Letta AI with Sarah Wooders - Weaviate Podcast #117! | Weaviate Podcast

In this episode of the Weaviate Podcast, the host interviews Sarah Wooders, co-founder and CTO of Letta, about stateful AI agents and the evolution of MemGPT. They discuss context compilation, structuring the context window, and self-editing prompts. Wooders explains Letta's approach to memory management, the agent development environment (ADE), and the use of tools for external data interaction. The conversation also covers the challenges and benefits of using context windows for memory and state, the role of databases in AI agents, and the future directions of AI, particularly in deriving insights from data.

Outlines

Part 1: Introduction to Stateful AI Agents

Part 2: Context Compilation and Tooling

Part 3: Multi-Agent Systems and Model Routing

Part 4: Future and Conclusion

Sign in to continue reading, translating and more.

Open full episode in Podwise

Letta AI with Sarah Wooders - Weaviate Podcast #117!

Weaviate Podcast

Part 1: Introduction to Stateful AI Agents

Introduction to Letta and Stateful AI Agents

The Evolution of Context Management in AI Agents

Structuring the Context Window for AI Agents

Self-Editing Prompts and Memory Management

Context Memory vs. Fine-Tuning

Part 2: Context Compilation and Tooling

Context Compilation and Tool Use

The Agent Development Environment (ADE)

ADE vs. LLM Observability Tools

Evals and Agent Development

Databases for AI Agents

External Tools and Agentic Architecture

Part 3: Multi-Agent Systems and Model Routing

Multi-Agent Systems and Shared Memory

Model Routers and Context Compilation

Reasoning Models and Context Window Structure

Agent Development Environments and Core Abstractions

Executing Tools on the Server

Tool Sandboxes and Future Directions

Part 4: Future and Conclusion

AI in Code and Conclusion

Letta AI with Sarah Wooders - Weaviate Podcast #117!

Weaviate Podcast

Part 1: Introduction to Stateful AI Agents

00:00Introduction to Letta and Stateful AI Agents

Introduction to Letta and Stateful AI Agents

02:36The Evolution of Context Management in AI Agents

The Evolution of Context Management in AI Agents

05:35Structuring the Context Window for AI Agents

Structuring the Context Window for AI Agents

09:55Self-Editing Prompts and Memory Management

Self-Editing Prompts and Memory Management

12:21Context Memory vs. Fine-Tuning

Context Memory vs. Fine-Tuning

Part 2: Context Compilation and Tooling

16:21Context Compilation and Tool Use

Context Compilation and Tool Use

19:15The Agent Development Environment (ADE)

The Agent Development Environment (ADE)

22:13ADE vs. LLM Observability Tools

ADE vs. LLM Observability Tools

25:18Evals and Agent Development

Evals and Agent Development

28:35Databases for AI Agents

Databases for AI Agents

32:34External Tools and Agentic Architecture

External Tools and Agentic Architecture

Part 3: Multi-Agent Systems and Model Routing

35:55Multi-Agent Systems and Shared Memory

Multi-Agent Systems and Shared Memory

38:54Model Routers and Context Compilation

Model Routers and Context Compilation

42:21Reasoning Models and Context Window Structure

Reasoning Models and Context Window Structure

45:51Agent Development Environments and Core Abstractions

Agent Development Environments and Core Abstractions

48:15Executing Tools on the Server

Executing Tools on the Server

51:25Tool Sandboxes and Future Directions

Tool Sandboxes and Future Directions

Part 4: Future and Conclusion

55:03AI in Code and Conclusion

AI in Code and Conclusion