Mindket
English
English简体中文
Sign inSign up
StoreMCPSkillsCalculatorLearnPriceBlog
Discover
Research308Writing59Business95Legal13Finance & accounting33Marketing & SEO172Developer1,379Productivity382Education81Companions14
Help and support
Language
English简体中文
IntermediateFree

Prompt optimization: let AI write, improve and test your prompts

Mindket editors · Updated · 4 lessons · 34 min

A black banner titled "Prompt optimization" over the line "Generate, improve and test prompts with AI".

This free course has four lessons on prompt optimization, for people who already write prompts and want a faster way to improve them. You will generate a first prompt with a model and improve it with OpenAI's optimizer or an edit meta-prompt. Then you compare versions on a small test set and let DSPy search for better ones against a metric you define.

What you'll be able to do

  • Generate a first prompt from a task description with Anthropic's metaprompt or OpenAI's prompt generation.
  • Improve an existing prompt with OpenAI's prompt optimizer or its edit meta-prompt, read what changed, and know what Anthropic says its prompt improver does.
  • Compare prompt versions on a small test set in a spreadsheet, a console or promptfoo before you switch.
  • Run DSPy's MIPROv2 optimizer to tune a prompt against your own metric.
All courses

Who it's for

For people who already write prompts for ChatGPT, Claude or a model API and want to improve them with less guessing. Know the basics of prompting first. The first two lessons work in a web console or a chat window, though lesson 1 also offers a notebook where you edit a few lines. Lesson 3 has a no-code route and one that uses a terminal and a YAML file, and lesson 4 uses Python.

Lessons

  1. 1AI prompt generator: let a model write your first promptAn AI prompt generator writes a first prompt from your task. Run Anthropic's metaprompt or OpenAI's Generate meta-prompt, then check what it wrote.7 min
  2. 2How to improve a prompt with AI: OpenAI's optimizer and Anthropic's improverHow to improve a prompt with AI: OpenAI's prompt optimizer and Anthropic's prompt improver, what changed in OpenAI's own example, and what to check.9 min
  3. 3How to test prompts: compare versions on a small test set before you switchHow to test prompts: build a small test set, run two prompt versions on it in a spreadsheet, Anthropic's Console or promptfoo, then read the results.7 min
  4. 4DSPy tutorial: let code and a metric optimize your promptsA DSPy tutorial for beginners: write a signature, pick a module, define a metric and let an optimizer such as MIPROv2 improve your prompt for you.11 min

Frequently asked questions

Is this prompt optimization course free?

Yes. Every lesson is free to read. Some tools in the lessons, such as model APIs, may bill for their use, and each lesson says where.

What is prompt optimization?

It is improving a prompt on purpose. A model drafts it, an optimizer rewrites it or a search finds better versions against a score, and you test the result against the old one before you switch.

Do I need to code?

Not for the first two lessons, which work in a web console or a chat window. Lesson 1 also offers a notebook where you edit a few lines. Lesson 3 has a spreadsheet route and a terminal route, and lesson 4 is a Python tutorial that you can read for the idea if you do not write Python.

Does an AI-optimized prompt always beat the original?

No. OpenAI's optimizer guide says the optimized prompt can perform worse than your original on specific inputs. That is why lesson 3 tests versions before you switch.

What should I learn after this course?

The context engineering course is next. It widens the view from one prompt to everything an AI agent sees at each step.

Share this course

Compare your prompts with working ones

Skills on Mindket are folders of instructions, examples and files. Open a few and compare how their authors word a task with your own prompt.

Browse skills

All courses · All articles