In practice people I know who've built startups on top of tech they built at universities haven't run into problems. A common strategy is to open-source any cod

In practice people I know who've built startups on top of tech they built at universities haven't run into problems. A common strategy is to open-source any code you might want to later use commercially, as a way of throwing it over the fence. Your startup can then use it on whatever terms you open-sourced it under (probably BSD/MIT, if you want to give your later self maximum flexibility). Granted, so can anyone else. But usually the research-prototype code itself isn't enough to build a competitor, and the common case for research code open-sourced by a grad student in some random .tar.gz code dump on their website (or worse, included as a file appendix to their thesis), is that nobody else will even so much as look at it. The university may have to approve this, but at major CS departments approving students to open-source the code that goes with their thesis is pretty rubber-stamp, at least if it's self-contained.