Ask HN: If you had 5 years of uninterrupted time, what would you build and why?

I would build a comprehensive framework for teaching Hanzi/Kanji to non-native learners of languages that use Chinese characters.

Anecdotally, I've seen what I assume is a disproportionate amount of passionate learners of languages that use Chinese characters give up their studies after a few years, and many of those I talked to cited difficulty in learning/using/recalling the meanings and readings of Hanzi/Kanji. There are many tools and strategies for learning Chinese characters already in existence (using SRS, mnemonics, calligraphy, etc.) but amongst students in an academic setting, their use seemed fairly minimal when I was a student (a couple of years ago).

If I had 5 years to build something, I would bring together a system using both new approaches enabled by uniting the various existant methodologies and tried-and-true methods that could take you from complete ignorance of Kanji/Hanzi to a degree of reading/writing fluency over a few years. I consider a push like this to be akin to the shift we are seeing in learning methodologies used to teach programming to newbies like Codecademy and TryRuby that rely on hands-on learning rather than lecture learning or trial-and-error learning.

Pre-built algorithms for procedural content generation would be a must-have!

For money: a system that manages data for preparing 510(k) submissions to the FDA for new medical devices and automates the process as much as possible. From watching how much effort my multi $Bn employer goes through to do this, there must be many, many smaller companies that suffer orders of magnitude more than we do. And they'd pay a handsome amount for a tool.

-or-

I'd find out just what it is that so many people hate about their CAM software and fix it. Bonus points if I get to build hardware as well: iPhone interface to a Haas OfficeMill anyone?

For fun: a walking robot with high payload (> 1/2 ton) capacity.

That is related to one of mine: an email/messaging client that doesn't suck. Or rather, multiple clients based on libraries for ‘mailboxes’ of various kinds (including those in which messages appear asynchronously, like mail and messaging servers, RSS, usenet, scrapable web forums, and so on). I don't know how continuous AV links would fit in this framework, though.

Another five-year software project I'd have fun with would be LLVM front ends for various historically interesting programming languages, and likewise code generators for dead or niche processors.

If I also had money for expertise and prototyping, I'd like to make a keyswitch that feels like an IBM Selectric (or at very least an IBM beam spring switch) but fits ISO's thickness requirement so that it's salable.

Computer vision, and augmented reality.

Real-time 3d reconstruction from simple video is currently possible with desktop computing power. I suspect that within 5 years it'll be possible with mobile computing power (note that the real dependency here is power efficiency). Obviously textured light techniques are even more powerful.

With projects like Glass and Myo, wearable computing is coming together.

We have the conceptual pieces we need to do useful augmented reality. Start by modelling lots of the world, both the geometry and also object categorization (the latter, admittedly, is still evolving fuzzily). Then build an app framework, for apps to help people execute tasks. One obvious example is step-by-step overlay instructions for doing repairs (changing your own car oil isn't that hard, right, but it's too intimidating for many people).

I think the short film Sight gets it mostly right, except I'm not talking about cybernetic augmentation, only wearable computing: http://www.youtube.com/watch?v=AFiE82Npbn4

There's lots of potential there.