What if I don't like the generated article?
There is still trial and error with the generation. Because the algorithm is a complete black box (as all machine-learning algorithms are), you need to test to see what works which is why we have added a 10% allowance for gibberish articles.
See our Best Practices for Seeding.
Either the generated article was good or it was gibberish, don't forget to rate them as this helps improve the system!