AI Fluency & Building · Prompt Engineering
Go deeper. Real-world tasks that need more research, planning and a polished result.
Task 1 of 3
Prompts for 5 marketing jobs, tracked over time.
What to do
Build a versioned prompt library for 5 marketing workflows (content calendar, ad copy variations, email newsletter, SEO blog outline, customer FAQ) for a brand you choose, with a changelog of every improvement.
You’ll make
Done looks like
A library of 5 prompts with version numbers and a changelog.
Task 2 of 3
Score a prompt on 15 cases and raise the score.
What to do
Create a 15-case test set for one prompt (for example, product descriptions) with clear pass/fail rules. Run it in a spreadsheet or promptfoo, improve the prompt, and show the score going up.
You’ll make
Done looks like
A 15-case test set with pass/fail rules and before/after scores.
Task 3 of 3
Zero-shot, few-shot and step-by-step prompts head to head.
What to do
Compare zero-shot, few-shot and step-by-step prompting on one job, such as sorting 30 customer reviews into positive, negative and feature requests. Publish the accuracy of each method.
You’ll make
Done looks like
A results table for 30 examples and a published accuracy comparison.
Learn before you build
Free course
A free open-source prompting course from basics to advanced.
Free course
A free 1.5-hour course by DeepLearning.AI and OpenAI.
Free tool
Open-source tool to test prompts and AI apps against test cases.
Code repo
Hands-on notebooks that take you from basics to advanced techniques.
Article guide
Official techniques: clarity, examples, XML tags, chaining and more.
Article guide
The most complete open guide to prompting techniques and research.