Especially with a hard to control textual query interface I suppose that a huge number of requests will be burned through for finding the correct wording to make Wolfram Alpha understand what you are trying to query.
They should really make a free dev account which either operates on a limited set of data, or deliberately delivers wrong numbers in their results.
A dev just needs a way to test his app without having to be careful about everyone of his tries costing real money.
The prices seem insanely high, and the most expensive plan also has a bizarre structure (http://products.wolframalpha.com/api/pricing.html). After you pass your initial 10 million request quota, the per request price is higher than during the quota! This means if you're doing 50 million requests, it's cheaper to buy 5 API keys w/10 million requests than to have one API key doing all 50 million. I would've expected that Wolfram encourage higher usage w/deeper discounts especially when you reach the top, not until you reach the top.
I also wonder if ~6 cents/request is supply and demand or maybe gree, or if the system is so complex and sophisticated that they actually need to charge that much to pay for the computational requirements.
All their plans have that structure. I'm guessing that they want to be able to plan what resources to give to Alpha on the basis of what their customers have signed up for, so they want to discourage "overuse". Also, you pay up front for the quota and presumably pay in arrears for any extras, so obviously they want to encourage you to pay earlier.
I find it amusing that they actually seem to think $60 is the right price point for "students, hobbyists, individual developers". How out of touch with the community can you get.
Looking at their visitor stats (http://siteanalytics.compete.com/wolframalpha.com/) I'd guess they have unused capacity for at least 2MM queries a month if back in May each visitor was hitting it twice. I don't get the impression they are using an infrastructure that they only pay for on demand.
If they want anyone to use their API (or their service as a whole) they should give away at least 1000 queries a month to the first 2000 developers that sign up.
What has been created here is a new kind of API, one the result of a lifetime of experience, and one that will no doubt influence the way you think about programming for years to come.
How fortunate you all must feel to use this API, and to interact with this amazing, lifechanging software. I don't think I exaggerate when I say that within this API lies the ability to understand our very universe.
I went ahead and posted this comment on the blog post. It'd be great to be able to at least have a sandbox to build a library for communicating with the API to test out libraries and clients under development. Without something like that, I really don't see open, robust client libraries being built, and driving adoption in applications and around the web.
---
I'm surprised that no one has said this yet, so I'll step up.
This is massively expensive.
Releasing an API without offering developers any way to build client libraries for Ruby, Python, and other languages short of ponying up $60 to test their client (or even offering a sandbox for sample data) is a bit much.
I'd be interested in putting together a Ruby library for interfacing with Alpha that would abstract the work of creating a request and parsing the response, making integration a snap for other developers and enabling deeper integration with Alpha around the web. I've thought a bit about how that might look, too:
But with such a strong financial disincentive to develop for the platform, it's not likely that I'll register as a developer to create something like this. Feel free to contact me if terms change.
They have temporary free plans called Pioneer Grants but you have to "discuss your application" with a representative. Oh silly wolfram, wasn't it enough to see microsoft learn a lesson the hard way from google?
I was looking forward to using the API in my Firefox add-on that shows the Wolfram Alpha results next to the Google results. Quick calculation of the cost: 18,000 users/day * say 3 searches/day = ~$2,300/day.
I think I'll stick with loading it in an iframe. I would have thought that this kind of thing fit perfectly with the example use case "augmenting web and meta-web search with computed knowledge" but I guess they have a different target market.
Their home edition is $300, but unlimited requests. Their API version is the equivalent of 5000 total requests over 5 months, or 4000 requests in 1 month ($240 for 3000 overage requests)
Wolfram|Alpha uses algorithmic computation! Sign me up!
"Wolfram|Alpha uses a unique paradigm for modeling data, based on algorithmic computation, that allows for implicit relationships among the data and the emergence of new knowledge."
Wolfram is industry, not Academia. They're used to billing sectors like aeronautics and oil for gazzilions. To them "web startups" are nobodies, and the price is there to keep the vagrants at bay.