AI Summit and CAT-MIP Initiative Cameron spoke at an AI summit hosted by N-able, a company formerly part of SolarWinds. His talk focused on the infrastructure gaps that Managed Services Providers (MSPs) face when handling AI. Consortium for AI Terminology (CAT-MIP): https://cat-mip.org/N-able (Host Company): https://www.n-able.com/AI Market and Stock News Google's stock is performing exceptionally well, and Gemini reportedly has 600-650 million daily active users. A new deal between Microsoft and OpenAI restructures OpenAI into a public benefit corporation, with Microsoft holding a 27% stake. As part of this, OpenAI pre-committed to purchasing $250 billion in cloud compute from Azure.Gemini User Statistics (Note: Research indicates 35M daily users and 450M monthly users): https://vertu.com/lifestyle/gemini-vs-chatgpt-user-numbers-comparison-features-2025-trends/The MCP Server Deployment Challenge Tom's project involves building an MCP (Model Context Protocol) server for massive.com, the new name for market data provider Polygon. Massive (formerly Polygon): https://massive.com/MCP Central Project Cameron's new project, mcpcentral.io, aims to solve this deployment problem. It builds the necessary infrastructure for the agent layer, making agent deployment and management easy for technical users and MSPs. The platform is designed for multi-tenancy, allowing MSPs to run services for clients without deep DevOps expertise.MCP Central Server Example: https://skywork.ai/skypage/en/ai-engineer-langchain-hub/1981168855956770816LangChain Agent Builder Demo A demo of LangChain's new Agent Builder, currently in private beta, shows how it creates agents from natural language. The tool includes pre-built triggers for services like Gmail, Slack, and Linear. An example email assistant agent successfully analyzed an inbox, categorized emails, and created a draft reply directly in a Gmail drafts folder.LangChain Agent Builder Documentation: https://docs.langchain.com/langsmith/agent-builderClaude Skills Explained and Demoed Anthropic's "Claude Skills" feature has been available for about a month. Skills are defined in skill.md files that provide instructions and can reference scripts, which then run in a secure, sandboxed cloud container. This method saves context and allows for repeatable, complex workflows. A demo of the "MCP builder" skill shows it successfully scaffolding a new, production-ready MCP server for the Massive/Polygon options trading API.Explanation of Claude Skills: https://simonwillison.net/2025/Oct/16/claude-skills/Getting Started with Claude Skills: https://neon.com/blog/getting-started-with-claude-skillsChapters00:00:00 Introduction & Weekly AI News00:01:33 Minimax M2 Model Analysis00:13:15 DGX Hardware Update00:14:19 AI Summit & Managed Services Providers (MSPs)00:16:24 CATMIP: A Terminology Standard for AI in IT00:20:53 Big Tech News: Google, Gemini, and Market Growth00:30:25 Demo: Options Trading MCP Server (Polygon/Massive)00:35:21 The Problem: How to Host and Share MCP Servers00:37:47 MCP Central00:44:20 ChatGPT Connectors & Developer Mode00:49:48 Grok Code Fast Model Update00:50:47 LangChain V1 Updates & LangSmith00:52:13 Demo: LangChain Agent Builder00:57:43 Demo: Email Assistant (Agent Builder)01:12:12 Agent Builder: Multi-Tenancy & Security01:13:43 LangSmith: Annotation, Datasets & Experiments01:18:54 Demo: LangSmith Agent Insights01:23:56 Introduction to Claude Skills01:25:12 Demo: Algorithmic Art Skill01:30:38 Claude Skills GitHub Repo01:33:25 Demo: Canvas Design Skill (Podcast Poster)01:37:41 Claude Plugin Marketplace01:40:00 Claude Skill Creator & MCP Builder01:45:26 Demo: Using the MCP Builder Skill01:50:43 Integrating LangSmith with Claude Code01:55:38 Quick Tip: Remote DGX Access with Cloudflare Tunnels01:56:15 Final Demo: Options Trading MCP Server Build Complete01:58:01 Wrap-up

Takeaways1- Tom discusses his recent trip to Texas and the importance of brisket.2 - Cameron shares his experience with the DGX and running DeepSeek OCR.3 - Tom highlights the funding announcement for LangChain and its implications.4 - Cameron expresses excitement about the new features in Google Gemini.5 - Tom discusses the performance of different AI trading algorithms.6 - Cameron shares insights on the OCR performance comparison between DeepSeek and Tesseract.7 - Tom explains the significance of the DGX in running large models.8 - Cameron discusses the potential for continuous fine-tuning in AI models.9 - Tom shares his thoughts on OpenAI's new browser and its market impact. 10 -Cameron and Tom explore new AI tools and their applications.

Takeaways1- Tom discusses his recent trip to Texas and the importance of brisket.2 - Cameron shares his experience with the DGX and running DeepSeek OCR.3 - Tom highlights the funding announcement for LangChain and its implications.4 - Cameron expresses excitement about the new features in Google Gemini.5 - Tom discusses the performance of different AI trading algorithms.6 - Cameron shares insights on the OCR performance comparison between DeepSeek and Tesseract.7 - Tom explains the significance of the DGX in running large models.8 - Cameron discusses the potential for continuous fine-tuning in AI models.9 - Tom shares his thoughts on OpenAI's new browser and its market impact. 10 -Cameron and Tom explore new AI tools and their applications.DescriptionSummary

In this conversation, Cameron and Tom discuss various developments in the AI landscape, including funding announcements, new tools like Google Gemini, and the performance of AI models in tasks such as OCR and crypto trading. They share personal updates, insights on the DGX system, and explore the implications of recent innovations in AI technology. The discussion also includes hands-on demos of new applications and a look into the future of AI workflows and automation.

Sound Bites

• "I want to have brisket."

Key Points

• Why “computer use” agents matter for real form-filling and CRUD tasks

• Live look at Gemini 2.5 Computer Use, Browserbase/Playwright, and WebVoyager-style tasks

• Operator vs Claude/Gemini computer use: accuracy, speed, and safety guardrails (CAPTCHAs, consent, impersonation)

• Where computer use fits vs MCP tools, local OS access, and classic scraping

• Veo 3.1 API update: reference images and start/end frames for consistent video

• Claude Code Plugins & community marketplaces (sub-agents, tools, slash commands)

• GitHub “Spec Kit” and spec-driven workflows for coding at scale

• Cerebras inference speed vs quality tradeoffs; why speed sometimes beats depth

• Local rigs and training: DGX Spark use cases and limits

• Karpathy’s NanoChat: small-scale train-your-own chat model and cost envelope

• “Agent Universe” demo: NAICS-led industry mapping → value flows → agent blueprints

• Architecture questions: vertical vs horizontal agents, data layer, tool connectors (HubSpot, Procore, Google Workspace)

A focused walkthrough of today’s agentic stack in practice. The episode tests Gemini 2.5 “computer use” for real browser tasks, compares it with Operator and Claude, and breaks down safety guardrails and why screenshot-loop agents remain slow. It covers where computer use fits alongside MCP and OS tools, then shifts to Veo 3.1’s new API features for reference-guided video. On the coding side, it explores Claude Code Plugins and community marketplaces, plus GitHub’s Spec Kit for spec-driven development on large codebases. The discussion touches Cerebras for ultra-fast inference, DGX Spark for local experiments, and Karpathy’s NanoChat for training compact chat models. It closes with the “Agent Universe” demo: mapping industries via NAICS, generating value-flow diagrams, and turning stages into deployable agent roles, with open questions on architecture, tools, and handoff into real systems.

09:51 Exploring Claude's Capabilities and User Experience19:28 Real-Time Code Generation: Claude Imagine Demo41:02 The Evolution of User Interfaces for AI Agents48:29 Challenges in Research and Data Analysis with AI58:22 Sora 2 and Luma.labs Live demo - Ray Reasoning Video Model01:04:21 Copyright Challenges in AI Content Creation01:12:59 Interactive Video Creation and User Control01:23:45 Building an AI-First Strategy and The GDPEVal Data Set01:37:38 Real-World Applications of AI01:51:42 Navigating AI's Impact on Employment

In this episode, Tom Spencer and Cameron Rohn discuss various updates in the tech world, including the challenges of AI podcasting, the implications of H1B visa changes, and the competitive landscape of AI models from Alibaba. They also delve into the limitations of Figma in web design and explore Cloudflare's new Vibe SDK. Cameron Rohn and Tom Spencer delve into the capabilities of Cloudflare's Vibe SDK and its integration with various AI tools. They discuss the potential of creating scalable applications using Cloudflare Workers, the ease of deploying custom apps, and the importance of integrating design tools like Figma into the development process. The conversation also touches on the features of Cloudflare's AI Gateway, including data loss prevention and real-time agent kits, as well as the advancements in voice cloning technology with 11 Labs for podcasting.Chapters00:00 Introduction and Overview of Topics03:02 Visa Policies and Talent Acquisition05:26 China's K-Visa and Global Talent Competition07:56 Alibaba's New AI Models and Innovations10:55 Deep Dive into Alibaba's Omni Model13:24 Comparative Analysis of AI Models16:09 Future of AI and Global Competition33:19 Exploring Alibaba's Innovations in AI Video Generation35:32 Diving into Figma and the Rise of NanoBanana42:35 Challenges with Figma's AI Integration and Design Workflow52:45 Utilizing MCP Servers for Enhanced Design Collaboration01:00:15 Transitioning to Cloudflare's Vibe SDK and Backend Infrastructure01:01:28 Exploring Cloudflare's Ecosystem01:07:41 Understanding Durable Objects and Stateful Applications01:13:52 Integrating AI Models and Customization01:20:19 Cloudflare AI Gateway and Observability01:25:32 Guard Rails and Data Loss Prevention in AI01:26:03 Innovations in Data Loss Prevention and Real-Time Evaluation01:28:26 Exploring Real-Time Agent Kits and Cloudflare's Infrastructure01:32:09 Voice Cloning and Podcasting with 11 Labs01:39:16 Creating Interactive AI Agents for Business Applications01:45:58 Future of AI in Communication and Collaboration

00:00 Introduction and Personal Updates

04:33 Transformative Use of ChatGPT in Family Projects

07:20 Discussion on AI Economic Index Reports

17:46 The Rise of ChatGPT: Analyzing Growth Metrics

22:03 Exploring User Interaction: Conversations with AI

27:45 Market Position: ChatGPT vs. Competitors

36:35 The Future of Development: Integration and Collaboration

42:14 DeepMind's Innovations: Exploring New Frontiers

43:32 The Virtual Agent Economy

49:21 Google's AP2 Protocol and Agentic Commerce

01:02:40 Exploring World Models and Spatial AI

01:11:32 Exploring AI and Hardware Leapfrogging

01:16:03 Robotics and the Built Environment

01:20:13 Advancements in Robotics and AI Models

01:25:23 The Future of Home Robotics

01:30:36 Privacy Concerns with Autonomous Robots

01:36:20 Creative Applications of AI in Music Production

01:41:46 Creative AI in Music Production

01:49:57 Challenges in Music Production and AI Integration

01:56:17 Exploring AI's Potential in Music Composition

02:09:19 The Future of AI in Music and Sound Design

Also the best ai generated takeaways from this episode is:

• Tom is experiencing a midlife crisis and exploring new interests.

Chapters

00:00 Welcome Back and Project Updates

03:30 Insights from the All-In Summit

06:21 Claude's Performance Issues and Alternatives

09:38 Exploring New Tools and Interfaces

12:38 Innovations in AI and Image Generation

15:34 The Future of Content Creation and AI

18:32 AI's Impact on Music Production

21:22 Reflections on AI in Creative Industries

24:35 Concluding Thoughts from the All-In Summit

34:03 Exploring Individuality and Ambition

35:32 Technological Innovations in Warfare

37:33 Cultural Nuances in AI Development

38:46 Emerging AI Technologies and Their Applications

40:39 AI Models and Future Predictions

45:45 OpenAI's Strategic Moves and Market Positioning

53:29 OpenAI's Acquisitions and Future Directions

56:38 Oracle's Role in AI Development

01:01:22 The Future of AI in Creative Production

Lots of demos, deep discussions and ai injections. Check the vault for the show notes, links and more at buildaipod.com

Chapters00:00 Introduction and Personal Updates00:46 Diving Deeper into GPT-5 Features and User Reactions01:48 Comparing GPT-5 with Previous Models04:08 User Feedback and Market Reactions06:37 Final Thoughts on GPT-5 and Future Directions10:57 Navigating Coding Agents and Abstraction Layers13:40 The Evolution of GPT Models and Their Applications18:07 Inference Compute and the Future of AI Investment19:57 Exploring Grok Desktop and Its Features27:25 Interoperability and Customization in AI Tools35:22 Introducing the Vault: A Centralized Resource for AI Applications42:02 Exploring Deep Agents and UI45:28 Running Deep Agents Locally47:19 Enhancing User Experience with UI Components50:13 Integrating Claude Code with Deep Agents56:37 Transitioning to Social Media Agents57:43 Complexity of Social Media Agent Architecture01:04:10 Monitoring and Managing Social Media Agents01:09:25 Streamlining Content Generation Processes01:12:54 Optimizing Code Repositories for AI Agents01:13:58 Navigating Corporate Culture Shifts01:16:49 Exploring Options Trading with AI01:22:47 Integrating AI with Financial Data01:28:42 Full Loop Development in AI Projectssign up to the vault to get all the details about the show - buildaipod.com

We recorded August 7th, right before ChatGPT launched.

We dove into GPT open source, OpenCode, Ollama Turbo, and deep agent setups.

I wanted to see LangChain’s open suite and test agent environments.

OpenCode stood out for its flexibility — multiple model providers, easy local setup, works with Ollama Turbo for $20/month.

LM Studio runs similarly.

I’m considering a high-spec NVIDIA rig and DGX Spark for local inference.

GPT-OSS is cheap, fast, and excellent for coding and tool-calling, but weaker on general knowledge.

Running it locally means more setup work but more control.

Hybrid local-plus-cloud routing feels inevitable.

We demoed OpenAgent Platform — fast, multi-provider agents without writing code.

Then explored LangChain SWE — an open-source, multi-threaded coding agent with planner/programmer loops, GitHub integration, Daytona sandboxes, and detailed token-cost tracking.

We looked at Vercel’s v0 API for quick generative UI, and the potential to run it privately for internal teams.

I closed with Google’s upcoming AI-mode ads and Societies.io — a virtual audience simulation tool for testing and optimizing content before publishing.

Chapters

00:00 Introduction to ChatGPT Launch and Demos

01:40 Exploring Open Code and LangChain

04:37 Local Inference and Olamma Integration

07:25 Cloud Acceleration with Turbo Service

10:11 Open Source Model Benchmarks and Feedback

Notes to come.