This year, and for the first time ever, we are applying to Google Summer of Code!
We already have two students that are interested in participating but if there is anyone else then do let us know :)
Me and Araq are the organisation admins this year as well as mentors and zahary is another mentor.
The ideas page is available on the github wiki here. I think we currently have enough mentors but having more can never hurt so if you would like to be one let me know.
It's a pity that those proposals were rejected. Maybe they'll be picked up by adventurous programmers anyways. The OpenCL codegen in particular would make Nimrod more interesting to many.
Nimrod is a very nice language and I expect (hope ;-) that in the next few years there will be many more wins than losses.
For anybody interested I just attended the GSoC rejection meeting. Here are the reasons why we were rejected: https://gist.github.com/dom96/9273843. The primary reason is that our ideas page just wasn't better than the ideas page of other projects, and as carols mentions we had stiff competition. But I'll let you read the logs for yourself, I'm sure that next year we'll do better.
The rest of the logs should be available here (2014-02-28) in case you would like to read why the other orgs were rejected.
Looking back through the Google Summer of code ideas I find little mention of the javascript backend for the compiler (in fact, as side bonus points for a different task). Ever since I started reading about Nimrod I've seen "oh, we could also generate js", but I've never seen yet a real example. Just like this forum is an example of what can be done with jester it may be a good idea to find a practical example of Nimrod code compiled to javascript to make this backend more visible to users.
What seems like a simple task would be to improve this very forum with local client javascript which in similar fashion to http://stackoverflow.com renders the HTML of the rst text being typed in real time. In theory this seems an easy and appropriate task: it could already be done with server support where through AJAX or other trickery a server daemon could render the text and return it for the HTML preview. The rst parsing/rendering code doesn't seem to be particularly dangerous or dependant on native modules, would there be any other problem with javascript rst rendering?
As usual the Google Summer of code ideas seems to be quite buried. The main website links to the Nimrod roadmap. The GSoC document could be linked from that. Also the section How to help from the website doesn't mention these documents, and I wonder how many people are reluctant to dive into IRC to learn about these fun opportunities of working for free.
A partially completed ISAAC implementation for Nimrod can be found here: https://github.com/Zshazz/Ramrod
I'll echo Mścigniew's sentiments regarding ISAAC and add that, at least from my tests, ISAAC is competitive with regards to speed to even non "crypto rngs." For instance, once you pay the high initial cost of ISAAC, it's actually FASTER than the Well512 implementation in the above link. Personally, I used to think Well512 (and similar RNGs like Mersenne Twister) were ideal general-use algorithms, but unless you're just pulling a very few numbers out of the generator, ISAAC is probably an appropriate replacement even for times where you don't need the additional security guarantees.
I'll also add that until both the algorithm and the particular implementation in question have been thoroughly reviewed by a few specializing in security/crypto, you should not use such things for crypto. So it's a really good thing that ISAAC is fast or there wouldn't be a good reason for anyone to use it.