Podcast
AI News Today | Julian Goldie Podcast
Hosted by Julian Goldie · EN
Latest Podcast
275episodes
Episodes
Newest firstAll episodes
How to Build Your Own Agent Operating System
5d ago00:15:52Tap to summarizeHow to Build an Agent Operating System (Agent OS) in 5 LayersThis episode explains how to build an “Agent OS,” a single system where multiple AI models and agents (like Claude and Hermes) work together with shared context to automate tasks such as lead generation, outreach, news research, SEO, video, and more. The creator describes replacing “14 scattered tabs” with a five-layer stack—foundation (one dashboard), memory (a shared Obsidian vault), routing (send each job to the best model, often free/local), agents (parallel execution via workflows like Kanban), and loop engineering (feedback that improves the system daily). Built from scratch in about eight weeks with minimal added cost beyond existing subscriptions, the system can be customized for any industry and is shared with a 3,900-member AI Profit Bot community.00:00 What Is Agent OS02:50 Why It Beats 14 Tabs05:20 Five Layer Overview06:18 Layer 1 Foundation07:36 Layer 2 Memory09:33 Layer 3 Routing12:04 Layer 4 Agents13:49 Layer 5 Loop14:36 Get The Full Setup15:51 Final Wrap Up
Transcribe →I Built an AI Agent Command Center
5d ago00:14:54Tap to summarizeHermes Agent OS Q&A: Skills Migration, Best Settings, SEO Automation, Memory, and Mobile AccessThe episode answers recent community questions about the Hermes Agent Operating System, which unifies tools like Hermes Oracle, a studio for image/video/voice generation, GLM 5.2 in quad code, Open Montage for AI film creation, and a memory system. It explains that Claude CLI skills can be transferred to Hermes on a VPS by moving the Markdown skill files, though smoothness may vary by API. Five recommended configurations include Hermes Oracle, media generation with Groq or Wolf, an outreach inbox for sending emails, Firecrawl for web search, and using /learn to ingest guides into skills. It also covers Windows installation guidance, SEO keyword research/content generation/deployment within Agent OS, model fallback via backup keys and auxiliary task fallbacks, hosting AI sites on Netlify, mobile access via Tailscale or Cloudflare, and a simple three-step SEO automation prompt flow, plus an Omi-to-Obsidian-to-memory workflow and how to view updates, changelogs, and update files.00:00 Agent OS Overview00:50 Community Q&A Kickoff01:10 Migrating Skills Setup Tips02:57 Windows Installation Options03:44 Showcase 360 Tour Build04:28 Automating SEO in Agent OS05:23 Model Fallbacks Multi Agent06:47 Fable Five Builds Benchmarks07:30 Using Local Files Screenshots07:51 Running Agent OS on Phone09:09 Hosting Sites with Netlify10:01 Simplest SEO Automation Flow10:43 Omi Obsidian Memory Pipeline12:26 Updating Agent OS Changelog13:37 Wrap Up Join Community
Transcribe →Laguna XS 2.1: New FREE + Opensource Local AI!
5d ago00:10:42Tap to summarizeLaguna XS 2.1: New Local Agentic Coding Model (Benchmarks, Demos, Setup Options)Julian tests the newly released Laguna XS 2.1 local agentic coding model from Poolside, highlighting its speed, lightweight setup, 256K context window, and availability on Hugging Face. He compares benchmark results (including SWE-bench) against Qwen 3.6, North Mini Code, Claude Haiku 4.5, and GPT-OSS, then demonstrates real outputs: a clean landing page, a functional to-do app, and another simple page, noting UI limitations but solid working code. He shares his hardware setup (Mac Studio M4 Max, 36GB) and plans to run broader evaluations on Goldie Bench. He explains ways to use the model via Agent OS local engine, Free Claude Code, or Hermes agent profiles, plus offline privacy benefits and a free OpenRouter API option.00:00 Laguna XS 2.1 Intro00:26 Benchmarks Breakdown02:20 Real World Builds02:47 Landing Page Demo03:16 To Do App Test04:18 Setup and Limits04:33 Goldie Bench Plans05:24 Ways to Use It06:19 Offline and Privacy06:50 OpenRouter Free API07:26 Hermes Terminal Setup07:57 Wrap Up and Next Tests08:05 Agent OS and Boardroom09:11 Community Examples10:41 Final Thanks
Transcribe →Gemma 4 is now 90% FASTER + FREE + Local 🤯
5d ago00:10:42Tap to summarizeGemma 4 Just Got Up to 90% Faster on Mac (Ollama + MLX)Google’s Gemma 4 has a new update that makes it up to nearly 90% faster on a Mac when run through Ollama using MLX, enabling much faster local model performance. The script explains that Gemma 4 is free on Ollama and MLX is a free open-source project, and that Ollama now supports MLX models directly so you can install via a simple command after updating. The creator reports seeing roughly 60% faster speeds in many tests, with a jump from about 50 to 95 tokens/sec in a shown comparison, and says coding prompts can hit the 90% boost. They also show Gemma 4 integrated into their Agent OS (with Hermes), letting it build and preview simple apps locally and autonomously.00:00 Gemma 4 Speed Boost00:49 MLX and Ollama Setup01:58 Agent OS Integration02:33 Real Speed Benchmarks04:07 Live Build Demo05:27 Why It Runs Faster05:56 Install in Terminal06:13 Projects Built With Gemma07:25 Limits and Best Uses07:56 Free Local Builds Anytime08:41 Agent OS and Boardroom Pitch09:43 Community and Support Features10:13 What Members Are Building10:41 Wrap Up and Thanks
Transcribe →How to Run Hermes FREE Forever!
5d ago00:10:36Tap to summarizeRun Hermes Free Forever: Gemma 4 MLX Update Makes It Up to 90% Faster (Ollama + Apple Silicon)Julian demonstrates how to run Hermes for free using a new Gemma 4 update with Ollama on Apple Silicon via MLX, claiming up to 90–95% faster local performance compared to before. He shows Gemma 4 set up as a free profile inside an agent OS, with examples like quickly responding to prompts, building a simple to‑do list app, and using Hermes’s /learn command to turn a tutorial into a reusable skill that runs in the background. For non‑Mac users, he notes an alternative: using a free OpenRouter API option for the 31B model. He explains that local models keep workflows private, offline, and suitable for long-running agent loops without token or subscription costs, and he briefly outlines setup by updating Ollama and selecting the newer MOX Gemma 4 models.00:00 Run Hermes Free Faster00:33 Agent OS Demo Setup01:14 Free API Alternative02:08 Hermes Learn Skill03:40 Install Ollama Gemma05:20 Loops and Subagents06:55 Local Loop Engine08:05 Easy Commands Objections08:58 Wrap Up and Offer09:20 Boardroom Walkthrough10:35 Final Thanks
Transcribe →Claude Code is now FREE: Here’s how…
5d ago00:09:21Tap to summarizeUse Claude Code Free Forever with Gemma 4 (90% Faster Local AI on Apple Silicon)The video explains how to run Claude Code for free by pointing the Claude Code CLI at Google’s Gemma 4 model, highlighting a new update that makes Gemma 4 about 90% faster on Apple silicon when run locally with Ollama using MLX. It outlines a simple setup: download Ollama, download Gemma 4, and connect it to Claude Code with a one-click install/one terminal command. The script shows demos of building apps (like a to-do list and a Space Invaders game) via an agent workflow and notes that Gemma 4 is best for basic tasks versus frontier models for complex work. It also describes a free alternative for non-Mac users by accessing Gemma 4 via OpenRouter’s free API and routing it into the open-source Claude Code project.00:00 Free Claude Code Setup00:30 Gemma 4 Speed Update00:53 Install in Three Steps01:32 Live Build Demos02:14 Hermes Agent Notes02:38 No Mac Use OpenRouter03:46 When to Use Gemma04:37 Agent OS Local Engine05:11 Why This Beats Tokens05:53 Common Objections Answered06:50 Testing and Leaderboards07:34 Why Local Models Matter08:07 Recap and Next Steps08:24 Join the Boardroom09:19 Final Thanks
Transcribe →Agent OS: Fable 5 vs Fusion, Windows Setup, and Free APIs
6d ago00:11:35Tap to summarizeAgent OS Q&A: Fable 5 Tests, Hermes Setup, VPS/Windows Tips, and Free API OptionsThis episode answers community questions about an agent operating system setup with memory and multiple agents (including newly added NotebookLM short videos), covering what people are using Fable 5 for, its credit/token limits, gating to Opus 4.8, Claude service-busy bugs, and upcoming Goldy bench tests and leaderboard updates versus Hermes-based mixtures. It suggests ways to link separate Claude accounts (terminal vs Desktop, VPS instances), and offers solutions for older hardware such as running OpenClaw/Hermes via a VPS, Raspberry Pi, or Tailscale. The host points to an existing Hermes Desktop masterclass but recommends the Agent OS for customization, shows a member-built offline job triage/outreach system, and shares Windows setup preferences (using Claude to configure Agent OS). It also covers Owl Alpha’s removal from OpenRouter and alternatives like OpenRouter free models, Minimax/Kimi plans, Groq Build via Twitter OAuth, local/NVIDIA options, plus VPS self-hosting and Cursor integration ideas.00:00 Agent OS Q&A Overview00:36 Fable 5 Limits and Tests02:16 Linking Claude Accounts02:48 Running Agents on Weak Hardware03:39 Hermes Desktop vs Agent OS04:33 Member Build Job Outreach05:26 Windows Setup Choices06:06 Free APIs and Owl Alpha07:17 VPS Self Hosting Guide07:49 Cursor and Open Design Models08:53 Tutorial Library and xMCP10:46 Community Access and Wrap Up
Transcribe →NEW Claude Agentic Operating System is INSANE!
6d ago00:14:01Tap to summarizeBuilding a Custom Claude Agent OS: Voice Agents, Orchestration, Memory Loops & Model SwappingThe speaker demos a custom “Claude agent operating system” built over two months to unify tools like Fable 5 and Hermes into a single, customizable mission-control UI with shared memory and organized outputs. They show how features evolved from simple chat into multiple model profiles, Hermes Mixture of Agents, a wake-word voice agent that can control the browser and local actions, and Hermes Oracle for pulling and linking breaking news and drafting social content. The system includes agent orchestration via group chat, pipelines, and a Paperclip-style org chart with departments, an inbox, dashboards, and progress history, plus Kanban workflows and integrations like NotebookLM video generation. It’s designed to swap models/CLIs quickly, run scheduled “loop” automations (outreach, news refresh), and maintain a growing memory/notes system. The OS and tutorials are offered through the AI Profit Boardroom.00:00 Agent OS Overview00:21 Why Mission Control01:12 Profiles and Mixture02:47 Keep It Simple03:11 Voice Agent Demo04:14 Oracle News and Content04:29 Building With Claude05:18 Agent Mastermind Chat06:05 Paperclip Agent Org08:07 Swap Models Fast09:16 Kanban and Video Tools10:12 Loop Engineering Automation11:24 Memory Feedback System12:12 Community and Access13:14 Courses Coaching Wrap14:00 Final Thanks
Transcribe →Hermes Mixture of Agents DESTROYS Claude Fable 5?
6d ago00:13:12Tap to summarizeFable 5 vs Hermes Mixture of Agents vs Fusion: Side-by-Side Build Tests (Games, WebOS, Landing Pages)The video compares Anthropic’s Fable 5, Hermes Mixture of Agents, and Open Router’s Fusion by testing 42 builds and showing side-by-side results across multiple projects, including a racer game, Nordic crypt game, orbit simulation, landing page, Dragon Realm, a WebOS, VoxelCraft, Frostfell, and a Doom-style game. Fable 5 wins the racer game, crypt game, and WebOS for usability and stability, but loses the orbit test to Hermes and the landing page test where Hermes produces the preferred design and Fusion beats Fable 5. Fusion wins VoxelCraft, while Frostfell is considered a tie, and Fable 5’s Doom build fails basic movement, giving the edge to Hermes/Fusion. The creator suggests mixing Mixture of Agents with Fable 5, recommends Hermes as a backup when Fable 5 tokens run out, notes Fable 5 subscription availability until July 7, and promotes tools, guides, and community access via AI Profit Boardroom and Goldie Bench.00:00 Showdown Setup00:18 How Agents Combine Models01:04 First Game Comparisons01:42 Nordic Crypt Results02:48 Orbit Simulation Test03:41 Landing Page UI Battle04:42 Why Mixture Works05:31 Dragon Realm Tie06:27 WebOS Usability Test08:30 VoxelCraft and Frostfell09:53 Doom Clone and Scorecard10:51 Final Takeaways11:59 Community and Resources
Transcribe →Claude Fable 5 is back!
6d ago00:09:58Tap to summarizeClaude Fable 5 Is Back: 42 Builds, Real-World Tests, Limits, and Better AlternativesThe script reviews Claude’s Fable 5 after its return, showing results from 42 builds and how to access it by selecting Fable 5 in Claude’s model list. It notes Fable 5 is included in a subscription only until July 7, after which it may require extra paid credits, and it sometimes auto-switches to Opus 4.8—especially on science-related prompts. The creator demonstrates strong coding and project-building speed (including a web “operating system”) but highlights weaknesses such as poor web UI output, buggy first builds, and inconsistent instruction-following compared with a custom-trained Opus 4.8. Side-by-side tests show panel approaches (Hermes Mixture of Agents, Fusion, Sakana Fugu) can outperform Fable 5 on some games and simulations, and the video promotes access to these systems and tutorials via the AI Profit Boardroom community.00:00 Fable 5 Returns00:32 Access and Pricing Window01:15 Showcase Web OS Build01:38 Panels vs One Shot Tools02:08 UI Weaknesses and Bugs03:15 Tutorial Skill Compliance04:33 Backlash and Model Switching05:31 Side by Side Build Tests08:07 Alternatives and Benchmarks08:41 Community and Training Offer09:52 Wrap Up and Links
Transcribe →