A GPT model is a generative pretrained transformer language model. Generative means it creates text or code, pretrained means it learned patterns from huge amounts of text and programming data, and transformer means it prioritizes the important parts of an input request. section 1.1
GPT generates responses by predicting what word or phrase usually comes next. This makes it particularly useful for producing predictable programming code, including HTML, CSS, and JavaScript, from plain-English instructions. section 1.1
A practical GPT workflow uses a short improvement cycle:
- Prompt: describe the page, content, or research you want. section 1.5
- Copy: move the generated code into a file. section 1.5
- Save: store the code as an HTML file. section 1.5
- Deploy: upload the HTML file to a web hosting provider, then refine the prompt when the page needs changes. section 1.5
Research prompt
Ask: “Help me research the history of coffee for a general audience. I’m interested in its discovery, global spread, and cultural impact. Provide key events, notable varieties, engaging trivia, and links to sources.” A web-connected GPT service can return links for fact-checking and further research. section 7.2.3
Help me research [topic] for a general audience. I'm interested in A, B, and C. Please provide X, Y, and Z. Also, include any interesting trivia or anecdotes that might be engaging for readers. The goal is to provide a depth of information that is accurate and detailed but also easily understood by a broad audience. Please provide links to your sources.
Use the reusable research-prompt template to specify the audience, broad themes, detailed aspects, desired anecdotes, and source links. section 7.2.3
