A GPT-powered agent capable of improving itself by engineering its prompt and adjusting its model settings.
- Execute a given task. For the initial execution, the provided task is used as the prompt.
- Based on the current prompt and the execution, improve the prompt.
- Based on the current settings and the new prompt, improve the settings.