Technical debt is a huge problem, when it is, well ... huge. What usually happens is this: we start a project, we work on its "prototype," we wait until it is fully ready, and then we start thinking about unit tests, continuous integration, and build automation. But it is too late because the product already is too big. The problem is that we hold the product in the prototype mode for too long. I believe that two weeks is the maximum a prototype should take. After that, it's a normal product, not a proof-of-concept anymore.

The video is here: https://youtu.be/4MP3xXGrpCY