diff --git a/posts/adventofcode/2017/02-corruption-checksum.md b/posts/adventofcode/2017/02-corruption-checksum.md
index 2116dc3..476496f 100644
--- a/posts/adventofcode/2017/02-corruption-checksum.md
+++ b/posts/adventofcode/2017/02-corruption-checksum.md
@@ -13,7 +13,7 @@ series: aoc2017
 
 [Today's challenge](http://adventofcode.com/2017/day/2) is to calculate a rather contrived "checksum" over a grid of numbers.
 
-[Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/02-corruption-checksum.py)
+[→ Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/02-corruption-checksum.py)
 
 !!! commentary
     Today I went back to plain Python, and I didn't do formal tests because only one test case was given for each part of the problem. I just got stuck in.
diff --git a/posts/adventofcode/2017/03-spiral-memory.md b/posts/adventofcode/2017/03-spiral-memory.md
index 970bdb2..c4ccafc 100644
--- a/posts/adventofcode/2017/03-spiral-memory.md
+++ b/posts/adventofcode/2017/03-spiral-memory.md
@@ -13,7 +13,7 @@ series: aoc2017
 
 [Today's challenge](http://adventofcode.com/2017/day/3) requires us to perform some calculations on an "experimental memory layout", with cells moving outwards from the centre of a square spiral (squiral?).
 
-[Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/03-spiral-memory.go)
+[→ Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/03-spiral-memory.go)
 
 !!! commentary
     I've been wanting to try my hand at [Go](https://golang.com), the memory-safe, statically typed compiled language from Google for a while. Today's challenge seemed a bit more mathematical in nature, meaning that I wouldn't need too many advanced language features or knowledge of a standard library, so I thought I'd give it a "go". It might have been my imagination, but it was impressive how quickly the compiled program chomped through 60 different input values while I was debugging.
diff --git a/posts/adventofcode/2017/04-high-entropy-passphrases.md b/posts/adventofcode/2017/04-high-entropy-passphrases.md
index 29ad8dc..85464c3 100644
--- a/posts/adventofcode/2017/04-high-entropy-passphrases.md
+++ b/posts/adventofcode/2017/04-high-entropy-passphrases.md
@@ -13,7 +13,7 @@ series: aoc2017
 
 [Today's challenge](http://adventofcode.com/2017/day/4) describes some simple rules supposedly intended to enforce the use of secure passwords. All we have to do is test a list of passphrase and identify which ones meet the rules.
 
-[Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/04-high-entropy-passphrases.py)
+[→ Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/04-high-entropy-passphrases.py)
 
 !!! commentary
     Fearing that today might be as time-consuming as yesterday, I returned to Python and it's hugely powerful "batteries-included" standard library. Thankfully this challenge was more straightforward, and I actually finished this before finishing [day 3](../day-03/).
diff --git a/posts/adventofcode/2017/05-a-maze-of-twisty-trampolines.md b/posts/adventofcode/2017/05-a-maze-of-twisty-trampolines.md
index 369667d..622a632 100644
--- a/posts/adventofcode/2017/05-a-maze-of-twisty-trampolines.md
+++ b/posts/adventofcode/2017/05-a-maze-of-twisty-trampolines.md
@@ -13,7 +13,7 @@ series: aoc2017
 
 [Today's challenge](http://adventofcode.com/2017/day/5) has us attempting to help the CPU escape from a maze of instructions. It's not quite a [Turing Machine](https://en.wikipedia.org/wiki/Turing%20Machine), but it has that feeling of moving a read/write head up and down a tape acting on and changing the data found there.
 
-[Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/05-a-maze-of-twisty-trampolines.cc)
+[→ Full code on GitHub](https://github.com/jezcope/aoc2017/blob/master/05-a-maze-of-twisty-trampolines.cc)
 
 !!! commentary
     I haven't written anything in C++ for over a decade. It sounds like there have been lots of interesting developments in the language since then, with C++11, C++14 and the freshly finalised C++17 standards (built-in parallelism in the STL!). I won't use any of those, but I thought I'd dust off my C++ and see what happened. Thankfully the Standard Template Library classes still did what I expected!