Weeks 1 and 2 gave you the skeleton — search, then reason. What goes wrong when you scale the search half — same queries twice, duplicate domains, too many results to read?
You'd waste API calls, get repeats from the same site, and the agent would slow down processing noise instead of signal?
Exactly. This week sharpens retrieval. Day 17 uses agent-computed semantic scores to pick the most relevant snippet, not the first one. Day 18 has the agent rank results. Day 19 caches search results so repeat queries stay fast. Day 20 dedupes by domain to avoid five links to the same site. Day 21 compares two queries side by side.
So by Friday the retrieval layer is dense — semantic, cached, deduplicated, comparable?
Exactly. Dense retrieval is what lets Week 4's capstone agent answer research questions in seconds. Week 3 is the upgrade from "raw search" to "intelligent retrieval."
Goal: retrieval that is semantic, cached, and deduplicated — not just raw search.
Create a free account to get started. Paid plans unlock all tracks.
Weeks 1 and 2 gave you the skeleton — search, then reason. What goes wrong when you scale the search half — same queries twice, duplicate domains, too many results to read?
You'd waste API calls, get repeats from the same site, and the agent would slow down processing noise instead of signal?
Exactly. This week sharpens retrieval. Day 17 uses agent-computed semantic scores to pick the most relevant snippet, not the first one. Day 18 has the agent rank results. Day 19 caches search results so repeat queries stay fast. Day 20 dedupes by domain to avoid five links to the same site. Day 21 compares two queries side by side.
So by Friday the retrieval layer is dense — semantic, cached, deduplicated, comparable?
Exactly. Dense retrieval is what lets Week 4's capstone agent answer research questions in seconds. Week 3 is the upgrade from "raw search" to "intelligent retrieval."
Goal: retrieval that is semantic, cached, and deduplicated — not just raw search.