Introducing `claude-prompt-engineer` ✍️ An agent that creates optimal Claude 3 prompts. Just describe a task, and a chain of AIs will: - Generate many possible prompts - Test them in a ranked tournament - Return the best one And it's open-source:

`claude-prompt-engineer` is a constrained agent -- meaning its behavior is highly-controlled, leading to better results than open-ended agents. It chains together lots of Claude 3 calls that work together to find the best possible prompt.

It's based on my previous `gpt-prompt-engineer` that went viral (5k+ stars on Github) with a few massive improvements: - Uses Claude 3 (writes far better prompts than GPT-4) - Auto-generates test cases, further automating the prompt eng process - Supports multi-variable prompts!

How it works, in a nutshell: - The user describes the task and input variables - Claude generates many candidate prompts to try - Each prompt generates against each test case, and the outputs are compared (by Claude!) for each combo, ELO tournament-style - Highest score wins!

I've been blown away by the quality of the generated prompts. Often, they outperform the prompts I'd write by hand (especially when I ask it to generate and compare 10+ prompts). I am already using this for my day-to-day prompting work... it takes the guesswork out of it.

If you'd like to try it or contribute, check out the Github repo.

How does it compare to dspy?

Haven’t worked with dspy yet. Do you like it?

This is again... great @mattshumer_ Will cover this on tomorrow's #ThursdAI 🫡

Thanks Alex! Happy to jump on tomorrow

What are those little things running across the top near the menu bar?

It's a fun setting I turned on in Colab years ago, I forget what it's called lol