« first day (3744 days earlier)      last day (1105 days later) » 
00:00 - 11:0011:00 - 00:00

12:01 AM
Scientifically speaking, pigs have a higher chance of being the impostor
 
Scientifically speaking, what's the probability of me kicking you for making Among Us references? :P
 
somewhere in the inclusive range between 0% and 100%
 
I would have said -69%
(negative sixty-nine percent)
 
It works!!!
 
@Lyxal What's a negative 69? A 96?
 
12:04 AM
My program now generates a list of the language names and byte counts of the answers
 
I was thinking of a reply and hit enter too early
 
get nae nae'd :P
 
How about I don't
 
aight, I think I have created an optimal solution for the gelatin metagolf and thus it is impossible to beat me now
it's a bit slow but it should be hypothetically guaranteed optimal
 
12:10 AM
0
A: Sandbox for Proposed Challenges

qwrRandom Move 2x2 Scrambling A simple way to scramble a 2x2x2 Rubik's cube is to make a sequence random moves. This is not how official scrambles are done, but can get pretty close to a uniform distribution of scramble states. Due to the way a 2x2 only has two layers, doing one turn on one face is "...

 
Is heapq a standard Python library?
 
Yes
 
if not i can implement a heap myself :P
 
It takes a plain Python array and does heap operations on it
 
It times out on TIO, so I'll have to test locally
 
12:11 AM
I think I've used heapq on tio without problems
Probably your code is way too inefficient? :P
 
it's pretty standard BFS
i don't do any intelligent optimizations
it's just that the search space is somewhat wide
 
BFS = Brute Force Something?
 
More like minimal cost path or uniform cost search
 
breadth first search
it's actually not, i'm stupid
this is standard djikstra's i think
or maybe not. i just remember how to implement it, i don't remember what algorithms are called anymore
 
Simple BFS does not need a priority queue actually
 
12:15 AM
data.map(p => groups[p[0]] ? (groups[p[0]][p[1][0]] = Math.min(p[1][0] in groups[p[0]] ? groups[p[0]][p[1][0]] : Infinity, +p[1][1])) : (groups[p[0]] = {[p[1][0]]: +p[1][1]}));
Readable™
 
But looking at the parsing rule of Gelatin, I guess a metagolfer should use a priority queue because the tokens can be added in chunks
 
i just have each of the five types of groups as a chunk (so, length anywhere from 1 to 3) and add them like monads
 
Yeah, I thought of that exactly
 
Testing it locally, it looks like it fails for one of the scoring inputs. I'm still testing, but something's up with it
Yep, got a memory error :/
 
as long as i don't put a 0,2 or monad after a dyad, i can just treat each chunk as a monad because of how the chaining rules work, and thus just pq with the length first
 
12:18 AM
The only way to combat memory error is to prune the search tree I guess
 
i guess i could try to optimize, like not adding _0 for example
i'll see if i can figure anything out
wait, _0 shouldn't even be getting added because i have a seen set to prevent issues like this
 
If there's no obvious way, you'll need to do something more aggressive, likely losing strict optimality
 
It threw a memory error for test cases where m-n is 517 or larger
That's not a lower bound, just a value I got from the scoring cases
 
There are around 2 million pairings of solutions on this site (in the same challenge)
 
i might need to change my ordering algorithm to try to focus on getting closer to the end value first at the cost of optimality
 
12:27 AM
Yeah, I don't think I can give you a score if it errors out for 6 of the 20 scoring cases
 
oh it's that bad? :/
 
Yeah :(
Just an FYI, I added in a "must be somewhat efficient" requirement (i.e. I must be able to score it), so your answer is (currently) invalid :(
I dislike disqualifying answers or changing the rules, but that seems like a pretty important one
 
0
A: Sandbox for Proposed Challenges

qwr[Draft] 2x2 algorithm with half the faces A fun fact about the 2x2x2 Rubik's cube is that due to the way a 2x2 only has two layers, doing one turn on one face is indistinguishable from doing the same direction turn on the opposite face and then rotating the cube. So algorithms, that is sequences ...

 
being able to run the code for a test-battery is somewhat important :P
it would be unreasonable for me to keep it around lol
 
12:34 AM
@RedwolfPrograms oh
That's a lot
 
At least you can score it as infinity if it doesn't work for some inputs
which makes it scorable and therefore technically valid? :P
 
@cairdcoinheringaahing I've updated my code; does this work now?
i'll undelete once it's confirmed to work since you can see it anyway :P
 
Lemme just run it (takes a few minutes) then go to bed cause it's past 1am here and I have to be up at 8am :/
 
Wait, I think I saw something important
 
is there any way to just output a constant whatever the input is?
 
12:37 AM
I have to do this
 
@hyper-neutrino Yep, scores 238
 
@cairdcoinheringaahing just give me the test data /s
 
Finishes almost instantly :P
 
technically you could do _~ first and then just work up from 0 for every single thing
my super intelligent heuristic is uh
 
@rak1507 No, as digits are single digits (i.e. 10 is 1 and 0, not 10)
 
12:38 AM
multiply the absolute difference to the desired value by the length of the program cubed
 
@cairdcoinheringaahing ah
 
@hyper-neutrino if you want to undelete and add its score (238), feel free
 
alright, thanks :)
 
Should Perl 5 and Perl be different languages? (Same for Java 8, JavaScript ES6, Python 3.8, etc.)?
 
Perl 5 should be a subset (so include the Perl 5 answers with Perl, but also show Perl 5 by itself)
 
12:41 AM
The site policy says different version means different language, but it's complicated when the version is simply unspecified
 
@cairdcoinheringaahing The extra clutter in the list might be annoying though
 
Alright, I need to go to bed, so if anyone needs their program testing for my metagolf challenge, it'll have to wait until ~9am UTC
\o
 
just discovered a bug in my program but I also have to go to bed, this is a nightmare
hopefully I dream the solution
 
12:44 AM
instead of sleepwalking, here on cgcc we sleepcode
 
I once made an esolang based on a dream
 
@hyper-neutrino I have actually dreamt golfs for Jelly answers before
 
lol, nice :P
 
I took a hard look at my life afterwards, but I saved 2 bytes so yay
 
Is Excel VBA the same as VBA?
 
12:59 AM
@RedwolfPrograms part of
but a really small part
 
1:51 AM
@OldSandboxPosts my thoughts exactly
...and any feedback?
 
@Wezl 1) I don't really get what the task is 2) Frick
@Wezl No problem I can think of. To include or exclude ^ is entirely your choice
 
2:11 AM
You have 9 hours to get ready to gwift I'll probably post it tomorrow morning then
morning in EST of course
 
2:52 AM
@OldSandboxPosts assembly function has already been posted
 
there's probably a way to optimize gelatin metagolf by working backwards
if we imagine the search tree branching out like a triangle then it should cut the search space into a quarter on average (though with exponential growth it might be even better)
too lazy tho. i'll be mildly disappointed if all submissions are just minor tweaks and optimizations to my djikstra's but oh well
 
all shall be hyper neutrino
 
3:09 AM
@hyper-neutrino I strongly believe mine is 99.9% close to strictly optimal, so I'll be very surprised if there's still room for outscoring
 
yours is definitely a much smarter optimization than mine which is just a randomly guessed heuristic :P
 
3:47 AM
Indonesian poker spam is striking again :/
 
Is it already yeeted?
Nice
 
yet another post nuked by flags before i could get to it; i think smoke detector put up at least 4 of the flags though
SD is just too good :P
 
What, it can flag multiple times on a single post by itself?
 
3:52 AM
no, there are a lot of users who have signed up for autoflagging (including myself)
 
I'm so irrationally nervous that it's somehow May 29th and not April 29th or the due date changed again and I should have finished my computer science projects a month ago lol
 
basically you give SD an write-access API token and when it is has enough confidence something is spam it will use itself and a couple other autoflag accounts to raise around 3-4 flags on the post
 
4:19 AM
is there a good algorithm to convert into complex bases? (number system uses sympy not python native numbers fwiw)
and also would it even be worth implementing into a golflang
 
Like, base 1+2i?
 
Can I get one more upvote on this so my NMP bot can access chat?
Thanks!
 
0
A: Sandbox for Proposed Challenges

Matt SanchezDrawing the Stack Overflow logo I thought of making this a monthly game. You will be using C as the language for this code golf challenge. The goal is to create the Stack Overflow logo using ASCII characters. It looks like this. Can you get it with the least bytes possible? Tags: code-golf, asc...

 
I'll add those credentials in at some point tomorrow and join it up with my New Posts port
 
@Bubbler yeah, smth like that for example
 
4:38 AM
The problem is under-specified because you need the set of digits to use as well as the base, according to wikipedia.
 
ah, right that makes sense. i guess [0, x) & Z doesn't really make much sense in this situation
 
If you don't mind getting some non-real digits in the result, Dyalog APL supports that, which uses the concept and algorithm described here.
The guarantee is that each digit is a Gaussian integer, and its magnitude is strictly smaller than the base
TFW I cite an APL Wiki article and Adám enters the chatroom...
 
Now I just need someone to post a new question so I can test out my NMP clone :p
 
To main?
 
Yes
It worked
It's instant!
 
4:43 AM
@Bubbler No connection. I'm just getting up.
 
I know. It just feels weird :)
 
This bot is so much better than NMP
(Also it's super insane that we got two posts within ten minutes of me hoping someone would post something :p)
 
@RedwolfPrograms Your hope is granted even though I didn't even look at chatroom before posting
 
aha, finally got in a review task before someone else sniped it
 
@hyper-neutrino Crap, is that mine one?
 
4:55 AM
no. it was a first post from a new user
although, regarding your post @okie, is it okay to output nothing when X=0, or must we still output "0"
 
output 0
 
alright
 
Looks like HN is trying to become Arnauld but in Python
 
wdym lol
 
Ouch i think i just edited my question with @hyper-neutrino editing at the same time, which i think canceled his edit
 
4:59 AM
Apr 25 at 1:48, by Lyxal
How many times do I have to say this? Arnauld is a machine learning AI that is designed specifically for answering code golf questions in js
 
@okie looks to me like it still went through
the edit about the base being positive?
(by the way "positive" automatically implies "non-zero" :P)
right, but i am also able to answer code golf questions in jelly, and also not nearly comparable to someone like arnauld at either :P
 
@hyper-neutrino Since you solve it first, do you think this question is easy?
 
it's pretty easy
 
Then it's good
I designed it to be an easy question for **new people in code golf** to answer
you just hurt new code-golf member(?)
 
2
Q: Print the Najdorf Sicilian

EnderShadow8Print the Najdorf Sicilian The Najdorf Variation of the Sicilian Defence is an opening often played at the highest level of chess. It is named after GM Miguel Najdorf known for being extremely sharp and theory heavy. Your task is to write a program that, given a Boolean-like input, prints a chess...

1
Q: Base-ically god

okieStory The god of base is opening heaven’s gate for mortals to join the base party. You, as one of the believer, wants to follow his step to ascend to heaven. But not everyone can get in. The god posts this question to let you prove you are worthy of his party. Task You will get 2 inputs First be...

 
5:05 AM
Finally caught up?
Poor NMP
Still continuing on, unable to work fast but still putting in the same ceaseless effort it did five years ago, unaware of its approaching fate.
 
You guys are coldblooded, even simple question is not spared and blatantly murdered
 
Welcome to TNB, @lefrost!
2
 
You forgot @user103574
Welcome to TNB, @user103574!
 
Please don't ping them twice. Also, they can't chat here anyway.
 
Sorry
Didn't meant to ping them twice
Can I make a upgrade version of a question, making it looks like a series?
I don't the answer is yes
but i want to give it a try
 
5:16 AM
It's fine, as long as it adds something interesting
 
Ok
 
I'd sandbox it though to make sure it's not too close
 
5:38 AM
every time Arnauld posts a JS answer there's pretty much no point in answering because he almost certainly posted the shortest possible, maybe I should use a different language
 
@ophact There are gazillion different languages to choose from besides JS :)
 
@ophact not true. Even the best golfers in each language can be outgolfed. xnor even encourages it with their bounties. You can get good to a point where you can challenge arnauld in JS. All it requires is persistence
 
Right, ill try to learn a golfing language (just for fun)
 
pretty much all of the mainstream languages right now have a bunch of top competitors
and it's not up to me to tell you what to do but you could either pick up a super obscure language or see it not as discouragement but a target to beat and someone to learn from
 
Not really, mainstream is pretty broad and many "mainstream" languages have only one or two users using it regularly
 
5:43 AM
actually mainstream is not well defined
and now that i think about it, in my mind i was probably thinking "languages that are used by a lot of people", which makes my first message entirely meaningless
 
Lol
 
in terms of cg&cc something like jelly is a lot more mainstream than, say, php
 
"Languages that are used by a lot of people" is very different from "languages that are used by a lot of people on CGCC"
 
who was the really really good PHP golfer again? i remember there being a user (maybe multiple) that specialized in php
 
I actually have my own language but it can barely do anything useful
 
5:44 AM
Yeah, Jelly is pretty mainstream on Code Golf :D
 
@ophact all the more fun to be had trying to use it
 
and 05AB1E
 
i wonder what ophact is going to learn
(psst, APL)
 
@ophact Try solving some of the most trivial challenges (Hello World, truth machine, infinite loop, etc) in it
 
depending on the capabilities a quine could also be either fun, or traumatic
 
5:47 AM
(and actually Hello World can be nightmare too)
(For the record, there's no HW answer in Flurry yet)
 
there's no flurry string metagolfer yet?
 
> flurry is like brain-flak
> evaluates to the K combinator
beautiful
 
@Razetime Not yet, and I don't think it will ever have one in the near future
 
i am still not able to get my head around lambda calculus
 
Have you tried Haskell?
 
5:53 AM
but it shouldn't be hard to translate an existing SKI answer
 
i had to learn lambda calculus for my cs class, lol
 
@Bubbler I haven't gone deep into haskell
i understood some of the functions for husk reasons
 
conceptially lambda calculus is very simple, it just looks like hell when you actually do things with it
 
But lambda calculus is also a topic for one-semester CS course
 
i feel like just like brainfuck, the simplicity in excess is what makes it so complicated
 
5:55 AM
^
 
(and I had to use OCaml for three different CS courses)
 
@Bubbler I guess i will wait for my course to start on that(I doubt it ever will)
 
know nothing about ocaml but standard ml is fun
 
Oh right, actually one of them was standard ml, not ocaml
@hyper-neutrino And lambda calculus mixed with combinators and side effects on the stack is just F crazy
 
lambda calculus with side effects sounds very un-fun
 
5:58 AM
yeahhhhhhh
 
am i limited to 20 review tasks per day? if so, why, if not, why does it say so :P
 
i'm wondering how to solve the flow free question in APL
 
Keeping pure in Flurry is possible (besides the stuff for I/O) but it's really verbose
@Razetime Good luck :P
 
@Bubbler I'll get there in a month
 
6:19 AM
i'm trying to understand sk lambda calc and i think i'm about to run out of neurons for this and i'm only like halfway through
 
Lol
 
ok i barely understand this; who on earth was big brain enough to even come up with this stuff in the first place
aw, flurry isn't on TIO
 
Yeah, you need to download the Haskell one (which is the reference impl)
or use my online interpreter (which has shortcuts for integer arithmetic but doesn't work for some programs because of that)
 
might include it into my interpreter
 
And in order to write meaningful programs, you also need to learn about Church numerals
 
6:32 AM
church numbers aren't too bad though
in my cs class we had to represent numbers as lists (using cons structures) of true/false i.e. as binary lmao
 
Well, I once got that kind of assignment too
but manipulating ADTs is much easier than manipulating Church encoded things after all
(or Scott encoded things for that matter)
 
huh, never heard of scott encoding before
 
6:47 AM
Hi, all. I hope you can see this deleted post:
I will repost to sandbox in 40 mins
With the question of whether its a good fit for the site
the final version of it I think will have a link in it to my own github repo of my personal solution which is the benchmark for correct solutions
I will also provide a formal proof of the algorithm so you there's no discrepancy of whether it provides the correct algo outputs.
Feedback?
 
I was about to ask about it (if there's a single canonical solution for every possible input). If so, that's good
 
There are many possible solutions, but my code will enumerate all solutions (up to standard form)
So maybe I should change it to provide any one solution (which is actually the original SGP)
 
Also, the winning-criterion tag for measuring time of submitted programs is , not
 
@Bubbler thx
fastest-cde
*code
So do you think it should be the original SGP - one solution will do?
If someone comes up with a polynomial-time solution. They win the P = NP millenium prize, lol
 
I guess finding one solution might be better (more room for algorithmic improvements?) but not really a problem either way
 
6:55 AM
It's an NP-complete problem afaik
 
And you need to prepare the inputs for testing before you post it to main
 
Okay, cool. It will take some more coding for me, maybe a month
before I have my fantasy code done
But I'll post it to sandbox with your suggestions in 35 mins
@Bubbler what are your thoughts about the SGP in general?
I found only one really great paper on it before, but I can't find it atm
I will try to recover that paper before posting to main
It had actual bounds proved in the write-up
 
Well, almost any NP-complete problem is interesting for running-time optimization
 
Do you think we should use a language other than Python?
 
Probably the biggest problem in the challenge is that the language is limited to Python 3 though
 
6:59 AM
I thought python would be best for ease of coding
 
Usually the challenges are open for all languages
and Python is a pretty slow language
 
That's difficult on my side then because I don't have the ability to build all languages so they would have to include build instructions.
Yes, but if your algorithm is faster in python it will be faster in any language
 
You can explicitly ask for build instructions in the challenge text
 
Yeah, but then again I have to go through all of them lol
What if I just said C++11?
 
And if the Big O is lower-bounded, constant-factor speedup is very important, which makes the language choice relevant
 
7:01 AM
It's about the algorithm though
 
But you're measuring the running time of programs I think
 
Okay, should I say just provide proof of your algorithm and the best O() wins?
 
That works if all you want is best big O
 
That way we wouldn't have to build things, but I could have a long example input and they have to pass back to me the long example output, which I can then just verify by file difference
 
and then it's indeed
 
7:04 AM
Cool cool
 
huh. i may have just denied myself a First Posts review by deleting the post too fast for the queue to pick it up
 
So I guess I'll do a dual verification. One by formal proof and one by example in/output
 
That's good enough I think
 
Thank you!
I'm excited, because this is new CS territory. No one has written or at least published this algo (an exact version)
I mean *an exact algorithm
 
Just a note that designing algorithms is not a prominent part of our site (though it's on-topic) and not so many users are interested in that kind of challenges (though we do have a few excellent algorithmists and computational mathematicians)
 
7:13 AM
@Bubbler what I'm having trouble with is how does the large file in/output verificaton work if I provide code on github?
They could just use the code to cheat
 
Our site rule requires answers to include code that solves the challenge, so when in doubt you can always run the submission yourself
 
I think that's the only way, to run the submission
 
7:31 AM
hello
sup everyone
is link to the official homepage of a language necessary when not giving a link to its online interpreter?
 
I believe yes
because anyone seeing the answer should have a way to run and verify it
 
0
Q: The Smallest Grammar Problem - no one has published an exact smallest grammar algorithm yet, as far as I know; here is a competition proposal

StudySmarterNotHarderHere is a write-up for the main site. Some changes I want to add are: You can use any language. You must formally prove the Big-O of the worst case of your algorithm in your answer. You must formally prove the correctness of your algorithm as well. As far as running the code, that's up to the v...

@Bubbler I posted it
 
@StudySmarterNotHarder hey you need to post in sandbox
not as a meta question
 
oh oops
where's that
 
@Wasif Either an official homepage, or a page on Wikipedia or esolangs
 
7:33 AM
@StudySmarterNotHarder head to sandbox and post it as answer here it is:codegolf.meta.stackexchange.com/questions/2140/…
don't forget to delete the meta question
 
@Wasif thank you. Deleted
0
A: Sandbox for Proposed Challenges

StudySmarterNotHarderThe Smallest Grammar Problem Here is a write-up for the main site. Some changes I want to add are: You can use any language. You must formally prove the Big-O of the worst case of your algorithm in your answer. You must formally prove the correctness of your algorithm as well. As far as runnin...

If anyone wants it
 
welcome :-)
 
It's not the best of write ups, but I willl do better on the final one. Still interested in where the users don't understand something
regarding the SGP problem
 
also you don't need to permalink it
it would automatically notified here through a new sandbox posts bot
 
7:42 AM
0
A: Sandbox for Proposed Challenges

StudySmarterNotHarderThe Smallest Grammar Problem Here is a write-up for the main site. Some changes I want to add are: You can use any language. You must formally prove the Big-O of the worst case of your algorithm in your answer. You must formally prove the correctness of your algorithm as well. As far as runnin...

 
7:52 AM
how does one prove turing-completeness of SKI?
 
2
Q: Phibonacci - Relation between Phi and Fibonacci

EliteDaMythYou want to see how quickly the ratio of two consecutive Fibonacci numbers converges on φ. Phi, known by the nickname "the golden ratio" and written as \$φ\$, is an irrational number, almost as popular as π and e. The exact value of \$φ\$ is \$\frac {1 + \sqrt 5} 2 = 1.618...\$ The Fibonacci sequ...

 
@hyper-neutrino By proving TC of lambda calculus and proving that any lambda term can be converted to equivalent SKI expression
 
ah, okay
 
According to this, TCness of lambda calculus was proved by Turing himself, apparently
 
oh wow, that's cool
 
8:05 AM
Lambda calculus is a pretty nice tool of proving TCness of subset of a practical language, since it's basically how anonymous functions work
When it is untyped, constructing S and K is sufficient. Otherwise you also need to provide Y
 
8:16 AM
ah. what's fix then (or whatever it's formally called)?
there was some challenge about making a TC subset of a language with minimal characters and someone constructed SK in haskell, but since haskell is strongly typed, apparently SK is insufficient
 
@hyper-neutrino fix is the same thing as Y, which is defined as fix f = f (fix f). But in strictly evaluated languages (e.g. OCaml, Purescript), you need to define it as fix f g = f (fix f) g instead
 
ah, okay i see. thanks.
yeah i've used (lambda f: lambda g: f(f, g))(lambda f, x: x / f(f, x))(v) in python a couple of times when trying to one-line things lol
 
That's called overcomplication :P
 
yes :P
 
8:47 AM
fix is effectively recursion right
 
pretty much
it lets you recurse without being able to name the function because otherwise the anon func can't reference itself
 
0
A: Sandbox for Proposed Challenges

Anders KaseorgUnlock Tim Cook’s iPhone This is Tim Cook, CEO of Apple. Your challenge is to write a program to output a picture that will be recognized as the same person by a facial recognition model. (Photo attribution) Test your picture using the face_recognition package—we’ll use the current version as of...

 
9:33 AM
@hyper-neutrino In APL, we use a symbol () for "self" so anonymous functions can recurse.
 
oh that's convenient to have
 
9:46 AM
waiting for in tacit functions
 
10:08 AM
@Razetime J has it, but there are interesting scoping issues.
 
NMP come on...
 
It can take up to 30 mins.
 
@hyper-neutrino last time I saw you you didn't have that diamond:)
oh btw did you see the haskell channel on th ppcg discord? (also @Bubbler)
 
4
Q: Generate all solvable 2xN mazes

A usernameYour job is to write a program that takes a number N as input and outputs all 2-by-N mazes that are solvable. Output format can be in the form of any two distinct values representing wall and empty - I use x and . here but you can use anything. You can have a matrix, array, string, ascii art, wha...

 
@Razetime Looks good!
 
10:21 AM
@xnor how do I apply a random hex color to an arbritrary string in python
 
@RedwolfPrograms I never got a welcome :P
 
:57830772 anyone know how to apply a hex color to a string in python
 
@Ausername thanx
 
@StackMeter Do you mean in console output?
 
@Ausername no, just to apply it to the string, such that I can store it
also is this valid for the "Is it true? Ask Jelly!" question: ````def a(n):
return n and "0" not in str(n)````
 
10:26 AM
@StackMeter Define "apply it"
 
@Adám but having it is half the struggle
 
@StackMeter I think so.
 
@Ausername if I were to then print it, it would have the hex color that we gave it - however, I can also assign that same string to a variable, and print that out, and it should display the text with the correct colour
 
@StackMeter There doesn't seem to be a way to do that - maybe create a class ColoredString and give it a color, value, and custom print method?
 
I'm really surprised how popular the XKCD KotH challenge is
 
10:34 AM
Your username is rendering as invisible.
 
@Ausername mine?
 
That's exciting. It renders for me though
 
@Lyxal Can you see @AncientSwordRage's username?
 
no repro
 
10:42 AM
woop
 
I have been unable to reproduce this issue
 
That's two good things so far today
I have a name, and I have a working answer to a challenge
 
therefore, I must mark this as
3
 
I'm on Firefox 87.
 
10:57 AM
@Ausername yes but I'm on mobile
 
00:00 - 11:0011:00 - 00:00

« first day (3744 days earlier)      last day (1105 days later) »