Not even close to 4K, I was just really bored this evening... Too lazy to add command line arguments, however.

import urllib, urllib2, re

def babelfish(langpair, text):
    url="http://babelfish.altavista.com/tr?lp="+langpair+"&trtext="+urllib.quote(text.encode('utf-8'))
    html = urllib2.urlopen(urllib2.Request(url,None,{'Accept-charset':'utf-8'})).read()
    m = re.compile("white class.s..div.style.padding.10px..([^<]*)</div",re.U).search(html)
    return m.groups()[0].decode('utf-8').replace(u'\n',u' ')

def log(f, num, text, encoding='utf-8'):
    out = (unicode(num)+u': '+text+u"\r\n")
    print out.strip().encode('ascii','replace')
    f.write(out.encode(encoding))
    f.flush()

def mutate(text, steps, langpairs, outfile):
    t = unicode(text)
    f = open(outfile,"wb")
    n = 1
    log(f, 0, t)
    history = [t]
    br = False
    while not br:
        for lp in langpairs:
            t = babelfish(lp, t)
            log(f, n, t)
            if n==steps: break
            n += 1
            if t in history:
                br = True
                break
            history.append(t)
        if n==steps: break
    if br: log(f, n, u"[loop found]")
    f.close()

mutate("Hay guys what's going on in this thread?", 20, ['en_ja','ja_en'], "nonsense.txt")