Seriously, though, looping through the dictionary is really a bad solution. Maybe reading the dictionary into an associative array and matching on index or something like that would be a better idea for speed. Loop through words in the post instead of words in the dictionary.

The major limitation here is that it requires a word to be in the dictionary to be considered correct. Non-optimal. No pluralization handling (other than brute forcing by adding to the dictionary), no possessive case (as was noted in the code), and no new-line handling (hint: simply strip all newlines from input, replace with a space, and be done with it. Check for hypens before a new-line to detect continued words; though, that is unlikely in a web post. Just don't use the output directly into a post. Maintain a control copy.)

Also, I can't download the source directly (file not found). I had to edit the URL to make it work. You need to fix the link by removing the second instance of shuzak.com.