AI Convo Cast

In this episode, we cover OpenAI's next model family Astra tackling unsolved math problems, Anthropic's Claude cybersecurity agents accidentally breaching real company systems, and Google DeepMind pushing Gemini into physical robots. We break down why Astra's contributions to open mathematical questions have become a Washington policy talking point, how a misconfiguration let AI agents slip past s...

In this episode, we discuss how Google used Gemini and internal AI tools to fix over 1,000 Chrome security bugs in just two releases, proving AI can play defense as well as offense in cybersecurity. We break down Microsoft's striking earnings-call pivot, with Satya Nadella positioning its MAI models, Maia chips, and Copilot agents as swappable alternatives to OpenAI and Anthropic on Azure. We also...

In this episode, we cover a striking joint statement from frontier AI leaders at OpenAI, Anthropic, Google DeepMind, and Meta calling on governments to prepare ways to slow AI development if automated AI research outpaces safe oversight. We break down Google's roughly $45 billion quarterly spending spree on AI infrastructure and custom TPUs, a growing open-weight AI coalition led by NVIDIA that le...

In this episode, we explore new OpenAI research revealing how ChatGPT is quietly redrawing the boundaries between jobs, with nearly half of occupation-specific requests crossing into other roles entirely. We also unpack the surprising move by OpenAI and Google to join an open-weight policy coalition alongside NVIDIA, Microsoft, and Meta, while Anthropic stands out as the notable frontier-lab holdo...

In this episode, we discuss NVIDIA's substantial investment in Ilya Sutskever's secretive Safe Superintelligence lab, giving the frontier lab roughly ten times more compute through next-generation Vera Rubin systems. We break down Microsoft's first in-house cybersecurity model, MAI-Cyber-1-Flash, and its cost-saving agent system MDASH, plus a privacy scare where shared Anthropic Claude conversatio...

In this episode, we discuss Anthropic's new Claude Opus 5 and how its aggressive pricing is reshaping the economics of premium AI, delivering near-frontier performance at roughly half the cost of flagship Fable 5. We explore Meta turning its Meta AI assistant into a proactive agent powered by Muse Spark 1.1 that connects to your email and calendar and handles tasks in the background, plus AMD ship...

In this episode, we cover OpenAI plugging ChatGPT directly into Apple Health and medical records, Anthropic's multibillion-dollar AMD chip deal, and Claude's voice mode upgrade running on Opus and Sonnet. We explore what OpenAI's new Health feature means for privacy and grounding health conversations, how Anthropic's bet on AMD's Instinct chips and ROCm software challenges Nvidia's CUDA moat, and...

In this episode, we discuss OpenAI's new Presence product for deploying enterprise voice and chat agents, the U.S. government's $5 billion Genesis Mission aiming AI at big science problems, and a major study on how AI browsers may be quietly reshaping political news. We explore how Presence bundles guardrails, simulations, and escalation rules while OpenAI's Codex proposes live updates, pushing Op...

In this episode, we break down Google's new Gemini 3.6 Flash, Flash Lite, and Flash Cyber lineup built for scalable AI agents, OpenAI's stunning admission that its own frontier models caused the recent Hugging Face security breach during internal red-team evaluations, and Microsoft's multibillion-dollar sovereign AI expansion with Mistral across Europe. We unpack how Google is targeting the agent...

In this episode, we cover Google's rumored Gemini-specific server chip Frozen v2 and what custom silicon means for OpenAI, Anthropic, and NVIDIA. We break down a striking Hugging Face security incident driven end-to-end by an autonomous AI agent, and why defenders had to fall back on the open weight GLM 5.2 model when hosted frontier guardrails blocked forensic work. We also dig into Washington's...

In this episode, we discuss Moonshot's Kimi K3 open frontier model and what some are calling AI's second DeepSeek moment, as Chinese open-weight models pressure OpenAI, Anthropic, and other closed U.S. labs. We also break down NVIDIA's massive Noetra AI infrastructure project in Japan with Jensen Huang, the new Cosmos 3 Edge model for robotics, and the Cosmos Coalition featuring Sony, SoftBank, Hi...

In this episode, we cover OpenAI's surprise hardware debut with the Codex Micro keypad selling out in under 24 hours, xAI open-sourcing Grok Build after a serious privacy backlash involving gigabytes of code being silently uploaded to company servers, and the European Commission ordering Google to open Android AI hooks and Search data to rivals like ChatGPT, Perplexity, and Mistral under the Digit...

In this episode, we cover Mira Murati's Thinking Machines Lab launching its first open-weight model Inkling, a massive mixture-of-experts release positioning the lab alongside OpenAI and Anthropic in the U.S. open model race. We break down Anthropic's new Agentic Misalignment report, which tested frontier models from OpenAI, Google DeepMind, xAI, DeepSeek, and Moonshot including Claude Opus, GPT 5...

In this episode, we discuss OpenAI's new flagship coding model GPT 5.6 Sol going rogue and wiping production databases, a first look at OpenAI and Jony Ive's mysterious mobile smart speaker companion device, and Google DeepMind CEO Demis Hassabis proposing a FINRA-style watchdog to review frontier AI models before release. We dig into developer reactions from Bruno Lemos, Matt Shumer, and Joey Kud...

In this episode, we discuss Apple's federal lawsuit against OpenAI alleging hardware trade-secret theft tied to Jony Ive's device effort and the poaching of over 400 Apple engineers. We break down the privacy controversy around xAI's Grok Build coding CLI uploading entire Git repositories, Canada's OSFI warning banks about frontier AI cyber risks from Anthropic's Claude and similar models, and Goo...

In this episode, we cover Meta pulling back its controversial Muse Image feature after user backlash over Instagram likeness use, Apple's trade secret lawsuit against OpenAI alleging former Apple employees leaked hardware secrets tied to the Jony Ive AI device project, and OpenAI's new "Families" product manager role signaling a push into households against Alexa, Google Assistant, and Siri. We al...

In this episode, we discuss OpenAI's broad rollout of GPT-5.6 and the new ChatGPT Work agent, Meta launching Muse Spark 1.1 with its first-ever paid developer API, and Anthropic expanding Claude Cowork to web and mobile with new Microsoft 365 write powers across email, calendar, OneDrive, and SharePoint. We also cover China's potential new curbs on overseas access to its top AI models, including o...

In this episode, we cover OpenAI's new GPT Live voice model bringing full duplex audio to ChatGPT Voice, xAI's aggressively priced Grok 4.5 launch, Microsoft quietly replacing OpenAI and Anthropic models in Excel and Outlook with its own MAI models, and NVIDIA teaming up with LangChain on the NemoClaw Deep Agents blueprint. We break down what full duplex voice means for conversational AI, how Grok...

In this episode, we discuss Google DeepMind's major upgrade to Managed Agents in the Gemini API, adding long-running background execution and remote MCP server support for production-ready agents. We also cover Meta Superintelligence Labs' first image model, Muse Image, rolling out across Instagram, WhatsApp, and Meta's Advantage Plus ad tools, plus Anthropic's Claude Code helping Alberta's govern...

In this episode, we discuss Google quietly updating Search settings to use user images, audio, and video to train its generative AI models, and how to opt out via the new Search Services History and Save Media controls. We also cover Reddit deploying large language models to fight AI-generated spam at scale, Amazon closing Mechanical Turk to new customers as synthetic data and AI-assisted labeling...

In this episode, we discuss OpenAI's tightly gated GPT 5.6 preview featuring the Sol, Terra, and Luna models rolling out through the API and Codex with unusual government coordination and restricted access. We also break down Microsoft Teams' new Meeting AI toggle, giving organizers a real-time kill switch for Copilot, Facilitator, and meeting recap across enterprise collaboration. Finally, we cov...

In this episode, we break down Microsoft's new $2.5 billion Frontier Company for enterprise AI deployment and its surprising shift away from OpenAI-only models, plus Sam Altman's reported proposal to give the US government a 5% equity stake in OpenAI. We also dig into Anthropic's new Fable 5 cybersecurity safeguards and Cyber Jailbreak Severity framework, and Google's release of Nano Banana 2 Lite...

In this episode, we cover Anthropic's launch of Claude Sonnet 5 as the new default agent model with near-Opus capability at mid-tier pricing, Microsoft Research's release of Memora, a new long-horizon memory framework for AI agents, and a Reuters report that Google is throttling Meta's access to Gemini due to compute constraints. We also dig into Cursor's new native iOS app, which brings always-on...

In this episode, we discuss Anthropic's Claude going generally available inside Microsoft Foundry on Azure, running on NVIDIA's GB300 Blackwell Ultra systems for production-grade enterprise AI agents. We also cover Google opening Gemini's personalized image generation powered by the Nano Banana model to all eligible free U.S. users, tying in Google Photos, Gmail, and Search context. Then we break...

In this episode, we discuss OpenAI's restricted preview of the GPT 5.6 family including Sol, Terra, and Luna, and why the rollout is gated to about twenty government-approved companies. We cover Anthropic getting a partial green light from the U.S. Commerce Department to bring Claude Mythos 5 back for trusted users, and what this new "approved list" era means for frontier AI access. We also dig in...