I chatted with John Kutay recently about Strands Agents and what our "model-driven approach" to building agents really means, such as swapping out RAG and prompt pipelines for Retrieval as Tool or "RAT" (coined by my colleague Anoop Deoras!). Let the agent's model ask for what it needs to know, instead of proactively stuffing the context window with unnecessary information Strands Agents: https://strandsagents.com/ Podcast links: Striim https://lnkd.in/gWhdzktk YouTube https://lnkd.in/g5GSYCGr Spotify https://lnkd.in/gHS8rdz6
Loved the episode so much, had to make a post about it!
Thanks for sharing, Clare
Ammar AA Jaai tau phir delet Karen ge
Founder & CEO at Chkk
3mo+1. I like the ‘Retrieval As a Tool’ vs ‘Context Stuffing’ framing. Our distantly-related, early experience using AI Coding Agents for Operational tasks like creating PRs for IaC confirms that 1/ Minimal 2/ Relevant 3/ Up-to-date context yields the highest quality results.