[mindcrime]

One idea you could try is this: take an "old" idea and "remix" it, improve it, etc. The book Rethink[1] has a lot on this. And it makes sense.. in computing, things really tend to go in cycles and what's old is often new again. Take Neural Networks for example. This is an idea that's so old it's practically archaic. But add some new ideas on training (backprop, discovered in the 80's or so), along with hardware advances due to Moore's Law and the advent of the GPU, as well as the surge in data volumes enabled by the Internet, always-connected devices and advances in storage technology, and BOOM... you have the hottest topic in the world today.

So go back in time... read old computing books, read old issues of InfoWorld or Network Computing from the 90's or early 2000's, or find some old issues of Byte magazine on Ebay, and mine through those... and find an old idea that can be combined with some new advances or new technologies, to make something that's mondo radical.

[1]: https://www.amazon.com/Rethink-Surprising-History-New-Ideas/...

[theoh]

I'm sure I've read a jokey guide to succeeding as a CS grad student that tells you to do exactly this, but tongue-in-cheek. Maybe someone else will recall the URL.