https://youtu.be/PGv5ymOhaNA

Prompting is hard. And tedious. Should I tip my model? Use in-context examples? We've worked with dozens of companies building and evaluating AI-driven applications, and the prompt engineering step is always the most frustrating. It takes tribal knowledge and guessing and checking to get right.

Think of endoftext as Grammarly for your prompts. As you write your prompt, we provide suggestions on improving it - simplifying instructions, clarifying the output format, etc. You can apply these suggestions to rewrite your prompt with a fix automatically. We also suggest new test cases to identify how your prompt might fail. Through your interactions, endoftext learns what types of prompt patterns and test cases you care about.

The endoftext beta has a set of simple suggestions and test case generations. We are actively working on adding more useful suggestions, and we'd love your feedback on what you'd like to see!

You can start using it at https://app.endoftext.app