The brutal ones are being saved for the weekend aren't they

File: tabular.png (118 KB, 708x580)

118 KB PNG

@cachefags need not apply

>>103574305
"Eric said it's okay to use AI after the leaderboard fills, which it has"

>>103574304
manual cache is not better than automatic cache

>>103574317
i am not using manual cache; i am using a list

>>103574330
that list is a cache

>>103574305
>all eric needed to do was to add a few slurs into the puzzle

>>103574334
yes, but @cache is not a substitute since i solve iteratively

File: im-1734617118087.png (55 KB, 692x437)

55 KB PNG

>>103574304
top-down pytoddler yields to superior tabular approach.

>>103574363
it's you again; your terminal emulator is really nice
yield accepted

eric dp my anus

File: 1720417967222186.jpg (39 KB, 1365x767)

39 KB JPG

had to look at how to solve part 2

File: carbon (10).png (266 KB, 1818x1284)

266 KB PNG

Eric has given up or what's up with these past 2 puzzles? These would be days 5-10 in past years.

>>103574429
you could tell me today was the 2nd - 5th and i would believe you

File: nick-point-laugh.jpg (31 KB, 533x475)

31 KB JPG

>>103574407
filtered

>>103574372
oh no i wasn't the one bickering with you about the dp solve, but i do recognize the superiority of tabular. some other coping anon.

>>103574471
no i mean i remember your terminal from previous threads; some other anon asked about the scan line shader / background

File: file.png (1.02 MB, 1739x1232)

1.02 MB PNG

>>103574458

File: terminal-bg-pinstripe.png (145 B, 10x8)

145 B PNG

>>103574483
haha yeah it's literally this

File: nickthespick.jpg (70 KB, 651x1024)

70 KB JPG

>>103574486
say nick the spick ONE MORE FUCKING TIME

>>103574304
>I am superior for doing more manual indexing

>>103574305
just tell it to roleplay as the grinch

>>103574538
you have to pass the state `(i)` in your recursive solution too, dumbfuck

do you people even enjoy doing this

>>103574573
At the very least it's something fun to talk about for the season

>>103574573
yes i enjoy Christmas puzzles

>>103574573
this year isn't as fun

>dpdpdpd
>topdownbottomuptopdownbottomuo
aaaaaaaaaaaaaaaaaaaaaaaa fuck you fuck you fuck you fuck i hate dp so fcukimg much aaaaaaaaaaaaaaaa

>>103574602
>dp memoisation top-down bottom-up push pull
cry more, faggot

>>103574567
Thankfully, Python strings have lots of methods that let you compare and manipulate strings without having to do manual indexing.
What is it with the "people" on this board managing to brag about the most retarded shit possible?

>>103574573
I did last year, but this year was dogshit

>>103574614
>his state is a sliced string
are you actually retarded?

>>103574619
this year is so boring I've mostly been going back on previous years and doing those instead. the christmas tree one was mildly entertaining for the visual aspect alone and the reverse engineering one was ok too

>>103574637
i think the reverse engineering one was the best by far

>>103574637
>>103574619
>>103574585
Filtered

>>103574724
can you faggots shut the fuck up please

>>103574744
Filtered

>>103574630
Obviously, it's the natural representation of the problem and even lets me reuse cached values for different input lines with zero effort.
And it somehow runs about three times faster than your gay solution, huh. Yes I actually tested it because you're a faggot.

>>103574770
>contributes nothing and shits up the thread
fucking kill yourself

File: baseddepartment.png (358 KB, 600x1079)

358 KB PNG

>>103574770
>contributes nothing and shits up the thread

File: carbon.png (989 KB, 1700x4604)

989 KB PNG

Fairly easy day, but had a long runtime before. Wrote a simple trie, runs in ~400µs.

>>103574649
not a high bar to beat when the rest are all snorefests

>>103574847
Good time. I got like 35ms using DP. I wonder whether the difference is in the approach, or the fact that I am constructing lots of strings. Should rewrite using sub string views to avoid memory allocations.

>>103574407
How is that possible if you already solved the day with the multiplying stones. You DO understand the algorithms you are implementing right anon?

File: codeimage-snippet_19(1).png (185 KB, 997x962)

185 KB PNG

>only need to check once per design index

It was really that easy.

>>103575146
You don't need to solve a day to try another

>>103575185
disgusting in every way, irredeemable

>>103575210
In that case you can go back and solve the stones now.

>>103575220
>t. used DP or caching as a crutch

File: tinytim.jpg (281 KB, 953x1390)

281 KB JPG

>>103575262
call me tiny tim bitch

File: 1728072886088731.png (879 KB, 2000x2000)

879 KB PNG

>>103571781
It took me a few tries to get a formulation that solves it in a reasonable time, but it's definitely doable with FOSS solvers.
Pic related solves both parts in about 10 minutes on my shit computer. The main issue seems to be that the search space is very symmetric (lots of valves have zero flow rate, for example), and the only "manual" optimization I did is to not even try opening those. I assume Gurobi / CPLEX would do much more automatically.

File: code.png (244 KB, 1282x1988)

244 KB PNG

Python

There is no need for cache.

Can someone explain the actual solution for part 2? I'm just trying to find unique tokens so I can get the factorial. So far I'm pretty sure its lower than 25! so I'm thinking its probably something like 10!-15! but I don't get the actual solve.

for a in designs:
    for b in designs:
        if a + b in alldesigns:
            if a + b in designs:
                alldesigns.remove(a + b)
        for c in designs:
            if a + b + c in alldesigns:
                alldesigns.remove(a + b + c)
            for d in designs:
                if a + b + c + d in alldesigns:
                    alldesigns.remove(a + b + c + d)
                for e in designs:
                    if a + b + c + d + e in alldesigns:
                        alldesigns.remove(a + b + c + d + e)

>>103574953
I got ~25ms with a queue, no new strings. Guessing it's the graph approach that makes the difference.

>>103575317
Is this bait

File: 1577742904314.jpg (149 KB, 1348x1470)

149 KB JPG

>>103575317
chinese remainder theorem

>>103575344
Obviously

>>103575344
>>103575373
holy shit I totally missread the problem my bad

File: hm.jpg (192 KB, 1330x1056)

192 KB JPG

>>103575317

Remember anon, 90% of brooters kill the process and start trying to do something "intelligent" just moments before the broot probably would have finished...

File: 1713024016199382.jpg (73 KB, 680x588)

73 KB JPG

>>103575560

File: Screenshot 2024-12-19 at (...).png (348 KB, 1280x1308)

348 KB PNG

Recursive solution in Swift.

>>103575587
cool go run it on an ipad dumbass fapple manboi

File: carbon.png (170 KB, 1268x954)

170 KB PNG

>>103574953
This pretty basic hashmap-backed memoization takes ~14ms, I think I got it down to around 6ms by using a vector instead of hashmap before I wrote the trie. Also, it checks if the current substring is in towels instead of .startswith().
>>103575320
I tried an existing trie but it was even slower than picrel, because it was allocating all found prefixes. So I think it's also important to keep the trie lean. Returning an iterator instead of a vector and using fairly small array without too weird of an indexing both helped a lot.

File: codeimage-snippet_19(3).png (109 KB, 958x559)

109 KB PNG

>>103575185
realised how much I was overcomplicating it

File: tim.png (78 KB, 234x296)

78 KB PNG

>>103575777
that's better
>t. tiny tim

File: unnamed.jpg (49 KB, 512x341)

49 KB JPG

wtf I'm still not filtered and its day 19. this is actually the furthest I've made it. I can't tell if I got smarter or this year is just way easier.

>>103575863
this year is easier imo

>this year is easier
you've been saying this every year, for years

>>103574248
>spent an hour looking for a bug
>one variable was an int instead of long long
i hate myself so much it's unreal

>>103575929
I think it is true, but also a function of returning anons getting better at solving them so they feel easier even though they may have struggled with these same puzzles in the past.

>>103575941
I was filtered by the amphipod tunnels in 2021. I could probably complete it now easily.

>same puzzle as last year
>in the same hotsprings
Eric doesn't even care anymore.

File: maxresdefault.jpg (81 KB, 1280x720)

81 KB JPG

Is there a word for the thought process where you solve every problem by adding more complexity, end up with something retardedly overcomplicated, and only when you step back (i.e. after posting) do you realise how dumb it is?

>>103575982
tunnel vision, decision fixation, etc.
having high energy without perspective is like missing your turn while flooring it in a race car. Congrats, you've got a really good engine, but you're wasting it and you'll be late.
>>103575935
hate yourself enough to use Zig? It's super anal about integer types and casting, and you would've gotten a runtime error from the overflow.

>>103575982
It's called not cheating. Ever notice that the diversity of solutions significantly decreases as the day goes on. That's because people are cheating and filtered.

I may be filtered but I'll keep going and learn from other's solutions. I will save christmas.

This year feels like when the main character in Idiocracy was taking the IQ test.

>no tree problems
>No cellular automata
>no wall of text follow 100000 different rules problems
>Just grid problem, grid problem, leetcode problem, recycled problem
Eric has given up

>>103576225
This is what day 18 was like. It set up an interesting premise and then hit you with
>how many buckets do you have?

>>103576233
There’s still 6 days left. He’s saving the best for last… right?

>>103576261
he's saving the last for the last
these feel like the rejected problems from previous years

>>103576308
kek he's dumping all the trash upon us in the last year

>>103575317
You need more loops.

>>103575317
depth-first search with memoization. That's it.
Write a recursive function over a string. If the string's empty, return true. If the string's not empty, find all valid prefixes and recurse on the string absent that prefix, and return the sum of the true values returned from that recursion. If there are no valid prefixes on a non-empty string, return false.
This will solve your example but be as slow as obsidian on your input. Slap a cache on the recursive function and it'll be fast.

File: 19.png (109 KB, 687x943)

109 KB PNG

Babby's first dynamic programming problem. Hoping for something crazy this weekend. Slightly washed C++, mostly squished to fit onto my screen for a picture.

File: 1570617259427.jpg (5 KB, 164x202)

5 KB JPG

>part 2 works for example, not for input
>spend two (2) hours trying to find a corner case bug in the code
>the problem was using int instead of long
FUCK

>>103576261
i hope an even harder version of the elf vs goblin puzzle

File: file.png (2 KB, 128x48)

2 KB PNG

>read part 2
>get PTSD from last year (filtered by memoization)
>go to sleep
>wake up, make coffee
>solve problem
FUCK YOU ERIC SUCK MY DICK BITCH

>>103576640
i'm doing that on and off and i'm guessing p2 is "figure out the minimum total health the elves need to win" or something, how close am I?

>>103576233
Cellular automata is tonight's problem.

>>103576100
>That's because people are cheating and filtered.
Or it's because the only people left actually know all the relevant data structures and algorithms.

>>103576640
Part 1: Elves vs goblins but you have to find the shortest path the reindeer can take to collect all the christmas orbs on your map before the goblins kill all the elves. Goblins and elves run away from the reindeer when he's within 3 tiles (manhattan distance). Each christmas orb on the map is a digit 1-9 that heals all remaining elves for that amount. Goblins, elves, and the reindeer cannot occupy the same tile.

Part 2: what is the maximum number of turns at least 1 elf can survive for without collecting any orbs? Turns out evil santa corrupted the orbs so they cause a 1 health/turn loss of all units within 5 tiles (manhattan distance). (Input designed so you have to use the fleeing mechanic, but also so that the simulation eventually ends by causing the last elf to flee into another goblin or degen zone)

Just redid 2023 d12 and was surprised by how hard it was. Much harder than today's problem.

>>103574784
actually the minimal representation of the state is a single integer
you failed to make this observation, so you're actually filtered; better luck next year

>>103575726
Adding a custom substring implementation (to avoid allocations) with custom hash function brought the time down from ~32ms to ~22ms. Your trie still wins. Or maybe C++ unordered_map just sucks.

>>103575777
>(&design[dx..]).starts_with(&towel)
you don't need the manual borrow

>>103576233
Could you create 500 interesting problems over 10 years?

>>103577000
Eric never had to. That he won't accept puzzle suggestions was always pointless and self-inflicted. Even if his mods are much more nuts about it, the IP fixation comes from Eric hiimself.

Wasted a few minutes trying to debug the answer because it looked way too big.

>>103576967
You're right. Leftover from before I realised starts_with was a thing.

File: 1708032521139740.png (583 KB, 2100x1158)

583 KB PNG

>>103577021
imagine a world where people can upload input.txt to github and submit puzzle ideas

>>103577021
whatever man if you made something really cool you probably wouldn't want a bunch of jeets and chinese stealing it either.
you wouldn't want a bunch of trannies creating committees to discuss the upcoming years' problems, and how to make AOC more inclusive, or whatever they do when trying to ruin everyone else's fun.

>>103577031
after solving part 1 (with an answer around 200), I quickly wrote something for part 2 and tried on my input without testing against the example, and I got something around 2500
it was wrong of course, but when I went back and made sure I had code that worked against the example, and THEN ran my input, I got a huge number (around 10^14) and thought, well, guess I'll try it and see if it works
was quite surprised when it did. my first attempt definitely was not thorough apparenlty

>>103575726
I used such a hashmap memoization. That vector memo approach is genius. How did you come up with that? Does that approach have a name?

>>103576665
WTF anon, is just a hash map.

C++ 23 + good old macros.

#include "aoc.h"
static V<Sv> patterns;
static M<Sv, N> sols;
static N count(const Sv d) {
    def it = sols.find(d);
    if (it != sols.ω) return it->second;
    N n = 0;
    for (def p : patterns)
        d.starts_with(p) && (n += p.ϟ == d.ϟ ? 1 : count({d.α + p.ϟ, d.ω}));
    return sols[d] = n;
}
int main(int argc, const char** argv) {
    MmFile in(argc >= 2 ? argv[1] : 0);
    def input = in | "\n\n";
    patterns = input.h / ", ";
    N silver = 0, gold = 0;
    for (def d : input.t / '\n') silver += count(d) != 0, gold += count(d);
    $o silver, gold;
}

>>103577226
nta but hashmap memo stuff is bananas. ive only ever seen / used it with aoc stuff. its actually so much more understandable than tabular dp for me. but I don't even know how the space complexity compares for junk like this

data = list(map(int, data.strip().split(" ")))
tab = defaultdict(int)
for stone in data:
    tab[stone] = 1
for i in range(75):
    step = defaultdict(int)
    for stone, count in tab.items():
        if stone == 0:
            step[1] += count
        elif len(str(stone)) % 2 == 0:
            power = pow(10, (1 + math.log10(stone)) // 2)
            step[int(stone // power)] += count
            step[int(stone % power)] += count
        else:
            step[stone * 2024] += count
    tab = step
    if (i + 1) % 25 == 0:
        print(sum(tab.values()), end=" ")

>>103577346
>ω

anon that isn't APL

>>103577021
>reddit mods
>power over the most meaningless shit going to their heads
tale as old as time

>>103577457
Why waste time typing five characters when I can AltGr + w to get ω?

>>103577292
I know that now. But when I didn't know that I had ten pages of notes and my desk looked like a schizophrenic basement

can I stay awake for another 8 and a half hours?

File: 2024-12-19-152612_692x72_scrot.png (16 KB, 692x72)

16 KB PNG

seeing this mathlet always makes me laugh. go back to grade school dummy

File: file.png (12 KB, 680x46)

12 KB PNG

mathlets_btfochads we are so back

>>103576597
i don't remember making this post
spooky

File: boblins.webm (1.24 MB, 396x582)

1.24 MB WEBM

did someone say goblins and elves?

>>103577511
>>103577639
>mathlets_btfo
>mathlet has a higher score
lol, lmao

>>103577971
>Day 26: Goblin Gangbang

>>103575146
I don't see the connection between these days.

>>103578011
An exponential process you can turn into a linear process via counting

>>103577973
You will never be a real programmer. You have no commits, you have no pull requests, you have no public repos. You are a low IQ brown man twisted by an unearned comp sci degree and coding bootcamps into a crude mockery of nature's autism.

All the “validation” you get is two-faced and from bots. Behind your back people say you're a useless DEI hire. Your parents are disgusted and ashamed of you, your “tech bros” laugh at you behind closed doors.

Real programmers are utterly repulsed by you. Thousands of forum posts have allowed programmers to sniff out frauds with incredible efficiency. Even devs who “pass” have no creative output and excessively verbose with AI generated tech buzzwords. Your GitHub profile is a dead giveaway, and even if you manage to score a 3 month internship at some tech company, they'll lay you off the second they find out that you haven't completed any non school related projects.

You will never be happy. You cope by saying that your bullshit software is using a "proprietary test framework" so you can't open source it, but deep down you know it's because you don't want people to see your incompetence where all you did was push README updates over and over, and that the core codebase was stolen off an open source GPL licensed project which you're shamelessly profiting off.

Eventually it will be too much to bear. You'll buy a google gift card, a spoofed business phone number, and start cold calling old people into redeeming your scam.

This is your fate. This is what you chose. There is no turning back.

Guys, I left my 12 thread bruteforce of part 2 going overnight but it isn't finished.

>>103577226
Well, you can substitute vector for a hashmap if you know the data stored is at least somewhat contiguous, like indices of a string. I think it's not that great if you need to cache values by multiple things at once or if your stuff is spaced apart.
If you mean the trie version, I think that'd be tabulation instead of memoization, since it's building upon earlier calls and not caching later ones. I just left the name there and changed the type definition.
>>103576960
A trie in cpp might be even faster if you swap that indirect node indexing for pointers. Or maybe the compiler sees that's basically a pointer and skips that entirely, I don't know.
I think it'd still be good to keep nodes in a vector, just to keep them close together in memory.

File: pretentious.png (193 KB, 832x983)

193 KB PNG

>>103574304
space-optimised bottom-up tabular push dynamic programming solution: buzzword edition

>>103578155
you need more threads

>>103578155
Quick! Lend anon your CPUs! He needs our power!

>>103578155
anon give me some of your input, i'll give you some answers

I give up, Im filtered
Especially annoying since I saw videos on YT explaining it that are around 6 mins long

>>103578286
>youtube videos
learn from the best: >>103574304, >>103578202

>>103577973
he got filtered by modulo math

Turns out cellular automata are awkward as fuck to program. What if one tile deletes adjacent tiles, but another tile moves? When the tile gets deleted depends on which cell is updated first.

rustbros do we have a #[cache] macro?

>>103578298
so use two grids and swap the references on each iteration?

I'm back home, slightly tipsy.
Time to use the ballmer curve or get filtered.

>>103578155
If every thread can find one valid combination every millisecond, you needed to start brooting when Jesus Christ was born.
Very topical.

>>103576960
>custom substring implementation
what's wrong with std::string_view?

>>103578298
store the next step in a new grid, alternate grids
if you want to be fancy, you can even use sets instead of arrays for easy infinite grids

File: day19.png (207 KB, 749x2217)

207 KB PNG

trie-solution. couldn't get it to run faster than that other rust solution from last thread.
about 0.5 ms.
tried to avoid any heap allocations, even though most of it made no difference performance wise.

>>103576915
"Natural" is not the same as "minimal."
And you curiously failed to address my objective reason to use strings as keys (easily reusing progress for multiple designs). How peculiar.

At least the idiomatic Rust guy seems to be at least competent, you are both confident and retarded.

>>103576960
std::unordered_map sucks a lot because the standard imposes requirements that force it to be implemented using a chaining hash table, try using a drop in replacement like robin_hood::unordered_flat_map

>>103578286
Say for some index X of a design, you know there are exactly 7 combos of towel pattern that reach it (i.e. there are 7 sequences of towels that cover from char 0 to char X-1 inclusive, so X is the next tile).

For every towel, check if it matches the substring starting at X. E.g. for a towel of length 3, you'd check if towel[0] == design[X] and towel[1] == design[X + 1] and towel[2] == design[X + 2] (hopefully your language has a builtin feature like comparing string slices).

If it does, you've found 7 patterns that reach index X+3 (any of the 7 patterns that reach X, plus the 3-length towel).

Of course there could be other patterns that reach X+3. But if you'd exhaustively checked every towel with every priori index, you could be sure you've found all the patterns. And there's one index you definitely know - 0 has exactly 1 path, so to speak.

So you create a vector of 0s that's the length of the design +1, and set vector[0] to 1, and then for every index from 0 onwards you see if every towel matches, and if it does you add vector[current_index] to vector[current_index + length_of_towel]. With each step you guarantee a correct total for 1 more index and when you've done the final index (i.e. design.len(), the hypothetical index 1 past the end of design that you'd reach when you place the final towel that covers the last chars of design) you have the number of patterns.

For part 2, add all these final values, and for part 1, add 1 for every value that's not zero.

>>103578293
damn, clever
I was aiming for similar stuff but I overengineered it and couldn't debug it

>>103578439
the minimal state is the canonical state; i don't care what is considered 'natural' in your twisted mind
my solution was slow because i omitted logic for early exits for the sake of terseness:

        if not dp[i]: # nothing to push
            if not any(dp): # no solution
                break
            continue

you can see my space-optimised solution here: >>103578202
though i expect it to run slightly slower due to extra modular arithmetic
now fuck off

>>103578449
now I get it
I made a big mistake of stopping taking index of a string into account, which screwed the results.
Lesson learned I guess

>>103578513
dp problems tend to be fairly simple
your state variables should be the minimal representation of the problem at each step; the value you're optimising is *not* a state variable
if you have e.g. 3 state variables, then your table should have 3 dimensions
the recurrence relations can be derived through thinking "which states can reach me" (pull dp), or "which states can i reach" (push dp)
you can often space-optimise by overwriting unreachable states - typically equivalent in effect to dropping 1 dimension - but in this case you do need a history of length `k`

>>103578523
Well, I tried both the "improved" monstrosity and even the original with a similar early exit, and guess what: it makes no fucking difference and is still slow as shit.
When will you just admit to being retarded?

>>103578580
>>103578202 is O(n**2) for any sensible defiition of n i.e. the input size. if say half of the input are designs you iterate over n/2 in the outer loop then you check against every position in every pattern in the worst case. (you do a compare of len(pattern))

>>103578601
performance: >>103574304

Days              : 0
Hours             : 0
Minutes           : 0
Seconds           : 2
Milliseconds      : 260
Ticks             : 22603467
TotalDays         : 2.61614201388889E-05
TotalHours        : 0.000627874083333333
TotalMinutes      : 0.037672445
TotalSeconds      : 2.2603467
TotalMilliseconds : 2260.3467

performance: >>103578202

> Measure-Command { cat .\input.txt | python .\aoc.py | Out-Default }
242 595975512785325


Days              : 0
Hours             : 0
Minutes           : 0
Seconds           : 1
Milliseconds      : 281
Ticks             : 12810889
TotalDays         : 1.48274178240741E-05
TotalHours        : 0.000355858027777778
TotalMinutes      : 0.0213514816666667
TotalSeconds      : 1.2810889
TotalMilliseconds : 1281.0889

stop lying on the internet

>>103578580
yea I clearly have a problem with implementing DP
Last year I was filtered by day 12 as well

>>103578628
>O(n**2)
false, which is why i ignored the number of designs and talked about the complexity of solving a single design
the length of each design isn't related at all to the number of designs, or to the length of any other designs;
you cannot call both the number of designs and the length of each design `n`

Now I see why Conway's is so simple. Oh, you want 1 cell to push another? Gotta check a 5x5 window every time to avoid double writing.

>>103578630
Oh damn you're right, it actually does make a 20% difference.

$ time python fag.py < input
315 625108891232249

real    0m2,662s
user    0m2,649s
sys    0m0,008s

vs

$ time python got.py < input
315 625108891232249

real    0m2,217s
user    0m2,184s
sys    0m0,014s

.

Now my very low-effort solution (>>103571076 for reference) runs only about 2.5x faster on my machine, how will I recover.

File: day-12.png (1.46 MB, 2952x8217)

1.46 MB PNG

>>103578655
i was also the tabular anon from day 12 last year :)

>>103578345
>what's wrong with std::string_view
Mostly that I didn't know it existed. I'm doing this year's aoc in C++ to learn the language, and STL is confusing as hell.

>>103578655
day 12 last year was my first dp problem and I hacked at it for well over a day before it clicked, this year when people mentioned the stones being similar to lantern fish I looked that up and did it first, it was easier to reason about but the same logic applied to the stones

>>103578690
good for you i guess?
i consider my approach canonical, and the principle of using minimal states simplifies harder problems: >>103578692
i think your machine has overhead somewhere, as the difference is >20% on my machine

>>103578665
no, but the sum of the lengths is what you need to consider... I am 90% sure you are larping but still, today was another boring day. Why is there no bigboy?

>>103578724
I only linked that other post to show that the bottleneck is not "early exits."
Your solutions are neither readable nor particularly efficient (even by Python's very low standards).
I wouldn't even respond if you just posted pictures without pretending to know what you are talking about.

File: 1733887359530571.png (629 KB, 822x744)

629 KB PNG

wheeeee

/,/ { split($0,q,", "); l=NF; next }

$0 {
  print NR
  s[++p] = $0
  while(p>0) {
    n = s[p--]
    if(n ~ /^0+ /) { p2++; if(!flag) p1++ }
    for(i=1;i<=l;i++) {
      if(n ~ " " i "[ $]") continue
      if(n !~ q[i]) { n = n " " i; continue }
      n2 = n
      gsub(q[i],"0",n2)
      s[++p] = n2 " " i
    }
  }
}

END { print "Part 1: " p1 "\nPart 2: " p2 }

>>103578665
>>103578764
in case you are not just run your solution on a list of a 500 times a 1000 'w' s as designs and 500 times 500 'w's as your patterns.
then scale all numbers by 2 and see what happens

>>103578801
shit, that's not my final version; that "print NR" should be resetting the flag to 0 and the flag should be set to 1 after we increment p1.

oops! still not filtered

>>103578174
>I think that'd be tabulation instead of memoization, since it's building upon earlier calls and not caching later ones.
Yes, I meant the vector named "memo" in the trie version ( >>103574847 ). Looks like a really fast solution and very predictable. For me it is much harder to see how well the caching approach is going to behave memory and speed wise.

>>103574538
looking through the replies i guess i hit a nerve by indirectly calling you a "@cachefag"
sorry for being rude on 4chan but you need to get a fucking grip

>>103574538
Yes, I am.

[Big brain idea] Pauli is operating his reindeer-powered exclusion machine; if any two cells move to the same location on the same turn, both are annihilated, leaving an empty cell. This makes the puzzle more fun and definitely isn't because I'm filtering myself coding it.

File: 1722588356504220.png (524 KB, 720x761)

524 KB PNG

>>103578864
Anime website, election toddler (and windows fag lol).

>>103578296
I was filtered by time.

>>103578883
There are anons on here who get upset if you use basic terms from Algorithms 101, like Dijkstra. It's amazing the strange hang-ups people develop. A bitter resentment towards their betters if you ask me!

>>103578964
yeah that's what happens when you broot

>>103579041
Mathematica isn't really the language for bruting.

>>103578927
technology board; at least post lain or something
>windows fag
i deny these allegations, my laptop and server run linux - this is just my vidya machine

File: quadratic.png (19 KB, 640x480)

19 KB PNG

>>103578864
so that comes out to about exactly n**2, when scaling the problem "most reasonably". i.e. by keeping the sizes of the patterns and designs limited but increasing their number. (or almost any other way as well)

    patterns = ["w" * 10 for _ in range(M)]
    designs = ["w" * 100 for _ in range(M)]

    input_size = sum(map(len, patterns)) + sum(map(len, designs))

gives almost perfect quadratic scaling of time/input_size on your solution.

File: 1602213874914.png (727 B, 32x16)

727 B PNG

/g/ saves Christmas 2025

Next year, I will create an alternative advent calendar for /g/. Please contribute your puzzles.
https://forms.gle/7f8849M1LLF8ZeAF6
01.12.2025 is almost a year away, so if you don’t have any ideas or time to implement them right now but would like to contribute later, just save this link and contribute when you're ready. (I probably won’t post it after the end of the current Advent of Code, as it would be weird to create a dedicated thread.)
Submissions are editable, so you can submit unfinished puzzles and complete them later.

Unfiltered, unbothered.
I wish we had 2023 again. As much as I hated some of the puzzles, they were absolute kino compared to most of the problems this year.

>>103579041
I sense deep and bitter mathlet resentment from you. Bad experience in grade school?

>>103579107
2023 was bullshit-hard, not fun-hard.

2022 was the last great year. I still have my cube and 3D-printed magma boulder.

Did any of you implement or use Aho-Corasick for yesterday's?

>>103579127
>Bullshit hard
That's what made it kino.
2023-21 was beautiful. Sadly Eric fucked up.

>>103579129
Does it work for 0-length patterns?

>>103579103
it's insane to refer to both the patterns and designs as `n`, and it only comes n**2 because you scaled both in tandem
O(n*m) final offer, where `n` corresponds to patterns and `m` corresponds to designs, or vice versa

>>103579106
i'm working with another anon on releasing a puzzle on the 26th and hopefully more next Christmas, if you're interested
irc.oftc.net #hoc

>>103579165
I missed a few days, are you guys still ripping on the guy from the disk fragmentation problem? If so keep it up.

File: 1723018180739193.jpg (102 KB, 640x920)

102 KB JPG

>>103579095
I recognize now that I was indeed seething at one point and I still believe you to be incorrect, but I think that we can put things behind us.
I'll even post a Nanami just for you.

>>103579183
I have "concepts of a plan" for a way to do puzzles on the board, if I have time later I'll post it.

File: lain.jpg (34 KB, 338x303)

34 KB JPG

>>103579188

>>103579198
we were planning on setting up a Tor site and generating inputs (and bigboys) for individual users

>>103579208
Why tor?

>>103578991
i can't remember a single aoc grid that required dijkstra's, i always use bfs

>>103579223
we want an edgy story about steinwitz collecting stars of david to save hanukkah
we also want to get away with it

>>103579183
I will check it out. Gonna go back to grinding the leetcode daily problem again after Christmas anyway. I just like puzzles mayne.

>>103579239
do something with more soul... anything but leetcode
https://cses.fi/problemset/

>>103579208
How are you going to handle users? it seems like the most difficult part to figure out.

File: madobe-nanami-japanese-wi(...).jpg (618 KB, 2134x1200)

618 KB JPG

>>103579188
I miss Nanami.

>>103579255
we could either use oauth or give people individual username:password combos, or both
e.g. if you ask on irc, we can pm you a login without botnet services

>>103579288
I know nothing about any of that stuff. Maybe you could just do a 4chan tripcode system where you just have a password that is hashed and logs you in. Idk.

>>103579295
could do, but you still have the problem of people spamming the leaderboard with their 50 alt trips
oauth and username:password on request makes that less of an issue

>>103579288
Why not just email+password?

>>103579321
Then your beeper goes off.

>>103579327
What

>>103579321
would need email verification and would only support botnet providers like gmail or proton
i don't think people want to give their real email to some /g/ user over Tor anyway, it's invasive

>>103579343
Ok, fuck tor, I don't want an advent of pol anyway. It will be clearnet next year with email/password. >>103579106

>>103579106
>>103579183
>/g/ aoc clone
Gets brought up basically every year, has never gone anywhere.
If you're going to go through with it, having it competing with AoC is not the way to go. At least do it in a different time span.

real "We should start a band!!!" energy

I'm here for AoC and will only be participating in that.

>>103579374
>competing with AoC
should be okay to run in parallel with the early problems
they only take a few minutes so people have time for ours too

>>103579367
okay :(

Just post a puzzle here every week over the summer or something
No need for a site

>>103579390
AoC might be over after this year

>>103579374
We'll see in a year. Maybe I will move it later, but I want christmas thematic. AoC has a lot of days when it feels not enough. Probably different time of a day and less leaderboard faggotry (e.g. filtered/not-filtered only) to avoid people worrying about their time.
>>103579381
Yeah, but you can't start a band without it.
>>103579403
It kills the idea of a second part.

bigboy/mediumboy for today. you can thank >>103578864
https://files.catbox.moe/1sgxqb.txt

day19 result:("4979", "1479223848389312371")

>>103579436
we could probably drop the edgy story if you want to just concentrate efforts
we have a few ideas for making LLM-resistant problems >>103579183

if you're willing to host on clearnet then i would contribute some puzzles

>>103579396
>should be okay to run in parallel
So are you going to piggyback off of /aocg/ (filling them with stuff unrelated to AoC), or are you going to fragment things and start up a second christmas general? This is more what I mean as a problem.
If it's more original/not overlapping, it's probably fine. If it's shoehorned to fit in, not so much. It's hard enough to convince anons to participate in AoC, trying to make a second unrelated service integral to /aocg/ is just asking for trouble.

>>103577163
it's not about the reasonable of the worry, but about the reasonableness of the degree of fear shown about it. If Eric had his entire site, backend, puzzle ideas, 30 years of elf fan fiction, and ASCII art library on a prominent git repository, then AoC would've been instantly cloned. If he hadn't had some branding sense and had hosted it on neocities, then the clone would also have a better domain name.
But inputs and story ideas are not a threat at all without him going public, making millions, and getting stockholders who can turn around and lawfare him to seize his incredibly valuable and irreplaceable assets.

File: 1701973326591240.png (57 KB, 1917x1152)

57 KB PNG

>>103579479
>edgy story
I am in this to fuck anxious elves with some Christmas chub, anon.

>>103579487
literally everything discussed so far is negotiable, but i'd imagine a separate series of threads hopefully linked in the aocg OPs

>>103579479
You'll definitely get more interest without the edgy story. As much as Shlomo's adventures in coin clipping would be amusing, not everyone has a dark sense of humor.

>>103579520
i discussed this at the start of the month and there was a lot more interest for wogbergs adventures, but i guess most of those people are filtered by now
the only thing i actually care about is that the problems are LLM-resistant

File: concepts-of-a-plan-concepts.png (59 KB, 498x281)

59 KB PNG

Here's how you can do two parts, but without any site infrastructure.

Part 1 is just what you see on the image, it'll be something reasonably simple, the way Eric does it. The harder Part 2 is another image contained within the bits of the first image (Steganography), which will be XOR'd with some number that is a function of the Part 1 solution. Part 2 can be verified as correct by hiding a success message, once again in the first at a given offset, and also XOR'd with the Part2 answer.

With this, there's no need for any infrastructure, just an anon who is willing to create a problem. The machinery involved in generating the puzzles can be automated away. These puzzles may optionally be licenced AGPL+Nigger to prevent it ending up on social media, and since tampering with the image destroys the puzzle, the notice cannot be cropped out.

>>103579605
good idea but the jannies will literally ban the poster for steganography

File: 1725974643409547.jpg (1018 B, 130x97)

1018 B JPG

fuck this shit, i got filtered on day 12, today i decided to come back to it because i just couldn't understand why the solution didn't work, and it turns out that the logic was correct but somehow rust's group by was yielding nondeterministic results, i then decided to reimplement the group by myself and it worked just fine fuckkkkkkkkkkkkkkkk

this is the solution that doesn't work

fn sides_of_dir(border: &HashSet<((isize, isize), Dir)>, dir: Dir) -> usize {
    border
        .iter()
        .filter(|(_, d)| *d == dir)
        .map(|(p, _)| p)
        .group_by(|p| match dir {
            Dir::U | Dir::D => p.1,
            Dir::L | Dir::R => p.0,
        })
        .into_iter()
        .map(|(_, sides)| {
            sides
                .into_iter()
                .map(|p| match dir {
                    Dir::U | Dir::D => p.0,
                    Dir::L | Dir::R => p.1,
                })
                .sorted()
                .collect_vec()
                .windows(2)
                .filter(|w| w[1].abs_diff(w[0]) != 1)
                .count()
                + 1
        })
        .sum()
}

this works

fn sides_of_dir(border: &HashSet<((isize, isize), Dir)>, dir: Dir) -> usize {
    let mut by_axis: HashMap<isize, Vec<isize>> = HashMap::new();
    for (x, y) in border.iter().filter(|(_, d)| *d == dir).map(|(pos, _)| pos) {
        match dir {
            Dir::U | Dir::D => by_axis.entry(*y).or_default().push(*x),
            Dir::L | Dir::R => by_axis.entry(*x).or_default().push(*y),
        }
    }
    by_axis
        .values()
        .map(|axis| {
            axis.iter()
                .sorted()
                .collect_vec()
                .windows(2)
                .filter(|w| w[1].abs_diff(*w[0]) != 1)
                .count()
                + 1
        })
        .sum()
}

rustfags owe me an explanation

>>103579466
You sure? Because I get:
Silver: 5000
Gold: 1060948046952

>>103579706
most stable rust behavior

>>103579631
I assume he means he'd host the images on a static site with clear instructions on how to extract the text and just link the site here.

>>103579466
>>103579707
No wait, your format is wrong. In the input there is a space between the ',' and the next pattern. Change that and I get the right results. 4.3 seconds.

>>103579719
if you have a static site anyway, just host normally
it would also let you take aoc market share if eric quits

>>103579706
And what have we learnt about using Rust?

>>103579706
you brought it on yourself by choosing the devil's language

>>103579605
>>103579719
>>103579732
If you are using a static site just encrypt part 2 with your answer to part 1 and write some js that tries to unencrypt it with user's input, nobody has time to xor your images or whatever.

>>103579631
>>103579719
I actually didn't know that was banned. Hosting elsewhere is another option I guess, but it would be cooler to just keep it all here. The sad irony is that hidden images could be posted without jannies being any wiser, and yet when we're honest about it, for an innocuous puzzle we run into trouble.

>>103579762
>nobody has time to xor your images or whatever.
If you cannot XOR, you don't belong in the challenge, it's not for you and you should go back to jeet code.

>>103579762
>nobody has time
Speak for yourself.

>>103579785
could maybe ask for permission but it's a humiliation ritual and 40% chance they say no and 60% chance they don't reply

Big brain solution: use the patterns to define and build a regex and then just check if each string matches the regex.

>>103579856
That might work for part 1, but there may also be some limits to the amount of backtracking your regexp has. It may even be a compile time restriction.

Was shocked how easy today's was for day 19, like if you did a recursive solution for part 2 all you had to do was make it return numbers instead of bools and add a basic memo.
I went back and did day 19 for last year (which i skipped), pretty fun problem got to use a lot of lambdas for part 1 and part 2 was a very different problem where you instead had to view the input as a tree of sorts.

>>103579963
No go back to 2015 day 19 if you want to experience real fun. If you got a lucky input, the problem is quite easy. Otherwise you’re in for a lot of pain.

>>103579762
Something like this:
https://pastebin.com/Puzkva2d

>>103580016
Is this bait? That's literally dp 101

>>103580025
Nope. As I mentioned, it’s very dependent on the input you get. I know this because at the time I happened to get the easy input, and then was confused as to why reddit found it so hard.

we still haven't had any gaussian elimination problems yet this year
we had like 5 parsing heavy versions of that in 2020

>>103580045
I solved it with top-down dp and computer the upper bound of the state count to be around 30 million.

>>103580059
Which is not a general solution.

>>103579706
>the logic was correct but somehow rust's group by was yielding nondeterministic results
Because HashMap and HashSet aren't stable, they iterate in random order

File: Advent1.png (359 KB, 1602x2154)

359 KB PNG

>>103574248
Just finished day one, how does it look?

>>103580079
Why do you think so?

>>103580092
Good morning

>>103580095
You received a lucky input, otherwise your reply wouldn’t have been “It’s just DP 101!”.

>>103580090
pytoddlers btfo

>>103580092
enable preview and it can get even smaller
https://openjdk.org/jeps/477

>>103580090
it shouldn't matter because i am only filtering and mapping and i called .sorted() before checking contiguity

>>103580114
Give me a bad input then. It's pretty easy to see that dp works in all cases.

File: evg.webm (501 KB, 884x652)

501 KB WEBM

bros I think my elves have autism, they won't stop forming lines

File: carbon(145).png (606 KB, 2936x4244)

606 KB PNG

Assembly on a $5 1GHz RISC-V cpu versus Python on a $450 5800x3D
Python is about 6x slower

>>103580166
i thought this was going to end up spelling the gamer word

>>103580184
Don't spoil part 2

>>103580166
What is this?

File: 3e9123fe1d7c70739258cacd3(...).png (106 KB, 841x975)

106 KB PNG

>>103579877
Got something close enough to it. No backtracking needed but takes about 1s to check each case.

File: aoc_2024_19.png (48 KB, 945x604)

48 KB PNG

Perl.
You can actually force the regexp to brute the solution with something like

$design =~ /^(?:$pattern)+(??{$gold++ if $& eq $design; return "(*FAIL)"})/;

, but I did not bother seeing it to the end on the actual puzzle. Curiously, using

^(?:$pattern)+(?{...})$(*FAIL)

actually triggers an optimization, regexp engine will not attempt to match left side of the string again if it has exhausted all combinations for it. So for patterns "a", "b" and "ab", trying to match "abab" it will match "a""b""ab" and "a""b""a""b" and then give up after reaching "ab".

>>103580248
:set shiftwidth=4

File: xarbon.png (15 KB, 556x660)

15 KB PNG

>>103580114
>>103580131
Here's my code, find the flaw then, oh ambassador of plebbit. Why do you even visit that shithole?

>>103580262
Shan't, helps keeping track of how nested the things are and to avoid going too much into it in smaller files.
And all the rest is already at 8 and switching would be a pain.

>>103579234
Nobody is going to shut down your website for having an edgy story. You can say nigger on twitter now.

>>103580248
>(??{ ... })
>(?{ ... })
These are so nice. Imagine if instead of the regex engine we had a general Prolog language with a good syntax for matching strings and arbitrary data structure (not the deprecated DCG shit) with the ability to put regular code inside (imperative, functional, and even control flow statements (next, last, goto) that could act on the outter quantifier as if it was a regular loop). It would be unparalleledly expressive.

I don't know for certain but I think it may have to do with the optimizations on the regexes's "main loop". The default main loop is try_match_at(0), try_match_at(1), ... try_match_at(length(string)), and the presence of anchors at the top level (not inside an alternation) is one of the situation where the main loop tries can be cut down so it does not try to match at every position.

>>103579396
For early problems I'd rather spend hours optimizing for bigboys than doing an entirely different and inferior problem
Don't compete with AoC, the best time to run it would honestly be like July when people aren't burned out

>>103580652
Hmm, I will think about it.

>>103580092
Looks good. I'd just add

throws IOException

to main instead of catching it, and

String#split(CharSequence)

takes a regex so you can also use \\s+ instead of multiple spaces.

>>103580279
m=defaultdict(lambda *x:[])
Is this AI slop? BTW how quickly does it compute an answer for you. I have an input that has been taking over a minute so far .

>>103580759
>Is this AI slop
no, why would it be? I wrote that. I like to mark arguments that I don't use as optional and you can't write lambdas with zero args.
>taking over a minute so far
It's an extremely unoptimized version, it took me 4 hours lol.
I'll write one in C++ that uses segment trees, that should run in less than a minute.

>>103579129
>Did any of you implement or use Aho-Corasick for yesterday's?
Hmm, I might do it just for fun and post a bigboy that can only be solved by it.
>>103579165
Unironically it works with zero-length strings if there are rules that prevent infinite usage.

>>103578314
>be future anon
>anyone has access to a streetcorner time machine, just a quarter per use
>working on an advent of code 2099 puzzle
>brute force will take 2000 years
>only 2000 years? that's easy
>equip laptop with an infini-battery™ and spend a quarter on time travel
>lock it in a box and bury it outside where your house will be built
>go back to present
>dig up box
>brute finished a century ago, input answer
>That's right! You've earned another gold star!
>brutebros win again

>>103580795
>you can't write lambdas with zero args.

>>> (lambda: print("hello"))()
hello

works on my machine
also you can just do defaultdict(list)

>>103580836
kek good to know, thanks
then I remembered wrong, I'll fix that

>>103580795
Well if you're the same anon, you were so damn arrogant about the problem. "Is this bait? It's just dp 101!" And yet... It doesn't seem to work.

I don't care for reddit either, but in 2015 that was the nexus around the puzzles. I went there to laugh at them feeling smug that my very naive solution "just worked" while they were pulling their hair out. Anons here were talking about that being a tough day and even as recent as a year ago there's still reddit discussions about it. Apparently the general solution is to create a parser for a context free grammar, with https://en.wikipedia.org/wiki/CYK_algorithm recommended. I want to explore this further.

>>103580870
It does work, just needs some optimizations. I promise I'll post it in the following days.
And it IS dp 101, but of course if you brute compute the states then it's slow. You have to query the minimum sum quick.

What's in store for us tonight, xisters?

>>103579706
>group_by on a hashset.
Did you even bother to read the docs?
https://doc.rust-lang.org/std/collections/struct.HashSet.html#method.iter

It literally says the order is arbitrary. You can't have a meaningful group_by using it.

>>103580977
i know that iterating over hashsets is not deterministic, but why is it a problem? an element is not put into a group based on its position

>>103580915
Djikstra (good things come in threes)

>>103581067
https://docs.rs/itertools/latest/itertools/trait.Itertools.html#method.chunk_by

Consecutive elements with the same key.
Note the example where key = bool.
Collect into a sorted vec first.

>>103581111 (checked)
Now cue the tard who will tell you that you’re trying to sound smart.

File: 1711026230973115.gif (192 KB, 1580x968)

192 KB GIF

>>103580915

>>103581140
that would be awesome actually
first fun puzzle of the year

>>103581115
Also note the name change. Probably because people got confused by the name group_by. Haskell's does the exact same shit btw. I don't know if the python itertools module partitions or not (doubt).

>>103581140
DELETE THIS

File: day19.png (70 KB, 1616x1220)

70 KB PNG

not horrendously slow for mathematica

>>103579706
Anon fell for https://github.com/rust-itertools/itertools/issues/374

>>103581140
I wonder what a sentient being in the game of life looks like.

>>103581171
Cool. Didn't think of using graphs.

File: D19_aho_corasick.png (942 KB, 2400x3910)

942 KB PNG

>>103579129
>>103580806
A little Aho-Corasick warmup before today's puzzle.

>>103581067
why do you think your second works and the first one doesn't

>>103581140
>Part 1
>The elves will move based on specific rules. How many elves after 1000 seconds?

>Part 2
>The elves are actually practicing a dance, but the ruleset was corrupted. What rules do the elves need to follow to eventually form a Christmas Tree?

File: file.png (865 B, 23x42)

865 B PNG

why is it off center?

>>103581182
Kek this basically confirms everything I've ever thought about rust.

>>103581401
It works the exact same way in python

>>103581401
>idk what an Iterator is
low IQ poster right here. The behavior of group_by makes perfect sense if you aren't retarded.

>>103581348
>>What rules do the elves need to follow to eventually form a Christmas Tree?
>The Christmas tree pattern that Eric meant this time

0000000000000
0000001000000
0000000000000
0000010100000
0000000000000
0000101010000
0000000000000
0001010101000
0000000000000
0000001000000

HALF AN HOUR

where my brootbros at

>>103581479
I'm ready to BROOOT today's puzzle

File: 1733748369024790.jpg (274 KB, 1280x1500)

274 KB JPG

>>103581479

>starting from 2015
>got to 2016 day 11
this one is hard. I keep getting 9 steps for the example, I think I misunderstood the rules in some way.

>>103581513
that's another puzzle that I take one look at and cringe because of how long it took me
I repressed the memory of how I solved it

File: yukari-2316.jpg (55 KB, 569x800)

55 KB JPG

any AOC++ leakers in chat?

>>103581462
Entropy/compression would still work on that

I think I will continue with one puzzle a night until I get all 500 stars, then again in a new language. Advent of Code is the gift that keeps on giving!

>>103581522
I hope todays is as hard. Eric is boring me this year.

>>103581309
How did it perform?

>>103581527
pathfinding on a grid 3 and it's easier than both previous problems

I told my friends I would stream my screen doing tonight’s puzzle so it is now guaranteed to be hard so I get filtered live.

>>103581527
You're given a list of equations
1 + 2 = 3
10 + 4 = 14
15 + 6 = 99
2 * 5 = 10
And you have to count the number of equations that are wrong.

>>103581527
Cellular automaton day

Poop checked.
Ready to go.

File: loli-sunglasses.png (813 KB, 899x899)

813 KB PNG

>>103581527
game simulation today

>>103581568
FUCK

>>103581566
thanks for cosmically guaranteeing us a difficult day today anon

two possibilities:
- Chinese Remainder Theorem day
- 3d Space (X, Y, Z) coordinates day

im so sleepy guys...

>>103581594
tree day
automata day

>>103581568
too hard, eric would never go that far

dijkstra today

>>103581602
We’ve already had that day.

>>103581552

DP silver: 78.168964ms
Aho-Corasick silver: 18.146983ms
DP gold: 29.145208ms
Aho-Corasick gold: 7.637811ms

The bigboy that will tell the real difference should include heavily overlapping patterns like

A, B, AB, AAB, AAAB,  AAAAB,  AAAAAB, AAAAAAB

>>103581607
but what about second dijkstra?

>eval(left)==eval(right)
Pssh, nothin personnel.

File: file.png (48 KB, 1200x1200)

48 KB PNG

Just under the wire. I thought I had a pretty good track record keeping these relatively non-lewd (except for day 5 which was low hanging fruit) but I couldn't help myself on the literal onsen problem

>>103581610
Nice. Pretty good decrease.

>>103581614

rm -rf = 12

Pssh, nothin personnel.

>>103581568
This except the input is formatted like
1 + 2 == 3
10 + 4 == 14
15 + 6 == 99
[1000 more lines]
sudo rm -rf --no-preserve-root /
[1000 more lines]
2 * 5 == 10

>>103581610
I very much doubt Aho-Corasick is worth it at pattern length 8 and I'm sure that DP could be optimized to be much faster.

>>103581623
I wish Eric had the balls to do something like this.

>>103581617
BLESS YOU ANON

>>103581597
I am too exhausted to solve anything right now. I plan to read the problem, then go to sleep and hope the solution comes to me in a dream.