

Language models are unsuitable for math problems broadly speaking. We already have good technology solutions for that category of problems. Luckily, you can combine the two - prompt the model to write a program that solves your math problem, then execute it. You’re likely to see a lot more success using this approach.
As I understand it, Google mostly ships new stuff that they let die because it’s one of the only ways to get a promotion at Google - to ship a product.
Once shipped, the newly promoted staff moves on to something else, and the business people take a look and see if the product actually makes any sense from a financial perspective, which is rarely the case.