Grant Miller outlines five evolving patterns for integrating AI agents with tools, starting from simple direct API connections to complex, secure token-based architectures. The discussion highlights the progression of these methods, emphasizing how authentication, user delegation, abstraction layers like MCP, and secure credential management using vaults improve security, observability, and scalability in agentic systems.
Aaron Baughman and Martin Keen present a unified framework using "periodic tables" to integrate AI and Data Science. They illustrate how elements like pipelines, embeddings, and RAG combine to build real-world AI applications, using a detailed document Q&A system example. The discussion emphasizes the critical interdependence of data science in grounding AI models and ensuring continuous improvement through an innovative feedback loop.
Prukalpa Sankar argues that while AI models are increasingly intelligent, their practical business utility is limited by a lack of contextual intelligence. She proposes a "context layer" as a crucial solution, akin to a "GitHub for context," to centralize and manage business knowledge, expertise, and norms. This system aims to provide AI agents with the shared, versioned, and continuously learning context necessary to overcome the challenges of isolated systems and context sprawl, ultimately differentiating companies in an AI-driven world.
Justin Schroeder argues for a paradigm shift in AI agent development from monolithic, context-inflated agents (inheritance) to modular, domain-specific agents (DSAs) that operate through composition. He explains how DSAs offer superior token efficiency, cost savings with smaller models, enhanced security through capability limits, and better scalability, predicting their widespread adoption by 2027 as a solution to rising AI costs and the need for practical, customer-facing AI.
Dr. Lech Kalinowski presents Vault, a unique AI-native handheld terminal built on an ESP32-S3, featuring a dual-display (OLED for live interaction, e-paper for static content) and powered by a single battery. It operates in four modes—shell, assist, control, and an LLM-native RPG—and controls autonomous OpenClaw agents against a local, self-hosted 120B parameter GPT model served by NVIDIA TensorRT-LLM. The talk delves into the AI-native architecture, why inference remains on the backend, real-time agent thinking surfacing on e-paper, engineering challenges, and LLM-native game design, highlighting the potential for calm, text-first AI interaction.
Dan Biderman and Jessy Lin of Engram introduce their "always training" paradigm, focusing on baking a team's knowledge directly into a model's weights to achieve true memory and continual learning. This contrarian approach, which they call a "RAG killer" for specific use cases, promises up to 100x token savings and superior performance by internalizing context rather than relying on ever-larger context windows or external retrieval, envisioning a future where everyone has their own continually learning, personalized AI model.