Hey, this is Lance. I want to talk about continual learning with agents in particular showing some examples with deep agents. So, Le has a really nice post on this theme of continual learning in token space and it makes the argument that a big gap between AI agents and humans as we know is ability to learn.Humans continually learn and improve over time. Agents knowledge is typically fixed and doesn't have the same adaptive capability. Now, there's different ways to teach AI systems to learn.So one is learni ...

Key Functionality & Benefits - Langchen introduces native provider tools from OpenAI and Entropic, enabling optimized model usage [1] - The new Langchain provider packages simplify agent building by eliminating manual JSON schema handling [2] - Cloud can dynamically discover and load tools on demand, facilitating operations on platforms like Cloudflare [3] - Native provider tools unlock real-world multimodal agentic applications [4] - The MCP tool set and tool search provider tool are more efficient than locally spinning up MCP servers [15] - The new tools API eliminates the need for handcrafting schemas and gluing adapters [17] Cloudflare MCP Agent Example - An agent connects to Cloudflare's MCP servers, granting access to the entire platform via a chat interface [5] - The agent can fetch DNS logs, run queries, and answer questions about a user's Cloudflare account [3] - The agent can access Cloudflare browsers to fetch websites and convert them to markdown or screenshots [7][14] - The MCP tool set tool from the entropic provider package allows defining configuration for every MCP server [10] - The tool search tool defers loading of MCP server tools, saving context window space [11][12]

Deep Agents Overview - Deep Agents is an open-source agent harness incorporating planning, computer access, and sub-agent delegation tools, commonly found in agents like Manis and Cloud Code [1][46] - The harness is designed to be easily adaptable with custom prompts, tools, and sub-agents [2][47] Key Features and Tools - Deep Agents provides built-in tools such as planning, sub-agent delegation, and file system operations [6][7] - The built-in tools enable interaction with the file system, shell command execution, planning via to-dos, and task delegation [8] - Custom tools, instructions, and sub-agents can be added to Deep Agents to tailor it for specific use cases [6][47] Quick Start and Research Application - The Deep Agent quick start repo offers examples for different use cases, starting with research [2][5] - The research quick start includes tools like a search tool (using Tavi search API) and an optional "think" tool for auditing agent trajectory [12][13][14] - Task-specific instructions and sub-agents can be supplied to Deep Agents for any given use case [12] Agent Loop and Middleware - Deep Agents utilizes Langraph for orchestrating the agent loop, which involves the language model (LLM) calling tools in a loop [29] - Middleware serves as hooks within the agent loop, allowing for actions like summarization when context exceeds 170,000 tokens [30][32] - Middleware can provide tools to the agent, such as file system middleware, and perform actions like summarization and prompt caching [31][34] File System and State Management - By default, Deep Agents writes to an internal in-memory state object, but it supports different backends like a sandbox or local file system [37][38] - File reading and writing operations occur within the Langraph state object, enabling easy retrieval into the LLM's context window [40] Deployment and Visualization - Deep Agents can be run in a Jupyter notebook for interactive inspection or deployed as an application using Langraph [10][44] - A UI can be connected to the local Langraph server for visualizing generated files and agent interactions [3][45]

Check this out. I just asked an agent to answer one of the world's greatest debates. Is Messi or Ronaldo the greatest soccer player of all time.This isn't an easy question to answer, and it definitely requires a good amount of research. The agent automatically spawned two parallel sub agents to look into each of their achievements. This meant searching the web over a dozen times, compiling a comprehensive report with cited sources.To be extra thorough, the agent then critiqued its own report and plugged any ...

Core Insights - Benchmark has experienced significant leadership changes, losing three General Partners (GPs) in the past two years, while also adding a new GP, Everett Randle, who has a strong background in AI investments [1][4][5] Group 1: Leadership Changes - Victor Lazarte, a key figure in Benchmark's investment in HeyGen, has left to start his own VC firm, marking a notable shift in the firm's leadership [1] - Everett Randle, previously a partner at KP, has joined Benchmark, bringing a wealth of experience in AI investments [1][2] - The firm has emphasized a flat organizational structure, valuing equal power and responsibility among partners [5] Group 2: Investment Performance - Benchmark's investments in AI have yielded impressive results, with many projects experiencing rapid revenue growth or securing multiple funding rounds [5] - Recent notable investments include a $25 million Series A round for an AI document product and a $105 million Series B round for Cursor, an AI programming tool [6] - The firm has also led significant funding rounds for AI-related projects, such as $85 million for Exa, which aims to create AI-centric search solutions [7] Group 3: Notable Projects and Valuations - Benchmark's early investments in AI projects like Cerebras have resulted in substantial valuations, with Cerebras recently achieving an $8.1 billion valuation after a $1.1 billion Series G round [8] - The firm has invested in various AI sectors, including AI coding, bug detection, and sales tax compliance automation, showcasing a diverse portfolio [6][8] - Benchmark's involvement in projects like HeyGen and Manus highlights its influence in the Chinese market, with HeyGen achieving $100 million in annual recurring revenue (ARR) [8]

Product Overview - xpander is a production-ready backend solution for AI Agents, managing memory, tools, states, version control, and guardrails [1] - The solution is designed to be plug-and-play and fully self-hostable [1] - It is compatible with various frameworks such as CrewAI, Agno, and Langchain [1] Technology and Implementation - xpander addresses the need for a functional backend in the development and deployment of AI Agents [1]

Langsmith Platform Overview - Langsmith is an observability and evaluation platform for AI applications, focusing on tracing application behavior [1] - The platform uses tracing projects to collect logs associated with applications, with each project corresponding to an application [2] - Langsmith is framework agnostic, designed to monitor AI applications regardless of the underlying build [5] Tracing and Monitoring AI Applications - Tracing is enabled by importing environment variables, including Langmouth tracing, Langmith endpoint, and API key [6] - The traceable decorator is added to functions to enable tracing within the application [8] - Langsmith provides a detailed breakdown of each step within the application, known as the run tree, showing inputs, outputs, and telemetry [12][14] - Telemetry includes token cost and latency of each step, visualized through a waterfall view to identify latency sources [14][15] Integration with Langchain and Langraph - Langchain and Langraph, Langchain's open-source libraries, work out of the box with Langsmith, simplifying tracing setup [17] - When using Langraph or Langchain, the traceable decorator is not required, streamlining the tracing process [17]

[Music] My name is John Gutsinger. Uh I work for Cisco. I'm a principal engineer and I work in the technical assistance center or TAC for short.Uh really I'm focused on AI engineering, agentic engineering in the face of customer support. We've been doing a IML for you know a couple years now maybe five or six years. really it started with trying to figure out how do we handle these mass scale issues type problems right where uh some trending issues going to pop up we know we're going to have tens of thousan ...