I was in the happy position of finding a job (post being a failed academic) doing stuff at the intersection of architecture and algorithms (high-speed regex implementation) and getting to to this for 11 years straight (2 years as a researcher, 5 years as a startup CTO, 4 years as a Principal Engineer at Intel). Got a decent if not ginormous exit too, which was nice. Also got to IMO change the industry - Hyperscan (github.com/intel/hyperscan) is used in a lot of places.

This nice run has come to an end for me and my team and it's made me realize that getting to work on nifty, complex algorithms and do real CS is not the default. To say the least.

I'd like to try to figure out how to do it again: the "algorithms startup" is a fun thing to do. You aren't going to be raking it in as having the 'nice library to do task X' doesn't allow you to capture huge amounts of value relative to the people selling boxen/UI/SAAS/etc. that wraps up your library. But you can work on a decent scale.

I am thinking about blogging about this; generally I've been sticking to tech stuff on branchfree.org so far but having done this successfully once makes me wonder if it could be done again.