Everyone Is Building AI Agents Wrong. Here's What Actually Works

Last week at MongoDB.local Bangalore I sat down with Sejal Khanna, Senior Developer Advocate at MongoDB, on The Ravit Show. Loved hosting her!!!!Sejal spends her days helping developers move from AI curiosity to actually shipping. Workshops, hands-on sessions, product storytelling, community. She works at the layer where hype meets reality, which makes her one of the most useful voices to hear from right now.Here is what we got into.We started with what she is building and teaching right now. The volume of AI content out there is enormous. What she keeps finding herself filling in when she is working with builders directly is the gap between watching a tutorial and actually getting an agent to behave in production.We walked through the announcements from the day. voyage-context-4 GA. Hybrid Search GA. Native Reranking in public preview. Search and Vector Search shipping in Community Edition and Enterprise Advanced. Sejal broke down which of these she is most excited for builders to actually get their hands on, and why the Community Edition move might quietly be the biggest one for developers in India.We spent real time on her workshop, The A to Z of Building AI Agents. Reasoning, tools, memory, agent architectures, and then actually building one using MongoDB as memory, a Claude model, and LangGraph for orchestration. She walked me through what she hopes people walk away able to do, which is a lot more concrete than the average AI workshop pitch.#data #ai #mongodb #mongodblocal #theravitshow