Also, implementing ideas and testing them on real data can be a very good heuristic to drive theoretical thinking by:
- giving you empirical evidences on the practical limitations of a positive theoretical results that might have been otherwise overlooked.
- giving you new insights on which approaches (e.g. class of algorithm) are the most promising for a given class of problems.
- make explicit some assumptions you might have implicitly made in your theoretical work and check whether they can actually be satisfied in practice.
When I said "concrete work" I didn't necessarily mean you can't do theoretical work, if that is your interest. I meant to say that you should actually be working together. It is very common for people to be part of a "single project" but the extent of their collaboration is that they have some meetings and talk about related issues. I don't think that is the same as working together.
It's a very good recommendation. I remember some PhD students reluctant to work on concrete implementations of their theorized approaches. But in the long-term, the most successful ones were doing a lot of concrete works and gained experiences for an academic career and while keeping a strong feet/competence into the private sector.