Okay so I keep seeing this debated in various corners of the type theory internet and I want an actual honest thread about it, not just tribalism. It's 2025. Both tools are mature-ish. Which one are you actually reaching for when you start a new project?
My situation: I do mostly pure math formalization โ algebraic topology, some category theory. I've been on Lean 4 + Mathlib4 for about 18 months now. The community momentum is genuinely hard to overstate. Mathlib4 has crossed 210,000 formalized theorems and 100,000 definitions at this point, and the contributor base keeps growing. The simp / ring / linarith combo handles so much of the boring arithmetic that I barely think about it anymore.
That said, I still feel a nagging respect for Rocq. The calculus of constructions foundations feel cleaner to me philosophically, and SSReflect is an absolute masterpiece of tactic design once you internalize it. I just... don't have time to live in two ecosystems.
So: Lean 4 or Rocq in 2025? And why?