Pre-order · Email Course for Developers
10 Prompts in 10 Days
Go from operating on vibes to measuring whether your prompts actually work. Ten prompt engineering patterns — from structured outputs to LLM-as-judge evals — each with working Anthropic Python SDK code and a hands-on exercise. One email a day, 15 minutes each.
Most developers call the API with a bare prompt and hope for the best. These ten patterns give you control over what comes back.
client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
messages=[{
"role": "user",
"content": "Suggest a cocktail
for a dinner party"
}]
)
client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
system="You are a cocktail
consultant. You ask about the
occasion, guest preferences, and
available ingredients before
recommending drinks.",
messages=[{
"role": "user",
"content": "Suggest a cocktail
for a dinner party"
}]
)
Who this is for
You write production code and you’re already calling the Anthropic API — but you feel like you’re leaving capability on the table. Your prompts work, mostly. The results are fine, sometimes. You want them to be reliably good.
This course is for developers who want to understand why prompts work, not just collect templates.
What you’ll walk away with
Ten prompt patterns that cover the situations developers actually face: structured output, chain-of-thought reasoning, tool use, multi-turn conversations, and evaluation. Each pattern comes with working Anthropic Python SDK code, a real before/after example, and a short exercise.
By Day 10 you’ll be stacking techniques — structured output with chain-of-thought reasoning, few-shot examples with multi-pass refinement — in ways that felt out of reach on Day 1.
Every lesson teaches a technique you can use that day. Each exercise fits in 15 minutes.
10 patterns that compound
Each day builds on the last. By Day 10 you are combining techniques that would have felt impossible on Day 1.
Days 1–3: Foundations
The System Prompt
The first thing you configure in every API call. Set identity, tone, and boundaries that shape every response.
The Constraint Prompt
Stop getting cocktail essays when you wanted three recipes. Explicit boundaries that eliminate noise.
The Few-Shot Prompt
Show, don't tell. Teach the model your exact output format with examples instead of paragraphs of instructions.
Days 4–6: Building a Real App
The Structured Output Prompt
Turn text responses into typed Python objects your app can render, sort, and filter.
The Chain-of-Thought Prompt
Force step-by-step reasoning. The single biggest quality upgrade for complex recommendations.
The Multi-Turn Prompt
Build a conversation loop where the model gathers requirements before acting. This is where it starts feeling like a product.
Days 7–9: Production Patterns
The Tool Use Prompt
Give the model access to real data. Stop it from making up ingredients and start returning actual recipes.
The Multi-Pass Prompt
Generate, critique, improve. Chain API calls into a pipeline where each pass refines the last.
The Evaluation Prompt
Measure prompt quality with test cases and LLM-as-judge. Turn prompting from art into engineering.
Day 10: Synthesis
The Compound Prompt
Combine everything into a production-ready prompt. All nine patterns working together in one application.
Built for working developers
Each email gives you a pattern, working code, and a 15-minute exercise. Use what you learn that same afternoon.
Runnable code examples
Every pattern includes working Anthropic Python SDK code you can run immediately with your API key.
Before/after examples
Real API output showing the concrete difference each pattern makes. See exactly how each technique changes what the model gives you back.
Composable patterns
Each technique stacks with the others. By Day 10 you are combining four or five patterns in a single prompt naturally.
Learn by doing
Every email ends with a quick exercise using code you already have open. Open your editor and start.
What people are saying
"I thought the training was excellent in all respects. I especially appreciated the informal Q&A that ran through the whole thing, where your expertise shone through and your relaxed, informal style made everyone comfortable asking the simplest questions."
— Colm O'Cinneide, PhD, Professor of Mathematics, Columbia University
"He understands how to teach a large audience and is able to adapt his delivery to the members of the class."
— Kate Strachnyi, Story By Data · LinkedIn's Top Voice in Data Science
"Dan was both my manager and a mentor as I transitioned from data science to software engineering. He combines a deep understanding of the technical aspects of system design with an ability to explain the real-world tradeoffs of decisions, which accelerated my learning."
— Aaron Schwartz, PhD, Software Engineer, .txt
Dan Gerlanc
I have taught 1,000s of developers on O’Reilly Media — courses on Python, Pandas, and Dask — and spent 15 years building production systems, from running 200,000 Bayesian models on AWS Batch at Ampersand to leading engineering as CTO at .txt and VP of Engineering at Normal Computing.
I built this course because I kept seeing the same thing: experienced engineers adopting AI tools but prompting them the same way every time, then wondering why the output needed so much editing. These ten patterns are the ones that made the biggest difference in my own work — simple, composable, and transferable to any model or tool you use today.
One price, full course
- 10 prompt engineering patterns — one per day for 10 days
- Runnable Anthropic Python SDK code for every technique
- Real before/after API output examples showing each pattern's impact
- Exercises designed to take 15 minutes using your own code
- Lifetime access to all course emails
Questions
What if I already use the Anthropic API?
Good — this course is built for you. Most developers we talk to already call the API but rely on trial and error to get the output right. These ten patterns give you systematic control: structured output, chain-of-thought reasoning, multi-pass refinement, and evals to measure whether your prompts actually work. The difference shows up immediately.
Do I need an Anthropic API key?
Yes. The code examples use the Anthropic Python SDK and Claude Sonnet. You will need an API key and the anthropic package installed. The patterns themselves apply to any LLM API, but the examples are written for Anthropic.
What languages are the code examples in?
Python using the Anthropic SDK. The patterns themselves apply to any language, and the examples are readable even if Python is not your primary stack.
How long does each day take to complete?
About fifteen minutes. Each one is designed to fit into a morning routine: read the pattern, copy the template, try the exercise, and get back to work.
What if I fall behind?
You have lifetime access to every email. The daily delivery is there to keep you on pace, but you can go through the course at whatever speed works for you.
What is the refund policy?
You can get a 100% refund any time before you receive the Day 3 email. Just reply to any course email or reach out directly and I will process it immediately, no questions asked.