AI News Today | Julian Goldie Podcast

Hermes Agent v0.14: Biggest Free Update Yet (180x Faster Browser, 1-Line Install, Grok 4.3 Studio + Memory Stack)The script walks through Hermes Agent v0.14, described as its biggest free update, including a one-line install (pip install hermes-agent), faster startup, and 180x faster browser control via a persistent Chrome connection. It shows how to update via the terminal (hermes update) and highlights new capabilities like Grok 4.3 integration with a 1M-token context window via OAuth, built-in X/Twitter search, a local OpenAI-compatible proxy to use OpenAI-style tools without API keys, Microsoft Teams support, computer use on any vision model, live session handoff between models, native Windows beta, new skill packs, and Discord history backfill. The presenter demos an “Agent OS” media studio for image/video/speech generation and Twitter search, explains a multi-layer “Goldie Agentic command stack,” and emphasizes an Obsidian-based memory system (“Infinite Context Engine”) for personalized automation outputs, while promoting an AI Profit Boardroom community and tutorials.00:00 V0.14 Biggest Update00:59 Update and Install01:48 One Line Setup02:47 Grok 4.3 Integration03:49 AI Media Studio Demo07:10 Proxy and X Search09:10 Teams and Faster Browser10:51 Computer Use Anywhere12:12 Handoff and Skills18:23 Memory Context Engine20:46 Command Stack Layers21:34 Recap and Community23:12 Live Q and A Wrap

OpenClaw 5.16 Beta: Groq Login (No API Key) + Silent Telegram Group Mode + Faster WorkflowsOpenClaw 5.16 arrives on GitHub as a pre-release beta with major upgrades: Groq can now be used via a Super Groq subscription login without an API key or extra billing, and the Telegram agent gains a “silent group” mode that listens to all messages to build context but only speaks when prompted. The script also highlights improvements like chaining scheduled tasks with a blocking wait/timeout, faster skill loading via caching, a Chinese (simplified/traditional) setup wizard, more reliable Discord reconnections, fixes for Matrix encrypted rooms, dashboard context-window estimation for local models, and a compaction bug fix to prevent repeated summarization. It promotes an “agentic OS” in the AI Profit Boardroom for managing OpenClaw, Claude, and Hermes with shared memory from one dashboard, and advises caution when testing the beta by backing up first and using the beta update channel with rollback options.00:00 Beta Update Overview01:03 Groq Login No Key02:09 Silent Telegram Groups03:37 Agentic OS Dashboard04:25 Workflow Speed Fixes06:10 Why Shared Memory Wins06:53 Safe Beta Upgrade Steps07:30 Final Takeaways Invite

Hermes Agent + Grok OS Upgrade: Real‑Time X Search, Image/Video Gen & Voice (Step‑by‑Step Setup)The video explains how a new Hermes Agent update combined with Grok OS (xAI/Grok via OAuth) enables real-time X (Twitter) search, image generation, video generation, and text-to-speech inside one AI agent system, often at no added cost if you already have an X subscription. It walks through the setup steps: update Hermes, select the XAI Grok OAuth model via the terminal command (Hermes model) and log in, then enable and configure tools in Hermes tools (including Xsearch, video generation, image generation, and TTS), noting that some features may require reconfiguring providers (e.g., switching video from PHAL to X). The script shows troubleshooting by importing documentation so Hermes understands tool calls, demonstrates generating an X trend roundup, images, a short video, and a voice note, and then integrates these capabilities into an Agent OS “studio” dashboard alongside other agents and an Obsidian-based memory context system.00:00 Big Upgrade Overview01:02 Connect Grok In Hermes03:49 Enable X Search Tool06:56 Personalized Automation Ideas09:13 Video Generation Setup12:50 Image Generation Setup14:41 Text To Speech Voice16:02 Why Agent OS Matters19:51 Memory Vault With Omi21:09 Build Studio Dashboard23:52 Mission Operator Mindset26:12 Wrap Up And Resources27:08 Community Q And A27:41 Final Goodbye

Agent OS: Build a Local AI Operating System for Claude, OpenClaude & Hermes (with Obsidian Memory)The script explains Agent OS, a local “operating system” dashboard that runs multiple AI agents—Claude, OpenClaude, and Hermes—together so they can coordinate, share context, and be managed from one mission control interface with direct chat, agent status, a wired-in Kanban board, and built-in analytics. It contrasts this with the common setup of separate tabs and terminals that don’t communicate or retain context, and highlights privacy and speed from running locally. The core differentiator is a memory layer using an Obsidian Vault integration that gives agents access to a personal knowledge base and automatically logs conversations back into Obsidian, compounding usefulness over time. The script outlines the “Goldie Mission Stack” (intelligence: Claude; execution: OpenClaw; research: Hermes; self: Obsidian) plus Omi for automated daily note capture and task creation, and gives basic steps to start building the system and a call to join the AI Profit Ballroom for prompts, guides, and coaching.00:00 Why AI Feels Weak01:09 Agent OS Overview01:48 Inside the Dashboard02:56 Local Privacy Benefits03:11 The Memory Layer04:30 Obsidian Vault Integration05:41 Memory in Action06:25 Goldie Mission Stack08:58 Omi Auto Logging09:45 Built in One Hour10:25 System Beats Tools11:50 Real Workflow Example12:48 Future of Agent Systems13:40 Build Your Own Stack14:35 Get the Full Setup15:39 Final Call to Action

OpenClaw 5.12: Stability Fixes, Lighter Installs, Telegram Reliability & Should You Update?OpenClaw 5.12 arrives after several buggy releases that broke setups, wiped memory, and disconnected channels, contributing to Hermes Agent overtaking OpenClaw/OpenRouter in API call volume. The update focuses on stability: a lighter core install with channel/provider libraries installed only when needed, Telegram reliability improvements by moving message polling to a separate worker with local message backup and better stuck-message detection, fixes for group chat “require mention” image handling, and stored stream recovery to prevent mid-conversation hangs by rotating to backup models/logins or showing errors. Other changes include a new default OpenAI/Codex subscription login path, web chat auto-scroll controls, clearer CLI error guidance, security hardening for Windows folder sharing and credential handling, stricter pairing, and more reliable plugin installs. Community reactions show low trust, so the advice is to wait, back up, know your rollback version, and check Discord for fixes.00:00 State of OpenClaw00:43 Lighter Core Install01:54 Telegram Reliability Fixes03:28 Stream Hang Recovery04:19 Coaching Pitch Break04:50 OpenAI Codex Defaults05:26 Dashboard and CLI Improvements06:27 Security and Plugins Update07:16 Community Trust Issues08:10 Hermes Overtakes OpenClaw09:33 Should You Update10:20 Tool Choice and Mastery11:31 Final Call to Action

Build a Local AI Agent OS (Claude + Hermes + OpenClaw) with Mission Control + Infinite MemoryThe video introduces Agent OS, a locally hosted “mission control” dashboard that connects and manages multiple AI agents (Claude, Hermes, OpenClaw/OpenClaude) in one operating-system-like interface with chat, session history, agent status, control rooms, skills/plugins, Kanban integration, and analytics. The presenter shows that the system logs every conversation across agents into an Obsidian-based memory layer containing goals, journals, and a personal knowledge base, enabling agents to retain context about the user and business and improve over time. They explain how they built the dashboard in one session by prompting Claude (using a stack like Next.js and Tailwind and a Claude CLI bridge), argue local-first is more private and usable than terminal workflows, and outline a four-layer “Goldie Mission Stack”: intelligence (Claude), execution (OpenClaw), research/orchestration (Hermes), and self (Obsidian). The episode also references community resources, prompts, a 30-day roadmap, and answers setup and architecture questions.00:00 Agent OS Superpowers01:26 Mission Control Tour03:52 Terminal vs Dashboard04:29 Building It With Claude06:05 Obsidian Memory Vault09:02 Automation From Context11:15 New Features and Journals13:31 Why Agentic OS Matters15:28 Four Layer Mission Stack18:04 Goals Journals Memory19:15 Mission Operator Mindset19:39 System Beats Tools20:14 Simple Dashboard Control20:37 Context Is Superpower21:42 Local Setup Benefits22:46 Recap Goldie Stack24:33 Prompts Roadmap Community25:46 Community Q&A Hermes30:23 Client Value Calculator33:14 Hermes Middleware Options34:52 Final Wrap Up

Hermes Agent: Build & Automate Anything for Free (Memory, Paperclip Teams, Scheduled Tasks + Free Models)The video explains how to use the open-source Hermes Agent to build and automate workflows for free, including setting it up from GitHub and running free models via News Research Portal (e.g., DeepSeek V4 Flash and Step 3.5 Flash) or other options like OpenRouter’s free Owl Alpha and local models with Ollama. It demonstrates installing and using Paperclip to organize a team of autonomous agents, assigning goals and issues, and checking the inbox for updates. For stronger “never-forget” memory, it recommends combining Omi (screen/mic note-taking) with Obsidian, exporting daily notes, and using Claude or Hermes to reorganize the vault into a structured system (e.g., PARA) with a knowledge graph. It also covers making agents proactive with scheduled tasks, setting up Claude MCP with Hermes, troubleshooting Hermes Workspace sync issues, deployment/privacy options (separate user, VPS, or cloud one-click deploy), and ways to offer Hermes setups as a service focused on lead and customer outcomes.00:00 Free AI Employee Intro01:08 Paperclip Team Setup04:35 Selling Hermes Services06:51 Run Hermes Free Models08:03 Career Focus Timeline10:19 Second Brain Memory Stack14:34 Scheduled Tasks Autonomy17:57 Claude MCP Control19:16 Token Saving Options21:23 Safe Deployment Choices22:41 Overwhelm Focus System24:24 Fix Hermes Workspace Sync26:55 Local Models For Websites28:20 LangGraph Integration29:03 Wrap Up And Community

Run a Fully Local Computer-Use AI Agent (Codex + Claude + Ollama + CUA Driver) — Free & PrivateThe video shows how to set up a fully local “computer use” AI agent that can control your machine (open apps, write notes, browse the web) with no data leaving your computer, using Codex or Claude Code with local models. It explains that running local models with Codex/Claude can be done via Ollama, notes hardware considerations for powerful models, and demonstrates downloading a model (e.g., Qwen 3.6+). The script then shows how to run Claude Code or Codex with another model (DeepSeek) via terminal commands and introduces the key tool for computer control: CUA Driver, used as a skill to open macOS Notes and write content in the background. It recaps the steps (hardware, Ollama, model download, Claude/Codex setup, CUA Driver) and mentions additional training on Hermes Agent and a community with coaching calls and support.00:00 Local Computer Agents Intro01:10 Running Models with Ollama02:23 Using Claude Code and Codex03:30 Computer Use with CUA Driver04:08 Live Notes Automation Demo05:32 Recap and Next Steps05:46 Training and Community Plug06:37 Wrap Up and Goodbye

Build a Second Brain for Claude with Omi + Obsidian (Free Infinite Context Engine Tutorial)The video explains how to make Claude far more context-aware by connecting it to a “second brain” that automatically tracks your daily work and feeds that information into Claude or other AI agents. The process uses Omi to record screens, conversations, tasks, goals, and daily activity, then syncs those notes into an Obsidian vault of markdown files. Since Obsidian isn’t a built-in connector, the script shows linking Claude to the vault via MCP Obsidian or by opening the local Obsidian folder in a Claude session, enabling Claude to review, summarize, and use the vault as context. Claude is also used to clean up and restructure the messy vault using the PARA method (projects, areas, resources, archive), create dashboards, maps of content, templates, and improve Obsidian’s graph view. The episode ends by promoting a full tutorial and community support inside AI Profit Ballroom.00:00 Build a Second Brain01:16 Tools Overview Omi Obsidian02:05 Omi Demo and Workflow02:43 Connect Claude to Obsidian03:14 Claude Reads Your Vault03:53 Automations from Context05:49 Clean Up the Vault06:43 PARA Dashboard Setup08:25 Refine Links and Graph09:09 Live Q and Pricing10:10 Graph View Improvements11:22 Wrap Up and Next Steps12:04 Community and Training Offer

Automate Anything with Hermes Agent OS (Build a Real AI Travel Agent in 30 Minutes)The video demonstrates how to automate tasks using Hermes Agent OS by building a real AI-powered personal travel agent for an AI Profit Boardroom member planning a 9-day trip to Manila. The presenter walks through the Hermes Agent OS framework: identify the time-consuming task, engineer a single-purpose agent with a clear role/goal/tools, deploy it on a real workflow, test and “catch the gaps,” then expand to new use cases and systemize via SOPs. They show Hermes and Claude building the project structure and files in parallel, creating a skill wrapper and a simple local mini-app (trip monitor), noting that an Amadeus API key is needed for travel data. The episode emphasizes shipping an 80% working agent over perfection, scaling via multi-agent stacks, and that setup can be done in plain English without coding, often in under 30 minutes, plus details on the AI Profit Boardroom community and tools/models used.00:00 Automate Anything Intro00:47 Travel Agent Use Case01:42 Hermes Agent OS Framework02:35 Build the Agent Skill02:54 Deploy Test Improve Loop03:36 Agent Stacks and Teams04:18 APIs and Local Mini App05:06 Scale with SOPs06:12 Debunk Common Objections06:54 Framework and Community Offer08:10 Live Q&A Wrap Up