Latest / Logosofia no Dia a Dia
How to Build Disposable AI Agents at Runtime
What if you could click a button and instantly get an AI assistant that knows everything about a single item in your home inventory—without pre-building hundreds of agents? This episode explores the engineering behind dynamically generated, disposable AI agents. We break down the architecture using the OpenAI Assistants API, LangChain, and simpler approaches, weighing tradeoffs between build complexity and runtime cost. The conversation covers retrieval optimization, context window management, and why good UI design can eliminate hard AI problems. Plus: why modern user manuals are terrible…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.