0:00 Microsoft just unveiled its first reasoning-focused foundation model, MAI-Thinking-1.0:06 In episode eighty-eight, we talked about AgentKit’s merger signaling the age of autonomous agents — today, that’s no longer a signal, it’s the landscape.0:17 Microsoft is building its own path, and it’s a direct challenge to its biggest partner, OpenAI.0:24 This isn’t just another model.0:27 It’s a strategic shift.0:28 Here’s what else is moving.0:30 While there were no new flagship models from Google, Meta, or OpenAI today, the action is happening one layer down.0:39 Anthropic just launched Claude Tag.0:41 It’s an agentic AI designed to live inside your Slack, autonomously writing code and analyzing data for your team.0:50 This is the agent-driven framework we discussed, now packaged as a product.0:55 At the same time, researchers just uncovered a critical flaw in the attention capabilities of even the most advanced AIs.1:04 The models are powerful, but they get confused on complex, long-horizon tasks.1:10 Think of it as a world-class sprinter who can’t run a marathon.1:15 That finding puts every new product launch into sharp perspective.1:20 Finally, Hollywood is bending the knee to OpenAI, and Superhuman just acquired the AI detection startup GPTZero.1:28 The board is being reset.1:30 Let’s go deeper on Microsoft.1:32 For years, the story was simple: Microsoft writes the checks, OpenAI builds the brains.1:39 That story just ended.1:40 The new MAI family of models, announced at the Build conference, is Microsoft’s declaration of independence.1:48 MAI-Thinking-1 is the key.1:50 It’s not a general-purpose chatbot.1:53 It’s designed specifically for long-horizon reasoning and planning.1:58 This is the exact capability needed to power the next wave of copilots and autonomous agents inside their own software, like PowerPoint and OneDrive, where it's already in preview.2:11 Why build it themselves?2:13 Two reasons.2:14 Cost and control.2:15 Running OpenAI’s biggest models is incredibly expensive.2:19 By building smaller, specialized models in-house, Microsoft can dramatically lower the cost of serving hundreds of millions of users.2:29 More importantly, it gives them control over their own destiny.2:34 They are no longer just a customer of the firm they invested billions in.2:39 They are now a direct competitor in the specialized model space.2:44 This isn't a pivot.2:46 It's a parallel path, a hedge against a future where OpenAI's priorities might not be Microsoft's.2:53 Now, here’s the turn.2:54 While Microsoft and Anthropic are shipping new agentic capabilities, a new research paper just put a ceiling on what those agents can do.3:05 Researchers gave top-tier AI models a classic psychological test of attention.3:10 The task was simple: read a list of words where the word itself is a color, but the text is printed in a different color.3:19 Like the word "BLUE" printed in red ink.3:22 The AI had to name the ink color.3:25 On short lists, the models were fine.3:28 But as the lists got longer and more complex, their performance fell off a cliff.3:34 They started making basic errors.3:36 They lost focus.3:37 This matters because the "attention" mechanism is the core engine of every modern large language model.3:45 It's what allows them to connect words and ideas across thousands of pages of text.3:51 This research shows that while attention is broad, it isn't deep.3:56 It degrades under cognitive load, just like a human's.4:00 It means that for all the talk of superhuman AI, the current architecture has a fundamental bottleneck.4:08 These models can write an email, but they can’t yet run a company.4:13 They can’t sustain the kind of complex, multi-step reasoning that true autonomy requires.4:19 So you have this split.4:21 On one hand, the product and engineering teams are racing ahead, shipping tools that push the current architecture to its absolute limit.4:31 On the other hand, the research community just found that limit's edge.4:36 The headlines today are full of new tools.4:39 The real story, the one that will define the next two years, is the race to solve this attention flaw.4:47 The company that cracks that won't just have a better model.4:52 They'll have a new kind of intelligence.
