Ask HN: Agentic search vs. RAG – what's your production experience?
There's been a clear shift from 2023's you need RAG to 2026's agentic search outperforms RAG.
We made the switch at Cosmico when building AI-native applications with Claude Code. Agentic search gives us better accuracy but at higher cost per query.
For those running production systems:
1. What prompted your switch (or decision to stick with RAG)? 2. What broke during the transition? 3. What's your hybrid approach (if any)?
Specifically curious about code search vs. document retrieval use cases, and how you handle the latency/cost trade-offs.
Our context: building custom software with AI agents in weeks, not months. Accuracy matters more than speed for our workflow, but that's not universal.
What's working (or not working) for you? In practice I’ve found agentic search does improve accuracy on harder queries, but the cost climbs pretty quickly once you move away from a single pass. More steps, more chances to call the model again, more room for retries. It compounds faster than you expect. Plain RAG is easier to keep predictable if your retrieval is tight. But once chunk count grows and you start pushing larger contexts, quality drops and you end up compensating somewhere else. What’s worked best for me is a hybrid. Cheap default path first, then only route ambiguous or high-value queries into the heavier agent flow. Also I think cost per successful task is more meaningful than cost per query. The expensive path only makes sense if it actually reduces retries or downstream fixes.