Zulip Chat Archive

Stream: new members

Topic: Blog post: The Math Is Haunted


Dan Abramov (Jul 30 2025 at 20:40):

Wrote my first blog post about Lean: https://overreacted.io/the-math-is-haunted/

It might be a bit messy and I'm not sure it teaches a lot but I wanted to get it out of my head so I can move to other topics.

Any comments/corrections are appreciated!

Arthur Paulino (Jul 30 2025 at 20:56):

Just some context: Lean 3, the version that got traction with mathlib, was first and foremost an interactive theorem prover. Lean 4, as a full blown programming language, is a relatively recent breakthrough.

suhr (Jul 30 2025 at 20:59):

Every dependent type theory is a programming language. What makes Lean different is that it's a usable programming language.

Dan Abramov (Jul 30 2025 at 21:00):

Yeah, that's fair! I appreciate the breakthrough-ness of the breakthrough although it also makes explaining what Lean is a bit tricky. I think positioning it as primarily a theorem prover which also gained the runnable capability (same as the historical context) is maybe the least confusing way to explain it to programmers.

Arthur Paulino (Jul 30 2025 at 21:04):

suhr said:

Every dependent type theory is a programming language. What makes Lean different is that it's a usable programming language.

Yes, that's encoded in my imprecise terminology "full blown programming language". I mean, there was pretty much no special reason for some programmer to use Lean 3 to build an application. The main point I wanted to make was that, if you have some extra context, Lean 4 being a programming language used mostly by mathematicians is the most likely outcome.

Dan Abramov (Jul 30 2025 at 21:52):

HN discussion, which might be a good place to engage with folks curious about Lean: https://news.ycombinator.com/item?id=44739315#44739780


Last updated: Dec 20 2025 at 21:32 UTC