You can automate your inbox, your Calendar, your Sheets. Now I want to ask: what still requires a human? What work do you do that no script can touch yet?
Reading meeting transcripts. Extracting the actual commitments from a two-hour call. Writing the first draft of a performance review. That stuff requires judgment.
Until now. AI agents can summarise a 90-minute transcript, extract structured action items — owner, task, due date — and draft the email you'd send after the meeting. Before we start building, rate yourself on six skills. Be honest — a 1 is a perfectly valid answer if you've never called an AI API.
I've used ChatGPT. Does that count?
Using ChatGPT and building an AI agent in Python are different things — one is a conversation, one is a function you call in a pipeline. By Day 30 you'll know exactly where the line is. Rate honestly. Day 30 answers show the delta.
Thirty days. One real project: an AI meeting summariser. By Day 28 you'll have a working agent that takes a meeting transcript, extracts ActionItem(owner, task, due_date) items into a structured list, and returns a formatted Monday-send email body with highlights, action items, and open questions.
| Week | Focus | Professional framing |
|---|---|---|
| 1 | Basic agents: run, output, summarise, classify, chain | "Write a summary of this exec update in two sentences" |
| 2 | Structured output: Pydantic models, Literal, list extraction | "Extract every commitment from this strategy memo as a typed dict" |
| 3 | Pipelines and batch: multi-agent, batch classify, word counts | "Process 200 performance reviews into themes — in an afternoon" |
| 4 | Search, tools, and capstone | "Build the AI meeting summariser your team actually sends on Mondays" |
Every lesson is a short Socratic conversation plus one coding challenge that runs against a real AI model.
Create a free account to get started. Paid plans unlock all tracks.
You can automate your inbox, your Calendar, your Sheets. Now I want to ask: what still requires a human? What work do you do that no script can touch yet?
Reading meeting transcripts. Extracting the actual commitments from a two-hour call. Writing the first draft of a performance review. That stuff requires judgment.
Until now. AI agents can summarise a 90-minute transcript, extract structured action items — owner, task, due date — and draft the email you'd send after the meeting. Before we start building, rate yourself on six skills. Be honest — a 1 is a perfectly valid answer if you've never called an AI API.
I've used ChatGPT. Does that count?
Using ChatGPT and building an AI agent in Python are different things — one is a conversation, one is a function you call in a pipeline. By Day 30 you'll know exactly where the line is. Rate honestly. Day 30 answers show the delta.
Thirty days. One real project: an AI meeting summariser. By Day 28 you'll have a working agent that takes a meeting transcript, extracts ActionItem(owner, task, due_date) items into a structured list, and returns a formatted Monday-send email body with highlights, action items, and open questions.
| Week | Focus | Professional framing |
|---|---|---|
| 1 | Basic agents: run, output, summarise, classify, chain | "Write a summary of this exec update in two sentences" |
| 2 | Structured output: Pydantic models, Literal, list extraction | "Extract every commitment from this strategy memo as a typed dict" |
| 3 | Pipelines and batch: multi-agent, batch classify, word counts | "Process 200 performance reviews into themes — in an afternoon" |
| 4 | Search, tools, and capstone | "Build the AI meeting summariser your team actually sends on Mondays" |
Every lesson is a short Socratic conversation plus one coding challenge that runs against a real AI model.