This is a concept puzzle product designed to improve children's dictation ability by using GPT-3 language model. The language model uses artificial intelligence to synthesize text similar to human text. It was announced by OpenAI in May 2020 and is the most powerful language model designed so far. In short, it works similar to the auto-completion function of a mobile phone, and can predict the next text based on previous data, but it is very powerful because it only needs two or three words to complete the entire article, Because it refers to all available books and millions of research and web pages. At present, most of the GPT-3 language models are applied to APP. It is assumed that parents can write a word. The product can be recognized and composed of words or idioms through technology to be played so that children can write it. It can not only promote the feelings between parents and children, but also improve children's dictation ability and enrich children's knowledge.