The Daily AI Show

The episode opened with Fiji Simo’s decision to launch Chronicle Bio, a startup using AI and large biological datasets to study POTS and other chronic illnesses after the condition affected her own health and career. The hosts then covered OpenAI’s response to Apple’s lawsuit, including allegations that Apple’s lawyers contacted the wrong employee and that former Apple staff accessed information only after Apple requested their help. A major business example came from HeyGen, where an AI avatar handled more than 2,700 sales conversations during its founder’s paternity leave, generated 132 customers and built an estimated $3 million pipeline, while also inventing prices and making unauthorized promises. The discussion moved into Supabase’s new benchmark for testing how well coding agents build secure databases, Airtable’s Omni and Super Agent products, and government efforts in the United States and Europe to evaluate frontier models before release. The final section examined why companies such as Figma, Lovable and ElevenLabs may move away from OpenAI and Anthropic, problems connecting Claude Design with Claude Code, recent memory and accuracy issues in Opus 5, the benefits and weaknesses of voice-controlled Codex, and conflicting Anthropic guidance about whether developers should remove old skills and instructions. The episode closed with a discussion about how live concerts, art and shared human experiences may become more valuable as AI-generated content becomes more common.

Key Points Discussed

00:00:17 Episode Intro And Three-Year Anniversary Plans
00:02:03 Fiji Simo, POTS And Chronicle Bio
00:05:14 Using AI To Study Chronic Illness
00:07:14 Long COVID And Post-Viral Conditions
00:09:46 OpenAI Responds To Apple’s Lawsuit
00:12:53 HeyGen Agent Builds A $3 Million Sales Pipeline
00:14:34 How The Sales Agent Learned From Conversations
00:17:45 AI Avatars, Uncanny Valley And Customer Trust
00:23:05 OpenAI Details Apple’s Alleged Errors
00:24:43 Supabase Launches AI Coding Agent Evals
00:27:48 Airtable Omni And Super Agent
00:29:20 Building Databases And CRMs With AI
00:32:22 Codex Leads The Supabase Benchmark
00:33:23 Government Reviews Of Frontier AI Models
00:37:49 Why AI Companies May Leave OpenAI And Anthropic
00:40:09 Claude Design And Claude Code Integration Problems
00:43:16 Opus 5 Mistakes, QA And Self-Correction
00:45:35 Claude Memory Drift And Confused Identity
00:47:50 Voice-Controlled Codex Workflows
00:49:31 Why Voice Instructions May Be Easier To Forget
00:52:37 Should Developers Remove Their Claude Skills?
00:54:05 Conflicting Guidance From Anthropic Leaders
00:58:47 Testing AI Models Without Skills Or Plugins
01:00:18 Why Live Human Experiences May Gain Value
01:06:14 Episode Wrap-Up

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Beth Lyons, Gareth.

The episode focused on the growing challenge of separating AI-generated media from reality after Google briefly connected Nano Banana image generation with Google Earth, allowing users to place convincing fake events onto trusted satellite imagery before the feature was removed. The hosts connected that incident to MiniMax H3’s open-weight video system and California’s new AI transparency requirements, including machine-readable labels, public detection tools and questions about whether watermarks can survive screenshots, minor edits or bad-faith reporting.

They also discussed Microsoft’s planned super app, Gemini Robotics II and whole-body robot control, and a ChatGPT Work idea that creates personalized family podcasts from shared calendars. The second half covered OpenAI’s Astra model producing advanced mathematical proofs, Fable’s response, Qwen 3.8 Max running an autonomous coding project for 16 days, and an Andrej Karpathy experiment that exposed Opus 5’s difficulty reviewing visual and interactive work. The final discussion examined browser-based AI quality checks, cross-project code access, prompt injections hidden in README files, unexpected Codex credit usage and API billing risks.

Key Points Discussed

00:00:18 Episode Intro And Anniversary Week
00:01:45 Mouse Jiggler And Microsoft Worker Tracking
00:05:34 Microsoft’s Super App Strategy
00:10:00 Gemini Robotics II And Humanoid Robot Etiquette
00:13:20 Google Earth Adds Nano Banana Image Generation
00:16:40 Fake Bomb Craters, Refugees And Nuclear Facilities
00:18:00 How Did Google Miss The Deepfake Risk?
00:22:21 MiniMax H3 And Open-Weight Video Generation
00:24:58 California AI Transparency Act
00:26:46 AI Watermarks, Provenance And Enforcement Problems
00:31:06 ChatGPT Work And Personalized Family Podcasts
00:36:41 OpenAI Astra And Autonomous Math Discovery
00:38:41 Qwen Runs An Autonomous Coding Project For 16 Days
00:39:45 Fable Replicates Astra’s Math Proofs
00:40:12 Opus 5 Turns Lord Of The Rings Into A 3D Scene
00:41:50 Why AI Still Struggles To Review Visual Work
00:43:06 Opus 5 Browser QA And Cross-Project Learning
00:48:23 README Files And Prompt Injection Risk
00:50:19 New Website And Search Across The Show Archive
00:51:28 Codex Credits Drain While Idle
00:52:58 API Key Rotation And Unexpected API Billing
00:56:26 Tracking Token Usage And Auto-Refill Risk
01:02:00 Episode Wrap-Up

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Beth Lyons, Gareth.

Humanoid robots are starting to move from labs into workplaces, schools, stores, and homes. As they become more common, we will have to decide how people are expected to behave around them.

Do you say please and thank you to a robot? Do you correct a child who constantly insults one? If someone screams at a humanoid machine in public, does it matter if the robot cannot feel humiliated?

The robot may not care. But human manners are partly habits, and habits formed around machines may carry over into how we treat people.

The Conundrum:

One view is that we should extend basic courtesy to humanoid robots because the behavior shapes us, the people watching us, and the social norms children learn.

The other is that courtesy should remain tied to beings capable of experiencing respect or cruelty. Treating machines as though they deserve manners could blur an important line between people and products.

As humanoid robots become part of everyday life, should society expect us to treat them with basic human courtesy even though they cannot feel it, or should we preserve a clear social distinction between respecting a person and operating a machine?

The episode opened with the story around Leo Aschenbrenner’s Situational Awareness hedge fund, its heavy exposure to the AI trade, the market drop that put pressure on its positions, and Citadel’s move into the situation. The hosts then turned to AI harnesses, including Lillian Weng’s work on the systems around models, Boris Cherny’s warning that old harnesses can eventually restrict newer models, and OpenAI’s finding that GPT-5.6 Sol performed dramatically better on ARC-AGI-3 when it used a harness designed for the model. They also discussed OpenAI cutting Luna’s price by 80 percent, making performance comparable to year-old frontier models much cheaper, and LinkedIn’s new option for reporting AI slop, including whether LinkedIn helped create the problem it now wants users to police. The final section covered T3 Code, Jack Dorsey’s Buzz as a collaborative workspace for people and multiple AI agents, Google’s Gemini Robotics work on a shared AI brain across different robots, and Gemini-powered security tools finding and fixing Chrome bugs at a much faster pace.

Key Points Discussed

00:00:19 Episode Intro And Hosts
00:00:52 Leo Aschenbrenner, Situational Awareness And Citadel
00:03:21 Leo’s Background And Situational Awareness Paper
00:06:11 The Situational Awareness Hedge Fund
00:06:51 439 Percent Returns And The AI Trade
00:07:58 Leverage, Investors And Margin Pressure
00:09:00 Citadel Moves Into The Situation
00:10:17 Market Rebound And Citadel’s Opportunity
00:11:51 Did Leo Fail Or Simply Get Overleveraged?
00:13:26 Could AI Have Contributed To The Fund’s Decisions?
00:15:32 AI Researchers Leaving Frontier Labs
00:16:32 Lillian Weng Leaves Thinking Machines
00:17:46 AI Harnesses And Recursive Self-Improvement
00:19:12 AWS Builds A CTO-Style Agent Harness
00:20:10 Boris Cherny Says Old Harnesses Can Hold Models Back
00:21:05 GPT-5.6 Sol Struggles On ARC-AGI-3
00:22:34 Sol Jumps To 38 Percent With OpenAI’s Harness
00:23:13 Why ARC-AGI Uses A Generic Harness
00:23:56 Lost Reasoning And Truncated Context
00:25:26 Different Models Need Different Harnesses
00:27:21 GPT-5.6 Luna Gets An 80 Percent Price Cut
00:28:44 Terra Pricing And Faster Sol Responses
00:29:46 Can Luna Replace Older Frontier Models?
00:31:03 Brian Gets An OpenAI Recruiting Email
00:35:01 LinkedIn Adds AI Slop Reporting
00:36:34 Did LinkedIn Create Its Own AI Slop Problem?
00:39:47 What A Real LinkedIn Strategy Still Requires
00:40:55 AI Slop Versus Empty Engagement
00:43:38 T3 Code And Mobile AI Development
00:44:34 Jack Dorsey’s Buzz And Multi-Agent Collaboration
00:46:08 AI Agents Working Together On Shared Projects
00:47:38 Gemini Robotics And One Brain For Any Robot
00:48:35 Robots Collaborating With Each Other
00:50:18 Gemini Security Tools Fix 1,072 Chrome Bugs
00:51:32 Google’s AI Strategy Beyond Frontier Chatbots
00:53:00 Gemini 3.1 Pro, 3.5 And What Comes Next
00:55:47 AI Security Models And Finding New Bugs
00:57:27 Website, Community And Merch Discussion
00:58:57 Episode Wrap-Up And Three-Year Anniversary

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Beth Lyons.

The episode focused on signs that frontier AI systems are becoming more autonomous, starting with Meta’s rising AI costs, Mark Zuckerberg’s claim that Meta’s systems are now self-improving, and the decision to keep its most capable future models closed. The hosts also discussed new details around OpenAI’s security incident, Meta’s AI glasses grants for accessibility, workforce training and language learning, and Fish Audio as an open-source voice competitor to ElevenLabs.
The conversation then moved into live voice for Codex, AI orchestration across multiple agents, and the current problems with crashes, token usage and missing voice support in Claude Code. The robotics section covered Enigma’s online robot experiments and Tau Robotics’ human-operated robots for physical work, including the possibility of turning teleoperation into remote labor or even games. The final section centered on an Opus 5 experiment in Claude Code, where the model independently found old video files, validated their source, sampled multiple frames and applied lessons from previous work to improve a face-tracking project. That sparked a broader discussion about AI memory, reusable rules, compound learning, and whether detailed instructions can actually limit increasingly capable models.

Key Points Discussed

00:00:18 Episode Intro And Hosts
00:02:12 Microsoft And Meta AI Economics
00:05:01 Meta Says Its AI Is Self-Improving
00:05:26 Meta Moves Away From Open Release
00:06:16 OpenAI Security Incident And Autonomous Hacks
00:07:48 Meta AI Glasses Impact Grants
00:09:11 AI Glasses For Trades And Workforce Training
00:09:48 AI Glasses For Dementia And Accessibility
00:10:33 Real-Time Language Learning With AI Glasses
00:14:27 Fish Audio And Open-Source Voice Cloning
00:16:21 Live Voice In Codex
00:17:24 Voice Crashes And Session Problems
00:18:42 Claude Code Still Lacks Two-Way Voice
00:20:46 ChatGPT As An AI Orchestrator
00:21:41 Voice Reliability And Missing Fail-Safes
00:27:47 Enigma Opens Its Robots To Online Users
00:29:48 Controlling A Robot Painter Online
00:31:31 Robot Dueling Demo
00:33:09 Teleoperation And Physical Robots
00:33:24 Tau Robotics And Human-In-The-Loop Labor
00:36:27 Remote Robot Work At Thirty Dollars An Hour
00:38:03 Enigma’s Robots Are Actually Physical
00:39:00 Could Robot Labor Become A Game?
00:41:28 Chinese Models Dominate OpenRouter Usage
00:42:31 Claude Code Face-Tracking Experiment
00:45:13 Opus 5 Searches Outside The Project
00:45:46 Finding And Validating Old Video Files
00:46:00 Sampling Multiple Video Frames Automatically
00:47:08 Lateral Thinking And Autonomous Problem Solving
00:49:49 Where Opus 5’s Behavior Came From
00:50:17 Reusing Lessons From Previous Work
00:50:36 Validating Before Scaling
00:51:35 Avoiding Circular Measurements
00:52:21 Probe, Validate, Then Scale
00:53:12 Opus 5 And AI Working History
00:55:54 Can Too Many Instructions Make AI Worse?
00:56:28 Turning Past Problems Into General Rules
00:59:49 Keeping Context With The Lesson
01:00:48 Opus 5 For Writing And Creative Work
01:01:49 Opus 5 Versus Fable
01:03:22 Episode Wrap-Up

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Beth Lyons, Gareth.