When to rewrite and when to refactor
Rewrites are almost always the wrong call, with a narrow set of exceptions that are worth naming precisely.
Estimating work honestly
The estimate is a probability distribution, not a number. Saying so up front makes the conversation at week six much easier.
Documentation people will actually read
Nobody reads a wiki. They read the README in the repository they just cloned, so that is where the important things go.
What we look for in due diligence
Test coverage matters less than deployment frequency. A team that ships daily has fixed things a test suite would only have described.
Technical debt is a decision, not an accident
Every shortcut is a trade made under time pressure. The problem is not taking them, it is not writing down which ones you took.