This week you write five small functions, each one a full round-trip to a real language model. By Friday you've called an agent five different ways — and you've chained two of them together in one function.
Five agents in one week sounds like a lot. Do they stack on each other or is each day independent?
Each day introduces one new idea: the call itself, then reading the output as a string, then shaping it with a system_prompt, then constraining it to a single word, then wiring two calls in sequence. Day 7 is the only lesson that combines patterns — by then your vocabulary is already in your head.
And every run hits a real model? Not a canned test response?
Every run. You pass a prompt in, the model generates tokens, you read them back as a Python string. That feedback loop is the whole reason Week 1 stays on basic calls — seeing your code talk to a live LLM is the best way to start.
Agent(model).run_sync(prompt) and .output.split()system_prompt that turns any agent into a summarizerGoal: by Friday you know every shape the smallest agent call can take.
Create a free account to get started. Paid plans unlock all tracks.
This week you write five small functions, each one a full round-trip to a real language model. By Friday you've called an agent five different ways — and you've chained two of them together in one function.
Five agents in one week sounds like a lot. Do they stack on each other or is each day independent?
Each day introduces one new idea: the call itself, then reading the output as a string, then shaping it with a system_prompt, then constraining it to a single word, then wiring two calls in sequence. Day 7 is the only lesson that combines patterns — by then your vocabulary is already in your head.
And every run hits a real model? Not a canned test response?
Every run. You pass a prompt in, the model generates tokens, you read them back as a Python string. That feedback loop is the whole reason Week 1 stays on basic calls — seeing your code talk to a live LLM is the best way to start.
Agent(model).run_sync(prompt) and .output.split()system_prompt that turns any agent into a summarizerGoal: by Friday you know every shape the smallest agent call can take.