Your agent remembers across turns now. That unlocks the next move — you stop running one agent on one input and start running typed agents across lists, side by side, sharing state.
So this week is the batch version of what we built in Weeks 1 and 2 — typed calls over lists, and two agents talking to each other through a shared context?
That is the shape. Five patterns: batch triage with Literal counts, typed task extraction with list[str], Pydantic event extraction, two agents that share a memory string, and a morning routine that threads compressed context through the whole pipeline.
So by Friday I can run an agent over a week's worth of inputs and have them talk to each other through memory?
By Friday your agents coordinate — they share a context, read each other's output, and compose into a multi-agent routine. Same Agent(model), bigger choreography.
triage_and_count: batch classify with Literal and count per labelauto_create_tasks: extract a typed list[str] of todos from proseget_upcoming_events: Pydantic list of Event(date, summary) extractionsorchestrate_calendar_and_tasks: two agents sharing one compressed contextmorning_orchestration: chain everything into a memory-aware routineGoal: by Friday your agents coordinate — typed batches, shared state, clean composition.
Create a free account to get started. Paid plans unlock all tracks.
Your agent remembers across turns now. That unlocks the next move — you stop running one agent on one input and start running typed agents across lists, side by side, sharing state.
So this week is the batch version of what we built in Weeks 1 and 2 — typed calls over lists, and two agents talking to each other through a shared context?
That is the shape. Five patterns: batch triage with Literal counts, typed task extraction with list[str], Pydantic event extraction, two agents that share a memory string, and a morning routine that threads compressed context through the whole pipeline.
So by Friday I can run an agent over a week's worth of inputs and have them talk to each other through memory?
By Friday your agents coordinate — they share a context, read each other's output, and compose into a multi-agent routine. Same Agent(model), bigger choreography.
triage_and_count: batch classify with Literal and count per labelauto_create_tasks: extract a typed list[str] of todos from proseget_upcoming_events: Pydantic list of Event(date, summary) extractionsorchestrate_calendar_and_tasks: two agents sharing one compressed contextmorning_orchestration: chain everything into a memory-aware routineGoal: by Friday your agents coordinate — typed batches, shared state, clean composition.