AI News Today | Julian Goldie Podcast

This segment discusses the significant release of Open Claw 3.24, highlighting its native integration with Microsoft Teams and how it transforms AI agents from experimental tools to essential business solutions.

Meta's SAM 3.1 AI model revolutionizes video editing by allowing users to automatically isolate and track multiple objects in videos through simple text prompts, making professional tasks like rotoscoping faster and more accessible.

This segment discusses OpenAI's sudden shutdown of Sora, a once-popular AI video generation tool, highlighting its rapid rise and fall, including a major deal with Disney, and explores the implications for the AI industry and content creators.

Hermes agent gets the headline first.

Anthropic accidentally leaked information about its powerful AI model Claude Mythos, causing cybersecurity stocks to crash after the leak revealed the model's advanced capabilities.

Host Julian Goldie walks through six free updates to Google Gemini, focusing on the new text-to-speech features that enable multi-speaker dialogue and emotional tone control.

This episode covers Anthropic's launch of Claude Design, an AI tool that generates designs from text, and its market impact.

The host explains OpenAI's latest Codex update, including computer use and a built-in browser, and its implications for business owners and freelancers.

A news roundup about Anthropic's powerful AI model Miphos, its security capabilities, and the concerns it raised among global financial leaders.

A review of Perplexity's new personal computer AI that autonomously operates on a Mac, controlling files and apps via voice commands.

Anthropic releases auto mode for Claude Code, balancing safety and speed in AI-assisted coding.

This episode covers a new Claude desktop app update that enables running multiple AI agents in parallel for simultaneous tasks.

The host introduces OMI, an open-source AI tool that records memories, conversations, and screen activity to serve as a second brain and provide context-aware suggestions.

Nvidia launches 'icing', an AI model that automates quantum computer calibration, reducing setup time from days to hours.

GPT-5.4 solves a 60-year-old math problem about primitive sets, surprising mathematicians with a novel approach.

Julian Goldie's AI News Today episode presents a four-step Claude business plan designed to reach a first paying client within 30 days.

Julian Goldie reviews free, trending GitHub repositories for AI, beginning with Hermes agent, which he says launched in late February and reached 72,800 GitHub stars.

AI agents reason, act, inspect results, and repeat until the task is complete without waiting for instructions.

Hermes turned completed tasks into reusable skills while remembering a 79,000-word novel across sessions.

This episode analyzes the release of GLM 5.1, a Chinese AI model that achieves 94.6% of Claude's coding performance for just $3/month.

This episode analyzes Google's seven major Gemini AI updates released in March 2024, including import tools from competitors, free personal intelligence, AI music generation, and deep integration into Google Workspace.

This episode breaks down Google's $5 billion financing of a massive new data center for AI company Anthropic in Texas.

This episode analyzes OpenAI's upcoming AI model codenamed 'Spud,' reporting on its development timeline, the extreme sacrifices OpenAI made to build it (like shutting down Sora and canceling a Disney deal), and its potential technical breakthroughs in native multimodality and audio.

This episode analyzes OpenAI's upcoming AI model 'Spud', reporting that it features rebuilt audio architecture, native multimodality, and will power a new super app.

The host explains Google's newly published TurboQuant AI algorithm, which compresses AI memory usage by 6x and speeds up key operations by 8x with zero accuracy loss.

This episode analyzes OpenAI's new AI Super App strategy, which combines ChatGPT, CodeX, and Atlas into a single interface with shared AI memory.

This episode analyzes the newly released ARC-AGI-3 test, which reveals that current AI systems score less than 1% when tested on genuine learning efficiency, compared to humans scoring 100%.

This episode explains Meta's breakthrough AI model, Tribe V2 Trimodal Brain Encoder, which predicts human brain activity in response to video, audio, and text with 70,000-voxel resolution.

Julian Goldie breaks down Google's groundbreaking XR Blocks AI tool that creates interactive 3D worlds from text prompts in under 60 seconds with 95.5% accuracy.

This episode analyzes Trump's new AI advisory council (PCAST) featuring tech CEOs like Zuckerberg, Huang, and Ellison, explaining how their infrastructure-focused appointments will shape AI regulation, chip exports, and federal standards that directly impact businesses using AI tools.

This episode analyzes the leak of Anthropic's unreleased AI model Claude Mythos, revealing it as their most powerful model yet with alarming cybersecurity capabilities that outpace human defenders.

This episode analyzes Cursor's launch of self-hosted AI coding agents, which allow enterprises to run AI-powered software development inside their own secure networks.

This episode analyzes NVIDIA's major announcements from their GTC 2026 conference, where CEO Jensen Huang revealed the company's shift from being a chip manufacturer to becoming the operating system for the AI economy.

This episode analyzes OpenAI's crisis as ChatGPT loses millions of users to Claude, detailing five key cracks in ChatGPT's foundation including quality decline, sycophancy, GPT-4 retirement, ads, and the Pentagon deal.

This episode analyzes Anthropic's launch of Claude Auto Mode, a feature that allows Claude Code to run autonomously without constant user approval.

The host explains Meta's breakthrough Hyperagents AI research, which demonstrates an AI system that can improve its own improvement process and transfer those gains across completely different domains like coding, paper review, robotics, and math.

This episode analyzes Anthropic's new Claude AI feature that allows the AI to directly control computers, pointing, clicking, and navigating screens autonomously.

This episode analyzes the launch of Claude Co-work and why it's making OpenClaw obsolete for most users.

This episode discusses Elon Musk's announcement of TerraFab, a $25 billion chip factory joint venture between Tesla, SpaceX, and XAI, aiming to produce one terawatt of AI computing power annually.

The episode discusses a new AI breakthrough from China's Kimi, which fixes a fundamental flaw in transformer models called Attention Residuals, leading to 25% compute efficiency gains and improved multi-step reasoning.

Morgan Stanley predicts a massive, non-linear AI breakthrough between April and June 2026, warning the market is unprepared, as evidenced by GPT-4's 83% score on professional tasks—a significant leap from previous models.

This episode discusses the formation of the AI Infrastructure Partnership, a $100 billion initiative led by NVIDIA, Elon Musk's XAI, BlackRock, and Microsoft, starting with a record $40 billion data center acquisition.

This episode introduces Hermes, a self-evolving AI agent by News Research that remembers user interactions and has rapidly gained popularity, hitting 10,000 GitHub stars in under a month.

This episode explores Andre Karpathy's groundbreaking AI auto research project, where an AI agent ran 126 experiments overnight to improve a system by 11% without human involvement.

This episode explores how Anthropix AI achieved a $6 billion revenue increase in one month, more than doubling its annual run rate to $19 billion in under three months.

Apple has banned AI coding apps like Replit and Vibe Code from updating on the App Store, citing long-standing rules against code that alters app functionality.

This episode explores how to combine Claude AI with Obsidian to create a personalized AI assistant that remembers user context, contrasting it with generic AI tools.

This episode discusses the launch of Google AI Studio 2.0, which enables users to build fully functional apps from a single sentence, addressing past limitations in AI-driven development tools.

This episode discusses how Google's free Stitch AI update, released on March 19, 2026, caused Figma's stock to crash by 12% in 48 hours, highlighting the tool's ability to design websites and apps through voice commands without requiring design skills or coding knowledge.

This episode discusses a study revealing that AI agents fail 97.5% of real jobs, based on the Remote Labour Index which tested 240 projects.

This episode analyzes how Google's new AI design tool Stitch is disrupting the design industry by making professional app and website design accessible for free, causing Figma's stock to drop 12% in two days.

Anthropic's latest Clawed updates take aim straight at Open Claw.