Ask HN: Idea for improvement on HN reading.

In a nutshell: Sort by being able to disallow keywords, usernames, dates, etc. to get the stories you'll like the most first when you check.

In my my brief research (i.e. Google search) it doesn't look like HN has an API. Possibly scraping the data using scRUBYt! (for ROR) or a similar technique would be effective enough. Then, you would re-weight the data based on additional specifications.

If you took the last x00 submissions (or top x00 in current popularity) and then taught the application what you did and didn't like, it could remove, raise or lower results as they came in.

As an example: In my case, Look at HN about twice a day. I generally like submissions with "Ask HN", "Tell HN" and "SASS" in the title. I don't like submissions discussing "Clojure" or "iPad" in the title. They aren't relevant to me. With my example, after scraping the data, the application could bump anything with a keyword I like up the list by 20%, and anything I don't like down the list by 50% (or 100% if I really hate everything on such a subject). There's so much more that could be said, but it seems quite simple.

I'm a designer mainly, and wannabe hacker. Any thoughts? Is this frowned upon by HN in general? Want to make it a reality (you hack, I design)?