Absolute limit of 60k means it's not worth investing in if there's a chance you will hit that limit.

Somehow, Google APIs always end up broken like that. Why don't they just have the balls to scale up like Amazon does?

I guess it might change. Google App Engine is finally beginning to make sense, be it mainly for intranet applications.

I see you two have some pretty big plans on your mind! Care to share what is it you're building that 60000 predictions per day just won't cut it?

... easy, benchmarking your own algos against the prediction API on large-scale data sets. I presume that's what 90% of the people who received their account during testing did anyway :-D I did 15k requests in about 10mins at times.

it also means that it's not something that they're using internally for anything important either

Does anyone have examples of apps that actually use the prediction API?

Wanted to use it with Google Analytics for a kind of recommendation engine of the form "users that liked this page also liked...". But they never approved my request and a 60k request limit won't cut it either.

an alternative: http://www.directededge.com/ (YC startup)

I'm using it in an app, since my request did get approved, but it's very rudimentary at the moment as I am still exploring it. I'll be sure to report back to HN at the point it is in full use.

To me a big weakness of the prediction API when I checked it out was that it cannot handle (i) missing values or (ii) sparse data formats. Both are very common in real life data.

Does anyone know what machine learning algorithms they use?

btw, you don't need Prediction API to detect a language. One could use Translate API [1]:

  >>> detect_language_v2("Tous pour un, un pour tous, c'est notre devise", open('api_key.txt').read().rstrip())
  [u'fr']

[1]: http://stackoverflow.com/questions/4545977/python-can-i-dete...