Weeks 1 and 2 built one agent per function. What happens when a task needs two agents, or the same agent run across a hundred inputs?
You'd want multiple agents sharing a function — one for tone, one for classification — or you'd need to batch through a list instead of calling the agent once?
Both. Week 3 is about composition. Day 17 is two agents with different system_prompt values returning a dict. Day 18 is a pipeline that keeps every intermediate output, not just the final. Days 19 through 21 are batch patterns — list comprehensions over prompts, min(..., key=len), and chained transforms on agent output.
So agents become building blocks — specialized, reusable, composable, just like regular Python functions?
Exactly. By Friday you can look at a list of inputs and build the right batch pattern without thinking. Same run_sync, bigger shapes.
system_prompt returning one dictmin(outputs, key=len) for the shortest response.output.split() + len() in one comprehensionGoal: you compose agents and batches as fluently as you call them.
Create a free account to get started. Paid plans unlock all tracks.
Weeks 1 and 2 built one agent per function. What happens when a task needs two agents, or the same agent run across a hundred inputs?
You'd want multiple agents sharing a function — one for tone, one for classification — or you'd need to batch through a list instead of calling the agent once?
Both. Week 3 is about composition. Day 17 is two agents with different system_prompt values returning a dict. Day 18 is a pipeline that keeps every intermediate output, not just the final. Days 19 through 21 are batch patterns — list comprehensions over prompts, min(..., key=len), and chained transforms on agent output.
So agents become building blocks — specialized, reusable, composable, just like regular Python functions?
Exactly. By Friday you can look at a list of inputs and build the right batch pattern without thinking. Same run_sync, bigger shapes.
system_prompt returning one dictmin(outputs, key=len) for the shortest response.output.split() + len() in one comprehensionGoal: you compose agents and batches as fluently as you call them.