diff --git a/17. Hard problems/17.13. re_space.md b/17. Hard problems/17.13. re_space.md
new file mode 100644
index 0000000..abfa28f
--- /dev/null
+++ b/17. Hard problems/17.13. re_space.md	
@@ -0,0 +1,40 @@
+# 17.13. Re-space
+
+## The spaces from a sentence have been removed. Most words are present in the dictionary but some aren't. Given a dictionary (a list of strings) and the document (a string), unconcatenate the document in a way that minimizes the number of unrecognized characters
+
+> example: 'jesslookedjustliketimherbrother' -> 'jess looked just like tim her brother'
+
+## First idea
+
+This is complicated... I could start first by unconcatenating the sentences considering they're all present in the dictionary, and then optimize. How do I unconcatenate in the first place? .tolower() first
+
+* Easiest case: 'iamalive' -> 'i am alive', only one possibility
+* Concatenation: 'theylookedible' -> 'they look edible', but 'looked' and 'edible' concatenate.
+
+I could do swaps of n-grams, start with the longest words and go downwards:
+
+> example: 'iamalive'
+
+1. swap of len(s) = 8, 'iamalive' in dict? no
+2. swap = 7, 'iamaliv', 'amalive' in dict? no
+3. swap = 6, 'iamali', 'amaliv', 'malive' in dict? no
+4. swap = 5, ¡'iamal', ..., 'alive'. 'alive' found and saved. How?
+5. swap = 4, nothing
+6. swap = 3, nothing
+7. swap = 2, 'am' found
+8. swap = 1, 'i' found.
+
+All the string is 'covered', done.
+
+> example: 'theylookedible'
+
+1. swap of len(s) = 10, nothing
+2. wap = 6, 'looked', 'edible' saved
+3. swap = 4, 'they', 'look'
+4. swap = 3, 'the'
+5. swap = 2, 'he' found
+6. swap = 1, nothing
+
+How to resolve the conflict of swap = 6? We could look for 'look' and 'ible'. If we find none, choose one randomly and 4 unrecognized words. If we find 'look', assign 'look' and 'edible' and discard 'looked'. How to save these? Don't know. This algorithm is O(n<sup>2</sup>)
+
+Also, once the substrings are saved, how to check for string completeness to break the loop and return?
\ No newline at end of file
diff --git a/17. Hard problems/17.13. re_space.py b/17. Hard problems/17.13. re_space.py
new file mode 100644
index 0000000..babc802
--- /dev/null
+++ b/17. Hard problems/17.13. re_space.py	
@@ -0,0 +1,52 @@
+import unittest
+
+dct = ['a', 'i', 'am', 'not', 'he', 'the', 'they', 'my', 
+      'alive', 'darkness', 'friend', 'hello']
+
+def re_space(s):
+
+    s = s.lower()
+    if s in dct:
+        return s
+
+    ins = s
+    keywords = []
+    n = len(ins)
+    for i in range(n-1, 0, -1):
+        j = 0
+        while j + i <= n:
+            tmp = ins[j:j+i]
+            if tmp in dct:
+                keywords.append(tmp)
+                ins = ins.replace(tmp, '')
+            j += 1
+        n = len(ins)
+
+    keywords = sorted(keywords)
+    touched = [0] * len(s)
+    for keyw in keywords:
+        idx = s.find(keyw)
+        if touched[idx] == 1:
+            continue
+        touched[idx:idx + len(keyw) + 1] = [1] * len(keyw)
+        s = s.replace(keyw, ' ' + keyw + ' ')
+    s = s.replace('  ', ' ')
+
+    if s[-1] == ' ':
+        s = s[:-1]
+    if s[0] == ' ':
+        s = s[1:]
+    return s
+
+class Test(unittest.TestCase):
+    data = [('iamalive', 'i am alive'),
+            ('hellodarknessmyfriend', 'hello darkness my friend'),
+            ('invincibleami', 'i nv i nc i ble am i')]
+
+    def test_rotate_matrix(self):
+        for test_num, expected in self.data:
+            actual = re_space(test_num)
+            self.assertEqual(actual, expected)
+
+if __name__ == "__main__":
+    unittest.main()