It got you 90% of the way there once you’d translated the problem. I can’t imagine LLMs ultimately saved more time per problem than WA did.