[Barooahn]

The jump in writing quality between GPT-4 and GPT-5.1 is hugely noticeable if you're specific with prompting (requires a bit of trial and error).

I've been using it to generate children's stories and reading comprehension questions for a UK curriculum app. The difference:

• GPT-4: Generic questions, often not engaging enough for 7-year-olds • GPT-5.1: Adapts tone, vocabulary, and complexity when you give it curriculum constraints

I realised the model isn't "smart enough" just to throw in a prompt like "write a story about..." - I needed structured prompts with specific examples of the target style.

It's still not perfect, but it's a massive improvement and can be called genuinely useful vs. just a novelty.