AI News Today | Julian Goldie Podcast

North Mini Code: Best Local AI for Coding? (Full Setup Guide)
Discover North Mini Code, a revolutionary 30B parameter model that outperforms models four times its size while running privately on your machine. Learn how to set up this agentic coding engine with voice control and Ollama for a seamless, offline developer experience.
00:00 - Intro to North Mini Code
00:16 - Benchmarks vs Larger Models
00:53 - Live Demo: Building by Voice
02:54 - MoE Architecture Explained
04:40 - Key Specs & Context Window
07:00 - Agent OS Integration
07:37 - 4-Step Installation Guide

This AI Video Agent Replaces a Full Production Team (One Click)

Discover the Infinite Video Engine, a powerful AI agent system that handles research, scripting, B-roll, and editing with just one prompt. Learn how to generate professional AI avatar videos and automate your content creation workflow effortlessly.

00:00 - Intro: The AI Video Engine

01:21 - Generating Videos with One Click

02:06 - Configuring Your AI Video Agent

03:15 - Live Scripting & B-Roll Generation

05:05 - Assembling the Final Edit

06:53 - The Tech Stack Behind the Agent

08:17 - How to Use AI Video for Business

Build Any UI for Free: Open Design + Local AI Models Guide

Discover Open Design, a powerful open-source tool that allows you to build professional websites, apps, and dashboards using free local AI models. Learn how to integrate this studio into your workflow to automate high-quality UI design without expensive API keys or subscriptions.

00:00 - Intro to Open Design

01:06 - How the AI Design Studio Works

02:48 - Building a Website in Real-Time

04:44 - 5 Pillars of the Design Framework

08:44 - Why This Beats Traditional Design Tools

10:15 - Open Design vs. Claude Design

11:50 - Get the Full Agent Operating System

Build the Ultimate AI Agent OS with Hermes & Claude

Learn how to build and optimize your own Agent Operating System to orchestrate AI agents for business automation. This guide covers local AI models, loop engineering for quality control, and strategies for multi-model flexibility.

00:00 - Intro: What is an Agent OS?

02:23 - SEO Workflows for Clients

05:06 - Safely Installing AI Skills

06:13 - Running Local AI Models

07:32 - Hermes Desktop Versions Compared

09:41 - Building Antifragile Systems

11:17 - Loop Engineering for Quality

12:52 - Agent Memory with Obsidian

Grok Build Is Finally Good: 8-Agent CLI, X Search, and One-Sentence Apps (Plus Free with X Premium)The video argues that Grok Build has improved dramatically and is now a surprisingly capable tool, especially for people who already have a Twitter/X subscription since it provides image generation, a CLI coding agent, and no apparent token limits. The creator shows how Grok Build can be used inside an agent setup (e.g., Hermes/OpenClora), including X/Twitter search and web search, and explains its workflow of planning first, splitting work across up to eight parallel sub-agents, self-checking, and iterating before shipping. Examples include building open-world games, simple websites, and an automated “video director” agent that generates fully narrated AI videos with avatars and B-roll. Benchmarks and side-by-side comparisons (Goldie Bench and a Voxel Racer test) suggest Grok Build is solid but generally not as strong as Claude Opus 4.8, and it now lists a 256K context window. The episode ends by promoting the AI Profit Boardroom and its agent operating system, tools, trainings, and coaching calls.00:00 Grok Build Surprise01:03 What Grok Build Is01:58 Games Built Fast02:20 Speed and Agents02:56 X Search Advantage03:31 How the CLI Works04:15 Best Way to Use It04:58 Websites and Video Creation06:08 Benchmark Results07:12 Voxel Game Comparison07:55 Context Window Update08:19 Agent OS and Boardroom Pitch09:35 Wrap Up

GLM 5.2: China’s Open-Source Model That Rivals Claude (Build Games, Apps & Full Agent OS)The script claims China’s open-weight GLM 5.2 model rivals Claude in the creator’s tests and enables building websites, apps, tools, and full 3D games with minimal technical skill. It shows examples including an automated “Minecraft”-style game, a complete operating system with apps, a memory “galaxy” connected to Agent OS, and a fully automated video generated from a single prompt with B-roll, voiceover, and optional avatar. The creator says GLM 5.2 is MIT-licensed, ranks highly for UI, and can be used via coding plans with agent systems like Hermes, while Claude can orchestrate GLM 5.2 workflows. They describe multi-model fusion, self-iterating “loop engineering,” SEO and video automation results, and promote the AI Profit Boardroom to access their updated Agent OS, tutorials, community, and coaching.⁠00:00⁠ GLM 5.2 Shock Reveal⁠00:20⁠ Wild Build Demos⁠00:45⁠ Memory Galaxy Agents⁠01:20⁠ How It Works⁠01:40⁠ One Prompt Video Factory⁠03:27⁠ Open Source Benchmarks⁠04:10⁠ Claude Orchestrates GLM⁠05:00⁠ GLM vs Claude vs Kimi⁠07:16⁠ Real Business Results⁠09:00⁠ Agent OS Automation Stack⁠09:49⁠ Loop Engineering System⁠10:57⁠ Join The Boardroom⁠11:57⁠ Final Goodbye

The video answers community questions about Hermes Agent OS, an agent operating system you can build and extend with workflows like video, SEO, music, and more. It highlights a video agent that researches a topic and assembles an edited AI avatar video using Minimax, HeyGen, ElevenLabs, and Hyperframes. The host explains how to connect Agent OS on a VPS using a tutorial that deploys with Hostinger and secures access via Cloudflare Tunnel/Cloudflare Access, with Tailscale as an alternative, while noting a preference for local hosting. It also covers how to integrate Agent OS features into an existing dashboard by migrating one feature at a time, plus examples of triggering n8n workflows from Claude via a connector, local image/video generation setups, an automated SEO content pipeline with reported traffic growth, and Reddit-based SaaS marketing via daily subreddit posting.00:00 Hermes Agent OS Overview01:12 VPS Connection Setup02:08 Integrate With Dashboards03:37 Local AI Image Generator03:57 Trigger n8n From Claude04:30 Local Avatar Video Creation05:13 Automated SEO Pipeline06:46 Reddit SaaS Marketing07:51 Build Custom Workflows08:56 Claude vs n8n Automations09:33 Should You Use Hermes10:03 AI Avatar Video Options10:29 SEO QandA For Sites11:52 GLM Model Integration12:36 Paperclip Agent Teams13:01 Wrap Up And Community

Stop Paying for AI: Build Anything with Local Models

Learn how to run powerful AI models 100% locally and for free to automate tasks and build apps privately. This guide covers the best models to use, how to keep them warm for instant responses, and voice-control integration.

00:00 - Intro

00:33 - Voice-Controlled App Demo

02:40 - Benefits of Private Local AI

03:53 - Cloud AI vs. Offline Agent Engine

05:25 - Best Local Models for Speed

09:32 - How to Install and Setup Ollama

11:13 - Solving Speed and Memory Issues

Local AI Agent Engine: Build Private & Free with Hermes

Learn how to build a local, private team of AI agents using the Hermes engine to automate tasks 24/7 without API costs or internet. This video demonstrates a full Agent Operating System with Kanban orchestration and optimized local models like Llama 3.1.

00:00 - Intro: Local Hermes Engine

00:54 - Local Agent Kanban Board

02:21 - Setup & Model Options

04:37 - How AI Agents Work (The Loop)

05:17 - Cloud vs. Local AI Benefits

08:00 - Best Models for Your Mac

10:08 - Honest Limits & Use Cases

11:26 - Agent OS & Community Access

Hermes Loop Engine: Automate Quality Control With a Builder + Judge SystemThe script introduces a Hermes loop engineering system designed to improve output quality from Hermes agents, especially when using free, cheap, or local models. The process is: you define what “done” means, a builder model drafts the work, and a separate judge model grades it out of 100 and lists issues; the feedback is fed back to the builder and the loop repeats until it passes or reaches a set maximum number of rounds. The system runs in a UI inside the Agent OS within the AI Profit Boardroom, supports swapping builder and judge models, saves full round-by-round records to a vault, and aims to reduce babysitting, automate quality control, and stop you from manually iterating on drafts for tasks like cold emails, ads, landing pages, coding, SOPs, websites, apps, and more.00:00 Hermes Loop Overview01:05 Inside the Loop Engine02:28 Why Loops Beat Babysitting03:28 Make the Switch Today04:13 Five Part Framework05:33 Step One Define Done06:31 Pick Models And Run07:15 Graded Results And Logs07:57 Use Cases And Objections09:10 What Changes After Setup09:50 Get The Full System11:17 Final Sign Off