Three weeks of building blocks. Week 1 read search. Week 2 wired an agent onto search. Week 3 sharpened retrieval. What happens when you stack the whole thing into one function?
A research-assistant shape — take a question, search for it, maybe hit multiple angles, then have the agent synthesize a structured answer with sources?
Exactly the arc. Day 24 builds the minimal RAG — retrieve then answer. Day 25 returns a structured Pydantic brief with fields the caller can read directly. Day 26 fans out across multiple queries and merges the retrievals. Day 27 adds a confidence score as a Literal. Day 28 is the full research-assistant capstone — every pattern from the track in one function.
So by Friday I write one function that takes a question and gives back a researched, structured, confidence-scored answer?
Exactly. And Day 30's explainer previews the next step — agents that remember. That's ai-for-makers: multi-turn, stateful, memory-augmented. This week is the last one where every call is stateless.
Brief(summary, sources, keywords) from searchLiteral confidence scoreGoal: an end-to-end retrieval-augmented research function you'd deploy today.
Create a free account to get started. Paid plans unlock all tracks.
Three weeks of building blocks. Week 1 read search. Week 2 wired an agent onto search. Week 3 sharpened retrieval. What happens when you stack the whole thing into one function?
A research-assistant shape — take a question, search for it, maybe hit multiple angles, then have the agent synthesize a structured answer with sources?
Exactly the arc. Day 24 builds the minimal RAG — retrieve then answer. Day 25 returns a structured Pydantic brief with fields the caller can read directly. Day 26 fans out across multiple queries and merges the retrievals. Day 27 adds a confidence score as a Literal. Day 28 is the full research-assistant capstone — every pattern from the track in one function.
So by Friday I write one function that takes a question and gives back a researched, structured, confidence-scored answer?
Exactly. And Day 30's explainer previews the next step — agents that remember. That's ai-for-makers: multi-turn, stateful, memory-augmented. This week is the last one where every call is stateless.
Brief(summary, sources, keywords) from searchLiteral confidence scoreGoal: an end-to-end retrieval-augmented research function you'd deploy today.