-solomon?

anyway we get 4 out of 13 bytes, which is like... 5-6? known bytes for reed solomon, out of 14-15, plus 7 bytes of reed solomon, plus whatever info is in the area near the rectangles

additionally this is ascii, so for example if there are 256 correct ways of constructing X bytes of data corresponding to X-1 reed solomon bytes (idk if that's true), then while we would normally have 65536 possible values, they would nearly all get rejected for being out of ascii range

this seems theoretically doable so far

just math-heavy

for starters i want to finish understanding reed solomon and then figure out what these 7 bytes (plus the data near 3 aligning squares) actually means

[safe monolith]

Eh? From my (past) understanding, the way reed solomon is constructed makes it each byte dependent on the byte that came before it

Changing just one bit of data will massively affect a majority of the error bytes

[wispy thistle]

yeah, that doesn't in any way conflict with what i said

[safe monolith]

So there shouldn't be many possible ways to construct X bytes of data is what I'm thinking

[wispy thistle]

I didn't mean constructing 256 sequential valid decodings

but 256 all over the place ones

[safe monolith]

Ah

Yeah then probably checks out

[wispy thistle]

there's a 1/8192 chance of random 13 bytes being ascii

so we should get around 8 possible values

out of which we can choose the one that makes most sense

10 tin cans moment

[wispy thistle]

there's a loooot of different reed-solomon decoders

surely one of them will work

(by "decoder" i mean "decoding algorithm", of course)

i think we can spare up to like O(n^7) which is enough for just about any practical algo

even the original algorithm in O(n!/(n-k!)/k!) may work

https://en.wikipedia.org/wiki/Lagrange_polynomial

okay lowest degree may be useless

https://en.wikipedia.org/wiki/List_decoding

here we go, this is what we need

Given the fact that bivariate polynomials can be factored efficiently, the above algorithm runs in polynomial time.

WAYTOODANK please keep the word "polynomial" to at most 1 in a sentence kthx

[zinc hinge]

. - .
i come back to neuro arg and you are just cooking some good methamtics

[wispy thistle]

this won't even result in anything official lol, but we do have a tradition of decoding anything sadecsss gives us

sorry i mean, for all we know he could be collabing with vedal on the arg, what if this time it's an important hint, we must decode this asap

[distant rune]

New clues?

[wispy thistle]

anyway a summary of everything so far:

#1201523099267502110 message has a heavily obstructed qr code

the qr is valid but seems hand drawn or otherwise a bit distorted, but can be converted into an image

the qr code has the lowest possible amount of error correction (up to 7% is allowed to be corrupted), so we can't recover it the simple way

we do have the entire error correction data (which is not enough)

we also know that the qr code most likely has 13 ascii characters, the last character is "3" (tentative, needs double checking as i used an old version of the qr code which has errors)

we can try using fancy math to find the ascii using the error correction data, but regular qr readers dont use fancy math so they're useless here, even reed solomon decoding libraries dont usually implement it

errors make me who I am

just like the QR code can be decoded only by using error correction :schizo:

https://cse.buffalo.edu/faculty/atri/courses/coding-theory/book/web-coding-book.pdf has some of the math

ft. python code https://github.com/ofekalg/error-correcting-code/blob/main/RS.sagews

[wispy thistle]

new version

~90 pixels are missing, which is 20%

qr is readable with 7% missing, which is 30 pixels, guessing 60 pixels is improbable

so we do have to do it the proper way

[strange parcel]

alr

[strange parcel]

how can we find the other piece?

[wispy thistle]

math

[strange parcel]

ok i will do my best

[wispy thistle]

see this piece we have? this is some polynomial that somehow encodes the data in some way

[strange parcel]

ohh

[wispy thistle]

normally it's used for recovering small pieces of data

in our case we have to recover most of it (we only have 5 bytes out of 15)

[strange parcel]

10 more?

[wispy thistle]

something like that

[strange parcel]

ohh ok

[wispy thistle]

13 characters specifically

[strange parcel]

tell me how

[wispy thistle]

i'm doing a qr code encoder just to verify all this and learn how it works lol

[strange parcel]

i want to help

[wispy thistle]

uh the problem is i dont know myself, but basically, https://en.wikipedia.org/wiki/List_decoding some of these algorithms should be used because we dont have enough data to use conventional algorithms

im researching qr codes and reed solomon encoding right now to understand the problem

from what i see this is decodable, but not using conventional algorithms (as already mentioned) but only experimental ones

[strange parcel]

if you need help you can ask me

[lost niche]

if yall tired you could probs cure cancer through scitzo, jesus

[wispy thistle]

okay it doesnt end with 3

alternatively, this is a number qr...

let me explain

heres the qr code, assuming its ascii and 13 chars, without the error correction data

the data above is the padding, 236, 17, 236, 17

at the right, there are lone pixels

they are the beginnings of ascii bytes (each ascii letter/number/symbol is below 128, so the most significant bit is always zero)

so far so good, but

note how there's 4 bytes missing between padding and the last letter

(the grey area is 11 bits, while it would be 7 for missing bits of an ascii char)

now, that's actually expected

because qr codes start with 4 bits for mode, 8 bits for data length, and after we add some bytes their length will be N*8 + 12, while before adding 236/17 padding their length must be divisible by 8, so we have to add 4 extra bits

those bits are 0, according to this site: https://www.thonky.com/qr-code-tutorial/data-encoding#step-4-break-up-into-8bit-codewords-and-add-pad-bytes-if-necessary

so, when we try to set them to 0

it turns into this

while the original qr code has this

so, the original qr code actually has 0001 in the position where ascii would have 0000 padding!

which means it may not be ascii?

if anyone has an interpretation of the qr code that stays consistent with this i'd really appreciate it...

oh hm, what if i use 11 chars?

that's more like it!

but theres still a mismatch aargh

wait maybe not?

to the bottom right of the large blue blob, is there any chance there could be a blue pixel? aaaargh

if not, this isnt ascii

if yes, this could be ascii

so, assuming ascii, this is the new version

but i dont think we should assume ascii anymore...

because i dont think this matches

and a closer match than this is impossible to get with ascii

well, worst case i find some potentially valid ascii, throw it away and we move on to numbers... i guess

[wispy thistle]

anyway i'm done for now (the result is a qr code generator, though without ecc for now, pretty cool i guess), but i have no idea whether the code being ascii is possible

[wispy thistle]

ecc seems like just long division of the message by a generator polynomial

so it shouldn't be too hard to create an algorithm to multiply that ecc data by the generator polynomial and sum it in various ways

though this is complicated a lot by the fact we don't have... any data

here, the generator polynomial is α0x7 + α87x6 + α229x5 + α146x4 + α149x3 + α238x2 + α102x + α21
(x4 means x⁴, etc)

oh wait fuck ecc is the remainder

so... the polynomial multiplied by any other polynomial, plus ecc is a potential solution?

hmm

if this is ascii, this should be solvable, but if not, there are too many possible values... i think?

lets see... we have 19 bytes, ~5-6 of them are known, so let's assume 13 are unknown; the polynomial has 8 terms or however they're called in english, which means we can indeed create around 2^48? possible solutions?

for example, for polynomial x+1 where we have 3 unknowns:

x+1
2x+2
...
255x+255 (after this it wraps around)

x²+x
2x²+2x
...
255x²+255x
and by doing a cartesian product, we get 2^16 possible values

so it seems the general formula of there being 2^(8*(unknown data length - ecc length)) possible values is correct

in this case, we can rely on ascii, and not much more

so even if this may not be ascii, it's useful to assume it is, because we can't get anywhere otherwise

in this case, assuming this is ascii will probably reduce the search range to 2^35... which, ugh, still is a lot

[wispy thistle]

if i take this pic, it has 10 bytes unknown, and 7 bits of ecc, so... 2^24 possible values?

i'm interested in generating just one though, simply to prove it impossible to find the answer (if it's indeed impossible)

[wispy thistle]

10 tin cans moment indeed, my math was wrong af

for some reason after i made an abstract statement about reed solomon codes of length N for data of length N+1 i thought this applies to this qr code, llm brain

so according to my new math there's not 8 possible values (which was an absurd statement), there's something like 2^24 values

[wispy thistle]

which is actually only 200mb of options

btw the chance of stumbling on one of such valid combinations is 1/2^56, so bruteforcing the entire qr is a stupid solution in this case

[safe monolith]

i thought there was something wrong, but i forgot too much to correct you

thanks for the schizo btw

[safe monolith]

I still think this is the correct interpretation, and going by what you said, which makes sense, the last block may indeed indicate this is not ascii

Multi encoding QRs can exist in theory, but for such a small qr code I dunno

[topaz cove]

qr code that just leads to another unfinished qr code

[wispy thistle]

200mb is also frankly not that much

if this is ascii, we can find the most coherent message in 200mb of ascii

if this is numbers, well rip because we sure as hell are not getting the canonical version of this qr code

though getting something that matches what we have should be a nice exercise

[tulip minnow]

New clue

Called it

Today is January 30th for me

It has a 3 in it

The 3 pattern continues

Also is that an ip address?

[daring acorn]

ip grabber

[tulip minnow]

Time to dox

Do you guys remember the pillow in the Halloween stream

What if they’re the same qr code

No they aren’t

[tulip minnow]

Also the blue screen crash reason is 0xe2

Which means it’s a manual crash

[safe burrow]

[tulip minnow]

For the bottom left corner

Idk what it is

Programmers come check that out

[safe monolith]

Slightly different approach from bred's. Also much slower than bred's cuz not Rust (too lazy ) & compares vectors and stuff (on an optimal count of processes, can do up to 3k/s on my system, while bred's does a nice 4k/s)

bred's approach fills in masked bits to check for a valid QR. I believe the QR also needs to be decodable. This script on the other hand is made to search 13 characters from abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890!@#$%^&*()+--= , generate a QR, and see which is the most similar to the qr.png based on a similarity score.

I believe it can be made faster if I made a custom factory to qrcode, but I'm lazy . By my calculations, this will merely take 77 * 13! = 479480601600 iterations to run, and if I can do 3k/s, that means 159826867s = 44,397 hours = 1,850 days = approx 5 years.

Only 5 years for a result that may not be correct

the only upside to this approach, is that it uses similarity scores, so the transcribed qr doesn't have to be 100% accurate. although the current scoring method is dumb, we can modify it so that there is a larger range between things that look similar

[safe monolith]

i kinda wanna try a wordlist

[hollow garnet]

idk what kind of words there'd be

[safe monolith]

genz slang + english + underscores?

also i made a mistake, it should be 13^77

[wispy thistle]

59379717571748521322244885951525396909802021751143365394748435217661338449728339899133

[hollow garnet]

i found this which might be useful https://www.reddit.com/r/datasets/comments/6bu1fb/top_1_million_words_on_reddit_after_analyzing/
though its reddit words so

[safe monolith]

surely achieveable in human lifetime

[safe monolith]

ok i try

[wispy thistle]

well as i mentioned above either we decoded the qr incorrectly or it isnt valid ascii

that's the real issue here

finding valid ascii isnt a problem with some math, but in this case it looks like there's 0 valid ascii

[hollow garnet]

maybe the qr is flipped in y=-x

[wispy thistle]

nah

it definitely isnt because 236 17 matches

[safe monolith]

can the math even return us ascii though, the ecc is the remainder right?

[wispy thistle]

as i mentioned above, the math can return 200mb of potential valid ascii

200mb isnt a lot

we can grep it for words like neuro

and do dictionary-based analysis

but, again, the problem is whether it's ascii at all

and if it is, then we misdecoded the image

[safe monolith]

yeah ok I follow, but this is assuming perfect transcription

[wispy thistle]

yeah and if it isnt perfect than what i wrote above (11 characters + padding) seems like the closest match

or it could be less than 11 characters

(+ padding)

either way each byte of ecc reduces the amount of possibilities by 256 bytes, which is why even though we originally have 20282409603651670423947251286016 possible values (multiplied by length 13 it's 263671324847471715511314266718208 bytes), with ecc it's 281474976710656 values (multiplied by length 13 it's 3659174697238528 bytes), then known length reduces one more byte, down to 1099511627776 values, then the ascii mode reduces by 128 to 8589934592 values, then ascii msbs reduce by 2^13 = 8192, leaving 1048576 values, which is 13631488 bytes of data, or 13mb, and though my math isnt perfectly correct, it's within a couple orders of magnitude

[safe monolith]

tbh as you've said, if it's not ascii then it'll be a stupidly hard thing to crack

[wispy thistle]

rather than hard, it will be a billion different numbers all decoding to a qr code that matches that picture

which is pointless

[safe monolith]

cuz if it's not ascii there will be too many possibilities, and probably no way to verify

yeah

[wispy thistle]

(to be fair billion is a small number for computers, so let's say trillion or quadrillion, you get the idea)

[safe monolith]

i'll try the wordlist with the similarity score approach just to test the waters

because i dunno if we can rely on an accurate transcription

it's also the same for the pillow qr code during halloween, textures and stuff make it hard to get a good transcription

[wispy thistle]

the padding is accurate as well, but the pillow thing doesnt look like a qr code because of diagonals

[safe monolith]

yeah 100% we got the padding right, it's also a telltale that the QR is at least valid

unless tomfoolery

wait a minute

is this the QR? (Bye) (Low Error Correction, Mask 7, Version 1)

[hollow garnet]

looks kinda like it

[safe monolith]

i haven't even touched the wordlist

imagine code authors actually running their code for longer than 2 minutes

couldn't be me

[wispy thistle]

yeah it looks like it

[safe monolith]

It's "Bye" btw

ok what went wrong with our transcription why is the middle part slightly different

[wispy thistle]

so i was right about the padding being longer than 4 bytes 😎

smart qr bruteforcing still sounds like an interesting topic

[safe monolith]

bro the script found this faster than the QR code i used to test it

literally no patience moment

[safe monolith]

yeah. i kinda wanna find out if there's a pattern between the padding bytes and the first few error bytes

or if there actually is pls enlighten

[wispy thistle]

don't think there is, it's padding and then ec

also I think the metadata strip (mask+ec level) is good practice for ec decoding because it's short

[daring acorn]

Maybe we should wait for the next video/stream?

[wispy thistle]

i mean we are always waiting, doesn't mean people will stop guessing

and in this case the chat got excited over a spinoff

not part of the arg

[safe monolith]

we both love and hate sadecccs

best kind of relationship

[wispy thistle]

i guess it's too late to create a google doc at this point, oh well

[safe monolith]

yeah it's fine, not related to the arg

[wispy thistle]

we did document most (all?) other puzzles by sadecccs

but yeah its fine

[onyx sparrow]

Dumb idea: can "KEY=128 bit" clue in Meaning of life be literal? Like, "128bit" is a key for something. Or keys must have specific length? (sorry, I don't know much about ciphering)

[wispy thistle]

this specifies the aes key length, like in some previous puzzles

(there's aes-128, aes-192, and aes-256, with key lengths of 16, 24, and 32 chars)

there could be additional ciphers involved

but we haven't had any ciphers with a key phrase (like vigenere) yet

and if 128 bit were literally the key, the key question is the alphabet, is it alphanumeric, is it ascii, is it numbers and letters, is it numbers, uppercase and lowercase letters? you can't really tell without trying, and given no other clues hint at anything like that, and that we already had "key = 128 bit"/"key = 256 bit" in previous puzzles and it did specify the aes key length, i don't even want to try using it as a vigenere key

not a bad idea in itself, just not really applicable to this situation imo

[onyx sparrow]

Understood. thanks for the explanation

[distant girder]

we are looking for a 16 character long key

[timid swift]

or 128 binary characters

[wispy thistle]

so far keys were ascii, but it's true, it could technically be binary

[timid swift]

true, but single tracking is a sure way to miss clues

[wispy thistle]

yep

[mint sonnet]

wait do we think the qr code is cannon?

[lost niche]

Trying to find that out

[haughty crescent]

What operation are you trying to perform? Ffmpeg should be the fastest and you can get the data into basically any structure. As for colorspace most codecs do support RGB, it's just that hardware decoders do not. You can use -pix_fmt to set a format on either input or output based on the position of the argument.

You also probably do not want to use av1 unless you need to. x264 or HEVC is better supported

[haughty crescent]

Ffmpeg can take a pipe as input.

ffmpeg -i pipe: arguments you can set fps using -r if your process is not real time.

[haughty crescent]

This is true. It is such a minor performance hit that it is probably not even trackable. However the RGB ---> YUV conversion isn't lossless.

Don't want to fill up this channel with a bunch of code talk, but I have a fair bit of ffmpeg experience if you want some help let me know. I've been lurking in the ARG thread for a while but haven't actually said anything.

[wispy thistle]

i'm talking about using ffmpeg as a library, not as a binary

[wispy thistle]

sadecccs's "Bye" qr code? no, of course it's not part of the arg

[timid swift]

100% not canon (:

[safe burrow]

But you, my sweet little ones, will cope with all obstacles because I believe in you with all my heart and soul, and with tea

[uncut cargo]

What does it mean though?