Apart from playing around with these models to test their capabilities, how are people using them - either for personal use or in production settings?

EDIT: Specifically interested in how people are using the reasoning models (o1, DeepSeek) in ways that GPT-4 / 4o wasn't already good at.