# Simple complexity

Output

                  #         #  #    #  # #####  ##    #
##        ####    #######  #   #    ##
####        ##      ##    #  ### #######
#  #         #       #    ##### ### #  #
#  # ### #####   #####    #         #  #
####### ###  #  ###  #    ##        ####
##    #   #  ###  #  #######        ##
#    ##  #####   ##### #  #         #
##    #                   #         #  # #####
#    ##                  ##        #######  #
#######                ####        ##    #  ###
### #  #                #  #         #    #####
#  #                #  # ### #####    #
####                ####### ###  #    ##
##                  ##    #   #  #######
#                   #    ##  ##### #  #
#  # #####  ##    #                   #
#######  #   #    ##                  ##
##    #  ### #######                ####
#    ##### ### #  #                #  #
#####    #         #  #                #  # ###
###  #    ##        ####                #######
#  #######        ##                  ##    #
##### #  #         #                   #    ##
#         #  # #####   #####  ##    #
##        #######  #  ###  #   #    ##
####        ##    #  ###  #  ### #######
#  #         #    #####   ##### ### #  #
#  # ### #####    #       #         #  #
####### ###  #    ##      ##        ####
##    #   #  #######    ####        ##
#    ##  ##### #  #    #  #         #
#
##
####
#  #
#  # ###
#######
##    #
#    ##
#####  ##    #
###  #   #    ##
#  ### #######
##### ### #  #
#         #  #
##        ####
####        ##
#  #         #
#         #  #    #  # #####
##        ####    #######  #
####        ##      ##    #  ###
#  #         #       #    #####
#  # ### #####   #####    #
####### ###  #  ###  #    ##
##    #   #  ###  #  #######
#    ##  #####   ##### #  #
##    #                   #
#    ##                  ##
#######                ####
### #  #                #  #
#  #                #  # ###
####                #######
##                  ##    #
#                   #    ##


Trailing spaces allowed. Shortest solution wins.

hint0

hint1:

thanks @Tahg for correcting hint1

• @ngn output only challenges without any explanation to the way the data was formed are usually frowned upon (see codegolf.stackexchange.com/q/126037), because the first answer tends to blow off the "mysterious" part – Uriel Sep 19 '17 at 18:32
• Personally I don't like challenges like this where finding out the rule / recipe is part of the task. Besides, once a person finds it, all others can just follow it, so why not post in the first place? – Luis Mendo Sep 19 '17 at 18:56
• @LuisMendo It's an interesting argument. Code golf in its purest form is about "implementing a known solution succinctly." Puzzles in their purest form are about finding a solution -- the implementation is either not relevant or considered busy work. One option would be to post the "trick" in spoilers. That way the pure code golfers to attack the problem as a golfing challenge, and those who like both the golf and puzzle challenges could avoid looking and solve both. – Jonah Sep 19 '17 at 19:13
• I think the "mysterious" part is something that I could definitely see more of. Although one answer might find the trick to it, the rest of it is still a regular code-golf challenge... And it provides a good challenge for those who aim to find the pattern in the beginning anyways. – totallyhuman Sep 19 '17 at 22:22
• @H.PWiz, I'm not sure that +26/-7 is really "well received". It looks more like "poorly received but managed to hit the HNQ fast enough to get a skewed score". – Peter Taylor Sep 20 '17 at 11:58

# SOGL V0.12, 1716 14 bytes

 #6{³IIč▓┼;I+§


Try it Here!

At a later update č▓ could be removed for 12 bytes - that converts ToS from an array of arrays of characters a multiline string to an array of strings - [["#","#"],[" ","#"]] -> ["##"," #"] - , because ┼ - horizontal append - doesn't deal well with arrays of arrays of characters - which I creates, because it's used for array rotation too. In SOGL an array of arrays of characters should be = array of strings, but many things don't support that yet..

Explanation:

 #            push "#"
6{          6 times do
³           create 3 total copies of ToS
II         rotate clockwise twice
č▓       normalize as explained above
┼      append horizontally
;     get the 3rd copy ontop
I    rotate clockwise
+   append vertically
§  reverse horizontally

• wait what how does this work – Conor O'Brien Sep 22 '17 at 21:07
• @ConorO'Brien just added an explanation :p – dzaima Sep 22 '17 at 21:08
• oh nice       – Conor O'Brien Sep 22 '17 at 21:08

## JavaScript (ES6), 233217213198182170163 122 bytes

f=_=>[...Array(64)].map((_,x,a)=>a.map(g=(i=(n=64,x),j)=>(n>>=1)?i&n?j&n?g(j,i): :j&n?g(i,~j):g(~i,j):#).join).join

document.write(<pre>${f()}</pre>) Edit: Saved 14 18 bytes thanks to @Shaggy. Saved 3 bytes thanks to @ngn. Saved a further 12 bytes thanks to the two working together. Saved 41 bytes by stealing @user202729's observations that the quarters use reflections rather than rotations. Ungolfed: function f() { var s = ''; for (var i = 0; i < 64; i++) { for (var j = 0; j < 64; j++) { var x = i, y = j, c = '#'; // Default to # // Each non-blank quadrant maps to to the original // image by doubling and a reflection. Repeat // this six times unless we hit the blank quadrant. for (var n = 0; n < 6; n++) { if (x >= 32) { if (y >= 32) { // Bottom right quarter is a diagonal reflection var t = x - 32; x = y - 32; y = t; } else { // Bottom left quarter is blank c = ' '; break; } } else { if (y >= 32) { // Top right corner is a horizontal reflection y = 63 - y; } else { // Top left corner is a vertical reflection x = 31 - x; } } x *= 2; y *= 2; } s += c; } s += '\n'; } return s; }  • Do you need to count the f= here? Nicely done, by the way. – Shaggy Sep 20 '17 at 10:32 • 217 bytes – Shaggy Sep 20 '17 at 10:36 • You don't need f= in byte count, but code snippet doesn't work without it. – user72349 Sep 20 '17 at 10:49 • @Neil Any chance you could add an explanation or an ungolfed version? – Jonah Sep 22 '17 at 1:31 • @Jonah I hope you find that helpful. – Neil Sep 22 '17 at 9:06 # LOGO, 375341297295 278 + 3 bytes Add 3 bytes because of -p flag, which enable perspective mode by default, thus don't need to run perspective command, saves 9 bytes overall. Use FMSLogo on Windows with Unix (LF) newline format (FMSLogo have problem with parsing CR newline format) to R rt 90 end to g :w :l R fd 2*:l R bk :l up 180 run :w R run :w fd 2*:l R bk :l run :w fd 2*:l up 180 end to h pu ask -1[setxyz 0 0 870]g[g[g[g[g[g[rt 45 fd .7 pd fd 0 pu bk .7 lt 45]1]2]4]8]16]32 repeat 64[sety 64-# repeat 64[setx #-1 type if pixel=[0 0 0]""#["\ ]](pr)] end  Unfortunately, no "Try it online!" link because I can't find any online interpreter support perspective mode. Idea: Draw a picture of the image, then retrieve the pixels from the picture and print as output. Picture breakdown to simple repeated parts: . Make use of the hint above. However, since LOGO does not support reflection, we can only simulate that by enter 3D (perspective) mode and turn the turtle 180 degree around an axis parallel to the computer screen. This defines a helper function g, which given 2 parameters l (image side length) and w (procedure used to draw the image), draw 3 copies of its reflected image. (see hint in the question) The procedure h perform the main procedure. • You can probably omit some spaces like 90 fd -> 90fd and save a few bytes. – Jonathan Frech Sep 20 '17 at 14:14 • @JonathanFrech Unfortunately, FMSLogo doesn't support that. – user202729 Sep 20 '17 at 14:15 • Why does papert support it then? – Jonathan Frech Sep 20 '17 at 14:19 • Is this the only answer that actually exploits the pattern rather than bit packing? – Jonah Sep 20 '17 at 15:58 • @Jonah The JS solution was first to exploit the pattern, though it does it in an interesting way - with modular arithmetic and bit operations instead of concatenating matrices. This LOGO solution is also interesting and unique with its pixel-reading approach. I don't know enough LOGO to understand the code but judging by the description it repeats certain patterns without going all the way down the fractal's structure (which I already gave away in hint1). – ngn Sep 21 '17 at 13:52 # 05AB1E, 30 26 bytes '#)6F©R®í®'#ð:®ζ)2ä€ø€J˜}»  Try it online! ## CJam, 30 bytes "#"]{_W%_32ff&+\_Wf%\z+.+}6*N*  Try it online! # Python 2, 205195 ... 145144142 144 bytes g=lambda n,i,j,*_:g(n/2,*[~i,j,i,~j][2*(j&n>0)-(i&n>0):])if n*(i&n<=j&n)else'# '[i&n>0] r=range(64) for j in r:print''.join(g(32,j,i)for i in r)  Try it online! Inspired by Neil's JS answer. • n>>1 is n/=2. – Jonathan Frech Sep 20 '17 at 16:21 • a is defined and then used once. Replacing its reference with its value could save four bytes. – Jonathan Frech Sep 20 '17 at 16:24 • I overlooked a useful simplification - rather than doing everything using rotations, I should have been using reflections instead. It saved me 41 bytes on my answer! – Neil Sep 21 '17 at 16:32 • I think you can save two bytes by simplifying your [...[x],...[x]][z] selection. – Jonathan Frech Sep 22 '17 at 7:19 • @totallyhuman Fair enough – TFeld Nov 10 '17 at 8:09 ## Haskell, 126125113106 103 bytes e=[]:e z=zipWith r=reverse m s=z(++)(r s++map(' '<$)s)$map r s++foldr(z(:))e s unlines$m$m$m$m$m$m["#"]  A direct implementation of hint1 in the spoiler. Function m generates the next iteration. The main functions applies m 6 times. Details to m: m s= -- s is the current pattern zipWith(++) -- concatenate pairwise the lines of (r s)++ -- the lines of s in reverse order, followed by map(' '<$)s   --    each char in s replaced by a space
-- and
map r s++     --    each line of s reversed, followed by
foldr ... s   --    the transposition of s

e=[]:e;foldr(zipWith(:))e     -- this right fold transposes a matrix


Edit: @ngn saved a byte and @Laikoni another 3. Thanks!

• n$n$n$n$n$n["#"] is shorter than iterate n["#"]!!6 :) – ngn Sep 21 '17 at 16:22 • @ngn: well spotted. Thanks a lot! – nimi Sep 21 '17 at 16:32 • You can use e=[]:e;foldr(zipWith(:))e instead of import Data.List;transpose and then shorten zipWith. – Laikoni Sep 22 '17 at 16:25 # Java 7, 259 238 237 200 bytes void f(){char[][]o={{35}},n;for(int s=1,x,y;s<64;s*=2,o=n)for(n=new char[64][64],x=s;x-->0;)for(y=0;y<s;n[s+y++][x]=32)n[s+~y][x]=n[y][2*s+~x]=n[s+x][s+y]=o[y][x];for(char[]b:o)System.out.println(b);}  Saved 2 bytes by removing the {} on the x loop, thanks ngn Saved 19 bytes from various assignment changes, thanks Jonathan Saved 24 bytes for print vs return (didn't know this was allowed), and Saved 13 bytes for loop changes, thanks Nevay Try it online! My first challenge, and I think respectable for Java. Uses hint 1, (which is wrong btw, but I can't write comments). It can probably be golfed further, this pass was basically as is, without pretty printing. • I am not sure if it might be because of a version difference, but I tested your code on TIO using Java 8, and it prints null characters instead of spaces. – Jonathan Frech Sep 21 '17 at 6:17 • It sure did, thanks for noticing. Apparently, Java is quite happy with displaying nulls as spaces in its Strings. – Tahg Sep 21 '17 at 6:36 • You can save a byte by removing s=1 from your for loop and replacing int s,t,x,y; with int s=1,t,x,y;. – Jonathan Frech Sep 21 '17 at 6:40 • n[s-1-y][x]=o[y][x];n[y][t-1-x]=o[y][x]; -> n[s-1-y][x]=n[y][t-1-x]=o[y][x];. – Jonathan Frech Sep 21 '17 at 6:45 • the proper transformation matrix is [[V,H],[_,T]] (Either that or there's some subtle difference in your algorithm, but that's what I needed) – Tahg Sep 21 '17 at 6:47 # J, 30 27 bytes |:@(|:@|.,|.@|:,.])^:6,.'#'  Try it online! ### Original version. ((|.,.|."1),' '"0,.|:)^:6,.'#'  Try it online! ## Python 2, 586 bytes import zlib,base64 as b print zlib.decompress(b.b64decode('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'))  Try it online. ## Python 2, 1032 1025 bytes I like this one more. But it is much longer. It could be golfed down but there is no need for that. officialaimm's approach is much shorter. -7 thanks to Jonathan Frech print''.join(' '*int(x)if'/'<x<'@'else("#"*(ord(x)-64),"\n")[x<"#"]for x in'99A9A2A4A2A1E2B4A9!99B8D4G2A3A4B8!88D8B6B4A2C1G8!88A2A9A7A4E1C1A2A8!88A2A1C1E3E4A9A2A8!88G1C2A2C2A4B8D8!88B4A3A2C2A2G8B55!98A4B2E3E1A2A9A55!88B4A559A9A2A1E2!98A4B99B8G2A2!98G88D8B4A2C!88C1A2A88A2A9A4E1!992A2A88A2A1C1E4A5!992D88G1C2A4B4!992B99B4A3A2G4!777A559A4B2E1A2A4!4A2A1E2B4A559A777!4G2A3A4B99B992!4B4A2C1G88D992!5A4E1C1A2A88A2A569!1E4A9A2A88A2A1C88!C2A4B8D88G98!2A2G8B99B4A89!2E1A2A9A559A4B88!55A9A2A1E3E2B4A98!55B8G2A2C2A3A4B88!8D8B4A2C2A2C1G88!8A2A9A4E3E1C1A2A88!8A2A1C1E4A7A9A2A88!8G1C2A4B6B8D88!8B4A3A2G4D8B99!9A4B2E1A2A4A2A9A99!99996A777!99996B992!99994D299!99994A2A992!99994A2A1C88!99994G98!99994B4A98!99995A4B88!9996E2B4A89!9995C2A3A4B88!9997A2C1G88!9997E1C1A2A88!9997A9A2A88!9997B8D88!9995D8B99!9995A2A9A99!9997A9A2A4A2A1E2!9997B8D4G2A2!9995D8B6B4A2C!9995A2A9A7A4E1!9995A2A1C1E3E4A5!9995G1C2A2C2A4B4!9995B4A3A2C2A2G4!9996A4B2E3E1A2A4!9995B4A559A5!9996A4B99B4!9996G88D4!9995C1A2A88A2A4!9999A2A88A2A1C!9999D88G1!9999B99B4A1!99991A559A4B')  Try it online. • The question allows [t]railing spaces, though your solution omits some spaces. – Jonathan Frech Sep 20 '17 at 7:19 • allowing trailing spaces means, that I dont have to output them. Is that correct? It would save bytes on my second approach. On the first one it is already done because I deleted the trailing spaces while saving my file... – Simon Sep 20 '17 at 7:24 • I do not know for sure, but I interpreted it as allowing additional spaces, not allowing missing spaces. – Jonathan Frech Sep 20 '17 at 7:26 • x=='!' -> x<'#' – Jonathan Frech Sep 20 '17 at 7:27 • if x.isdigit() -> if"/"<x<"@" – Jonathan Frech Sep 20 '17 at 7:28 # Mathematica, 112 90 bytes Thanks to Jonathan Frech for help saving 2 bytes! ""<>Nest[{{(r=Reverse)@#,r/@#},{" "+0#,#}}~Flatten~{{1,3},{2,4}}&,{{"#"}},6]~Riffle~" "  Try it online! (Mathics) For some reasons, Mathics prints leading spaces in all lines except the first one when print multiline string. Also Mathics doesn't support operator. Explanation: • {{Reverse@#,Reverse/@#},{" "+0#,#}} : Represent the reverse-horizontally, reverse-vertically, replace-all-by-" ", and transpose ( is transpose operator in Mathematica), corresponding to different ways to reflect or rotate the image. • ~Flatten~{{1,3},{2,4}} : Flatten in particular dimensions. • Nest[ ... ,{{"#"}},6] : Apply the function inside to {{"#"}} 6 times. • ~Riffle~"<literal newline character>" : Riffle a newline character between each line. • ""<> : Join all strings together. • You can define replace Reverse with R and define R=Reverse; to save two bytes. – Jonathan Frech Sep 21 '17 at 13:33 # C# (.NET Core), 10161002980 955 bytes Saved 14 bytes thanks to Kevin Cruijssen! Saved 47 bytes thanks to Jonathan Frech! o=>{foreach(var l in new[]{35223185965568,52841249260288,0xf00c0c277f00,0x9004043ee900,0x977c7c200900,0xfee4e4300f00,0xc22727f00c00,73934616658944,0xc2000020097c,73667282210788,0x7f0000f00c27,0xe9000090043e,9895614577696,0xf0000fee430,0xc0000c227f0,4398050918032,325982355457<<21,1092065689603<<20,835235872783<<20,291765223433<<20,0x7c20090000970000,-15289957744513<<17,21955973480545<<17,68788263321667<<16,68799053409<<17,206425089091<<16,0xf00c27277f0000,618546478825<<16,650622541833<<16,0xfee430300f0000,208473439235<<18,72203117569<<18,1<<21,3<<20,15<<20,9<<20,9895936,127<<17,97<<17,67<<16,15969<<17,3829596160,662634496,16105<<16,8201<<16,806289408,4027318272,9217L<<18,537463164,806293476,4027321383,2416182334,2541517856,4276413488,3257346032,1128152720,3254779936,1124073520,2130706672,3909091472,150995095,251658494,201326786,67108931})System.Console.WriteLine(System.Convert.ToString(l,2).PadLeft(64,'0').Replace('0',' ').Replace('1','#'));}  Try it online! # Explanation The format of the output is stored inside an array of signed 64-bit numbers, which fits perfectly since each line is 64 characters long. Empty spaces are represented by 0 and # is represented by 1. The numbers are then converted to their binary string, zeros are padded left until the string is 64 characters wide and then the 0 and 1 characters are replaced with   and #. The code is stored inside a lamba function, more specifically a System.Func<string>. Some constants in the long[] are shortened by performing some bit-shifting. • Is there not an unnecessary space in long[] n=new[]? – Jonathan Frech Sep 20 '17 at 9:17 • @JonathanFrech Ah you are right, apparently it snuck in there somehow. – Ian H. Sep 20 '17 at 9:18 • You would save 4 bytes by not having to specify the padding character. – Neil Sep 20 '17 at 11:07 • Well, if the trailing l is not needed, you can probably save even more bytes. – Jonathan Frech Sep 20 '17 at 11:11 • Looks like the bottom of this is bugged. – Shaggy Sep 20 '17 at 11:35 # Charcoal, 42 bytes #↓##ＦＥ⁵Ｘ²ι«‖↑Ｊ⁻×³ι¹±⁰ψ⟲ＯＯ⁴×²ι⟲ＣＪ⁰±φＴ×⁴ι×⁴ι  Try it online! Link is to verbose version of code. Would be 25 bytes if this worked: #ＦＥ⁷Ｘ²ι«‖↑⟲Ｃ→⁴⟲ＣＪ⁰±φＴιι  Explanation: #↓##  Manually generate the first recursion, since it's not possible to rotate around (1, 0.5). (Such a rotation would only make sense if it was 180°.) ＦＥ⁵Ｘ²ι«  Loop over the first five powers of 2 (1, 2, 4, 8, 16). ‖↑  Reflect the canvas vertically. This completes the top left quarter of the result. Ｊ⁻×³ι¹±⁰ψ⟲ＯＯ⁴×²ι  Rotate the canvas 180° around a point half-way up the right-hand side. The copy ends up in the correct position for the top right quarter of the result. ⟲Ｃ  Rotate the canvas 90° around the bottom right corner. The copy of the top right corner ends up in the correct position for the bottom right corner of the result. The copy of the top left corner is extraneous. Ｊ⁰±φＴ×⁴ι×⁴ι  Trim the canvas to the size we want. The trim starts at the cursor or the top left of the canvas, whichever is bottom rightmost. The cursor is therefore sent to (0, -1000) to ensure that it doesn't interfere with the trim. # Dyalog APL, 29 bytes ⎕←(((⊖,⌽)⍪⊢∘' '¨,⍉)⍣6)1 1⍴'#'  Try it online! # Python 2, 715 711 bytes • -4 bytes thanks to @Jonathan Frech for k in range(64):print"".join(ord(i)*" #"[j%2]for j,i in enumerate('            <?<<<89:11222 202 4 $" 	!%$%$&%$($%&'))[k<<6:][:64]


Try it online!

• 711 bytes. – Jonathan Frech Sep 20 '17 at 5:46
• @JonathanFrech Thank you, I didn't think there were no new-lines. :D – officialaimm Sep 20 '17 at 5:52

## Perl 5, 452 bytes

451 bytes code + 1 for -p.

Packs all the data into the string, rather than stealing @Neil's correct answer.

$_=unpack"B*","\x00\x00\x20\x09\x09\x7c\xc2\x33\x00\x30\x0f\x0f\xe4\x43\x33\x00\xf0\x0c\x0c\x27\x7f\x33\x00\x90\x04\x04\x3e\xe9\x33\x00\x97\x7c\x7c\x20\x09\x33\x00\xfe\xe4\xe4\x30\x0f\x33\x00\xc2\x27\x27\xf0\x0c\x33\x00\x43\x3e\x3e\x90\x04\x33\x00\xc2\x00\x00\x20\x09\x7c\x00\x00\x43\x00\x00\x30\x0f\xe4\x00\x00\x7f\x00\x00\xf0\x0c\x27\x00\x00\xe9\x00\x00\x90\x04\x3e\x00\x00\x09\x00\x00\x97\x7c\x20\x00\x00\x0f\x00\x00\xfe\xe4\x30\x00\x00\x0c\x00\x00\xc2\x27\xf0\x00\x00\x04\x00\x00\x43\x3e\x90\x09\x7c\xc2\x00\x00\x20\x00\x00\x0f\xe4\x43\x00\x00\x30\x00\x00\x0c\x27\x7f\x00\x00\xf0\x00\x00\x04\x3e\xe9\x00\x00\x90\x00\x00\x7c\x20\x09\x00\x00\x97\x00\x00\xe4\x30\x0f\x00\x00\xfe\x00\x00\x27\xf0\x0c\x00\x00\xc2\x00\x00\x3e\x90\x04\x00\x00\x43\x33\x00\x20\x09\x7c\x7c\xc2\x33\x00\x30\x0f\xe4\xe4\x43\x33\x00\xf0\x0c\x27\x27\x7f\x33\x00\x90\x04\x3e\x3e\xe9\x33\x00\x97\x7c\x20\x20\x09\x33\x00\xfe\xe4\x30\x30\x0f\x33\x00\xc2\x27\xf0\xf0\x0c\x33\x00\x43\x3e\x90\x90\x04\x37\x00\x20\x37\x00\x30\x37\x00\xf0\x37\x00\x90\x37\x00\x97\x37\x00\xfe\x37\x00\xc2\x37\x00\x43\x36\x00\x7c\xc2\x36\x00\xe4\x43\x36\x00\x27\x7f\x36\x00\x3e\xe9\x36\x00\x20\x09\x36\x00\x30\x0f\x36\x00\xf0\x0c\x36\x00\x90\x04\x36\x00\x20\x09\x09\x7c\x34\x00\x30\x0f\x0f\xe4\x34\x00\xf0\x0c\x0c\x27\x34\x00\x90\x04\x04\x3e\x34\x00\x97\x7c\x7c\x20\x34\x00\xfe\xe4\xe4\x30\x34\x00\xc2\x27\x27\xf0\x34\x00\x43\x3e\x3e\x90\x34\x00\xc2\x00\x00\x20\x34\x00\x43\x00\x00\x30\x34\x00\x7f\x00\x00\xf0\x34\x00\xe9\x00\x00\x90\x34\x00\x09\x00\x00\x97\x34\x00\x0f\x00\x00\xfe\x34\x00\x0c\x00\x00\xc2\x34\x00\x04\x00\x00\x43"=~s/[3-9](.)/$1x$&/ger;y/01/ #/;s/.{64}/$&
/g


Reversible output from xxd for 451 byte file:

00000000: 245f 3d75 6e70 6163 6b22 422a 222c 2200  $_=unpack"B*",". 00000010: 0020 0909 7cc2 3300 300f 0fe4 4333 00f0 . ..|.3.0...C3.. 00000020: 0c0c 277f 3300 9004 043e e933 0097 7c7c ..'.3....>.3..|| 00000030: 2009 3300 fee4 e430 0f33 00c2 2727 f00c .3....0.3..''.. 00000040: 3300 433e 3e90 0433 00c2 0000 2009 7c00 3.C>>..3.... .|. 00000050: 0043 0000 300f e400 007f 0000 f00c 2700 .C..0.........'. 00000060: 00e9 0000 9004 3e00 0009 0000 977c 2000 ......>......| . 00000070: 000f 0000 fee4 3000 000c 0000 c227 f000 ......0......'.. 00000080: 0004 0000 433e 9009 7cc2 0000 2000 000f ....C>..|... ... 00000090: e443 0000 3000 000c 277f 0000 f000 0004 .C..0...'....... 000000a0: 3ee9 0000 9000 007c 2009 0000 9700 00e4 >......| ....... 000000b0: 300f 0000 fe00 0027 f00c 0000 c200 003e 0......'.......> 000000c0: 9004 0000 4333 0020 097c 7cc2 3300 300f ....C3. .||.3.0. 000000d0: e4e4 4333 00f0 0c27 277f 3300 9004 3e3e ..C3...''.3...>> 000000e0: e933 0097 7c20 2009 3300 fee4 3030 0f33 .3..| .3...00.3 000000f0: 00c2 27f0 f00c 3300 433e 9090 0437 0020 ..'...3.C>...7. 00000100: 3700 3037 00f0 3700 9037 0097 3700 fe37 7.07..7..7..7..7 00000110: 00c2 3700 4336 007c c236 00e4 4336 0027 ..7.C6.|.6..C6.' 00000120: 7f36 003e e936 0020 0936 0030 0f36 00f0 .6.>.6. .6.0.6.. 00000130: 0c36 0090 0436 0020 0909 7c34 0030 0f0f .6...6. ..|4.0.. 00000140: e434 00f0 0c0c 2734 0090 0404 3e34 0097 .4....'4....>4.. 00000150: 7c7c 2034 00fe e4e4 3034 00c2 2727 f034 || 4....04..''.4 00000160: 0043 3e3e 9034 00c2 0000 2034 0043 0000 .C>>.4.... 4.C.. 00000170: 3034 007f 0000 f034 00e9 0000 9034 0009 04.....4.....4.. 00000180: 0000 9734 000f 0000 fe34 000c 0000 c234 ...4.....4.....4 00000190: 0004 0000 4322 3d7e 732f 5b33 2d39 5d28 ....C"=~s/[3-9]( 000001a0: 2e29 2f24 3178 2426 2f67 6572 3b79 2f30 .)/$1x$&/ger;y/0 000001b0: 312f 2023 2f3b 732f 2e7b 3634 7d2f 2426 1/ #/;s/.{64}/$&
000001c0: 0a2f 67                                  ./g


Try it online!

# Jq 1.5, 538535488 476 bytes

This is a straightforward representation. I haven't attempted any fancy encodings yet. Replaced 0,0 pairs with Z function.

Thanks again to Jonathan Frech for helping eliminate 3 bytes!

def B:recurse(if.>0then./2|floor else empty end)|.%2;def S:256+.|[B]|reverse[2:]|map(if.>0then"#"else" "end)|join("");def Z:0,0;[[Z,3,1,1,4,5,2],[1,4,5,2,2,3,Z],[Z,Z,2,3,Z],[Z,Z,3,1,1,4]]as$m|[[],[9,15,12,4,124,228,39,62],[Z,Z,Z,Z,124,228,39,62,32,48,240,144],[32,48,240,144,151,254,194,67,194,67,127,233,9,15,12,4],[124,228,39,62,32,48,240,144,32,48,240,144,151,254,194,67],[194,67,127,233,9,15,12,4,9,15,12,4,124,228,39,62]]as$c|$m[]|range(16)as$l|map($c[.][$l]|S)|join("")


Character count

$wc -c picture.jq 476 picture.jq  Sample run $ jq -Mnr -f picture.jq
#         #  #    #  # #####  ##    #
##        ####    #######  #   #    ##
####        ##      ##    #  ### #######
#  #         #       #    ##### ### #  #
#  # ### #####   #####    #         #  #
####### ###  #  ###  #    ##        ####
##    #   #  ###  #  #######        ##
#    ##  #####   ##### #  #         #
##    #                   #         #  # #####
#    ##                  ##        #######  #
#######                ####        ##    #  ###
### #  #                #  #         #    #####
#  #                #  # ### #####    #
####                ####### ###  #    ##
##                  ##    #   #  #######
#                   #    ##  ##### #  #
#  # #####  ##    #                   #
#######  #   #    ##                  ##
##    #  ### #######                ####
#    ##### ### #  #                #  #
#####    #         #  #                #  # ###
###  #    ##        ####                #######
#  #######        ##                  ##    #
##### #  #         #                   #    ##
#         #  # #####   #####  ##    #
##        #######  #  ###  #   #    ##
####        ##    #  ###  #  ### #######
#  #         #    #####   ##### ### #  #
#  # ### #####    #       #         #  #
####### ###  #    ##      ##        ####
##    #   #  #######    ####        ##
#    ##  ##### #  #    #  #         #
#
##
####
#  #
#  # ###
#######
##    #
#    ##
#####  ##    #
###  #   #    ##
#  ### #######
##### ### #  #
#         #  #
##        ####
####        ##
#  #         #
#         #  #    #  # #####
##        ####    #######  #
####        ##      ##    #  ###
#  #         #       #    #####
#  # ### #####   #####    #
####### ###  #  ###  #    ##
##    #   #  ###  #  #######
#    ##  #####   ##### #  #
##    #                   #
#    ##                  ##
#######                ####
### #  #                #  #
#  #                #  # ###
####                #######
##                  ##    #
#                   #    ##

• Three omittable spaces in B: rec and 16) as $l|. – Jonathan Frech Sep 21 '17 at 8:02 • if.==0then" "else"#"end -> if.>0then"#"else" "end. – Jonathan Frech Sep 21 '17 at 8:03 # Python 2, 174164 154 bytes (based on matrix operations) • -10 bytes thanks to ngn. • -10 bytes thanks to Jonathan Frech. from numpy import* s=[['#']] exec'b=fliplr(s);s=vstack((hstack((flipud(s),b)),hstack((eye(len(s)),rot90(b)))));'*6 s[s>'#']=' ' for l in s:print''.join(l)  Try it online! • import * -> import*; in s.tolist() -> in s – ngn Sep 28 '17 at 18:53 • – Jonathan Frech Nov 10 '17 at 8:25 # Canvas, 10 bytes #６［↔ω⤢∔ω↶×  Try it here! # JavaScript (Node.js), 1233 bytes _=> 18# 9# 2# 4# 2# #5 2#2 4# 18#2 8#4 4#7 2# 3# 4#2 16#4 8#2 6#2 4# 2#3 #7 16# 2# 9# 7# 4#5 #3 # 2# 16# 2# #3 #5 3#5 4# 9# 2# 16#7 #3 2# 2#3 2# 4#2 8#4 16#2 4# 3# 2#3 2# 2#7 8#2 17# 4#2 2#5 3#5 # 2# 9# 16#2 4# 19# 9# 2# #5 17# 4#2 18#2 8#7 2# 17#7 16#4 8#2 4# 2#3 16#3 # 2# 16# 2# 9# 4#5 20# 2# 16# 2# #3 #5 4# 20#4 16#7 #3 2# 4#2 20#2 18#2 4# 3# 2#7 21# 19# 4#2 2#5 # 2# 4# 2# #5 2#2 4# 19# 4#7 2# 3# 4#2 18#2 4#2 4# 2#3 #7 16#4 5# 4#5 #3 # 2# 16# 2# #5 4# 9# 2# 16# 2# #3 #3 2# 4#2 8#4 16#7 2# 2#7 8#2 18#2 4# 2#5 # 2# 9# 19# 4#2 10# 9# 2# #5 3#5 2#2 4# 10#2 8#7 2# 2#3 2# 3# 4#2 8#4 8#2 4# 2#3 2# 2#3 #7 8# 2# 9# 4#5 3#5 #3 # 2# 8# 2# #3 #5 4# 7# 9# 2# 8#7 #3 2# 4#2 6#2 8#4 8#2 4# 3# 2#7 4#4 8#2 9# 4#2 2#5 # 2# 4# 2# 9# 42# 42#2 40#4 40# 2# 40# 2# #3 40#7 40#2 4# 41# 4#2 33#5 2#2 4# 32#3 2# 3# 4#2 34# 2#3 #7 34#5 #3 # 2# 34# 9# 2# 34#2 8#4 32#4 8#2 32# 2# 9# 34# 9# 2# 4# 2# #5 34#2 8#4 4#7 2# 32#4 8#2 6#2 4# 2#3 32# 2# 9# 7# 4#5 32# 2# #3 #5 3#5 4# 32#7 #3 2# 2#3 2# 4#2 32#2 4# 3# 2#3 2# 2#7 33# 4#2 2#5 3#5 # 2# 32#2 4# 19# 33# 4#2 18#2 33#7 16#4 32#3 # 2# 16# 2# 36# 2# 16# 2# #3 36#4 16#7 36#2 18#2 4# 37# 19# 4#2.replace(/(.)(\d+)/g,(_,c,n)=>c.repeat(n))  Try it online! • Ooh, what's the problem now? Erm... Am I missing something? – totallyhuman Sep 19 '17 at 22:49 • I didn't downvote, but, fyi, it can be compressed much more using this. – user72349 Sep 19 '17 at 23:54 • I didn't either, but this isn't in the spirit of the question. The pattern was generated, presumably, by some simple recursive procedure. The challenge is determine the rule, at which point your solution will be extremely short. – Jonah Sep 20 '17 at 1:35 • Convert to base 36 to save ~50 bytes. – Shaggy Sep 20 '17 at 7:30 • I am completely aware that this is not the optimal solution or the shortest solution or a clever one. I simply tried using an algorithm and it ended up being this. This is, however, a perfectly valid solution and hence doesn't deserve downvotes. Boring solutions should not be upvoted but they shouldn't be downvoted either. – totallyhuman Sep 20 '17 at 22:01 # C# (.NET Core), 976 969 bytes o=>{foreach(var l in new[]{35223185965568,52841249260288,0xf00c0c277f00,0x9004043ee900,0x977c7c200900,0xfee4e4300f00,0xc22727f00c00,73934616658944,0xc2000020097c,73667282210788,0x7f0000f00c27,0xe9000090043e,9895614577696,0xf0000fee430,0xc0000c227f0,4398050918032,0x97cc20000200000,0xfe44300003<<20,0xc277f0000f<<20,0x43ee900009<<20,0x7c20090000970000,-0x1bcff0ffff020000,0x27f00c0000c20000,0x3e90040000430000,9017629528424448,0x300fe4e4430000,0xf00c27277f0000,0x90043e3ee90000,0x977c2020090000,0xfee430300f0000,0xc227f0f00c0000,0x433e9090040000,2097152,3145728,15728640,9437184,9895936,16646144,12713984,4390912,2093088768,3829596160,662634496,1055457280,537460736,806289408,4027318272,2416181248,537463164,806293476,4027321383,2416182334,2541517856,4276413488,3257346032,1128152720,3254779936,1124073520,2130706672,3909091472,150995095,251658494,201326786,67108931})System.Console.WriteLine(System.Convert.ToString(l,2).PadLeft(64).Replace('0',' ').Replace('1','#'));}  Try it online! • Hi, welcome to PPCG! This looks almost exactly the same as @IanH's C# .NET answer.. So if you have any improvements for his, make a comment instead of creating a new answer that is almost exactly the same. As for your question, you can use <s>969</s> to cross out the previous bytes. – Kevin Cruijssen Sep 20 '17 at 11:26 • I do not have the reputation needed to comment. – my pronoun is monicareinstate Sep 20 '17 at 11:26 • Please don't try to circumvent the rep requirements. – Shaggy Sep 20 '17 at 11:33 • Does this mean that in PPCG I cannot try to help others if my reputation is too low? – my pronoun is monicareinstate Sep 20 '17 at 11:35 • @Shaggy to be fair, stackexchange is a bit harsh on newcomers, let's not make it harder for them by assuming malice too quickly – ngn Sep 21 '17 at 8:51 # C# (.NET Core), 739 bytes _=>{var r="";for(int i=0,j,k=0;i<626;i++)for(j=0;j++<@"4#+#$#&#$##'$$&#=*&&)#%#&:&*(&#%#):##+#)#&'#%###:###%#'%'&#+##:)#%#%#&*&:&#%#%#)*=#&$$'%'##$#+#<$&#5#+#$##'5#&$4$*)$#5)2&*$&#$%2%##$#2#$#+#&'7#$#2#$##%#'&#;&2)#%$#&$:$4$&#%#$);#5#&$$'###*###'$$&#5#;)$#%#&$4$:$&#$%#)2&;#&'#%##$#2#$#7'&#+#$#2#$##%2%$#&$*&2)5#$)*$4$&#5'##$#+#5#&$<#+#$##'%'$$&#=*)#%#%#&:&*&#%#%#):##+#&'%'#%###:###%#'&#)#+##:)#%#&(*&:&#%#)&&*=#&$$'##$#&#$#+#^#a$^&^#$#^#$##%Z)[$&#\#&$S'$$&#S%#%#&T#%#)T'#%###T#+##T*&R&*T##+#V#+##&###'F*&&)#D&*(&#%B##+#)#&'C###%#'%'&#G)#%#%#&F&#%#%#)G#&$$'%'##$#F$&#5#H#&$4$G)2&F%##$#2#$#J#$#2#$##%F&2)G$4$&#H#5#&\$"[i]-34;){r+=i%2<1?' ':'#';if(++k%64<1)r+='\n';}return r;}


Try it online!

Uses the same approach is this other answer.

# K (ngn/k), 3231 27 bytes

6{,[|x;""x],'(+|+x),+x}/"#"


Try it online!

6{ }/ 6 times do

+x transpose

|x reverse vertically

+|+x reverse horizontally

, concatenate vertically

,' concatenate horizontally

,[A;B] is the same as A,B. it helps avoid parentheses around A and around the whole expression

""x use the elements of x as indices in the empty string. out-of-bounds indexing produces spaces, so this expression will return an all-spaces matrix, same size as x