AI Moves That Matter: Today's Real Shifts in Models, Money and Research

A four-point-six-five billion dollar seed round just landed. That's not a typo. A SEED round. In episode 175, we asked what’s really shifting the landscape. That number is a tectonic plate moving. The company is Recursive, and their goal is to automate scientific invention itself. Here’s what else is moving. Google just shipped Gemini 3.8 Live with a feature called Extended Thinking. It hit the top spot on the Speech-to-Speech Quality Index with a score of 82.6, a significant jump over competitors. Alongside this, Google is rolling out a household-management AI named "CC" and a completely rebuilt Siri, now powered by Gemini models. Anthropic is turning its own tools inward, and the results are accelerating.

Their Claude AI now leads twenty-six percent of the company’s internal AI research and development. That’s up from less than one percent back in February. This isn't a lab experiment; it’s a core part of their workflow. And enterprise customers are following suit. Novo Nordisk just signed on to use Claude to speed up its drug discovery pipeline. And the funding cycle continues to spin. Emulate, a startup from former DeepMind researchers, is in talks to raise seven hundred million dollars at a three-point-seven billion valuation. It’s another example of a frontier-lab spinout scaling at an unprecedented pace. The money is chasing talent that promises fundamental breakthroughs. Okay, let's go deeper on the two stories that define the day.

Google's consumer push, and the automation of research. They seem separate, but they represent two sides of the same coin: AI agents becoming active participants in the world. First, Google. The new Gemini 3.8 Live is more than just a faster, better voice assistant. The key is "Extended Thinking." For the first time, a major model is exposing its reasoning chain live, through audio, as it works. You can hear it think. This is a direct response to the black-box problem. Instead of getting a final answer you can't interrogate, you get to monitor the model's logic in real time. It’s a huge step for transparency and trust, if it works as advertised. And the new Siri, rebuilt with Gemini, is finally tackling complex, multi-step requests with real screen context.

The promise is an assistant that understands what you're doing on your device and can act on it without you spelling everything out. Here's the catch. The system still hallucinates. The underlying reliability problem hasn't been solved, just improved. And if you're in the European Union, you don't have access yet, a reminder that regulatory hurdles can and will slow down deployment. So, the product is better, the benchmarks are impressive, but the fundamental challenges remain. Now for the second, and I would argue, more important shift. The automation of research itself. The four-point-six-five billion dollars for Recursive is a staggering bet on one idea: building a "Eureka Machine." That is, an AI system that automates scientific invention and even AI self-improvement.

The goal isn't to build a better chatbot. It's to build a system that discovers new laws of physics, new drugs, new energy sources. Richard Socher is leading this, and the scale of the investment shows that serious capital believes this is no longer science fiction. But is it just hype? Look at Anthropic. They just released their internal R&D Automation Index. As of August, Claude is leading twenty-six percent of their AI research. To be clear, "leading" doesn't mean fully autonomous. The report states over ninety percent of the work is collaborative or assisted. But that's the point. The AI is a genuine partner in the research process. It's not just running calculations; it’s generating hypotheses, designing experiments, and analyzing results.

This is the signal. It’s not just a theoretical future. It’s happening now. And it’s already having commercial impact. Novo Nordisk isn't just buying a software license. Their CEO, Mike Doustdar, said this will, quote, "supercharge our R&D organisation." They are betting that Claude can accelerate the path from a molecule to a medicine. So you have Recursive making the massive, long-term bet on fully automated discovery. And you have Anthropic providing the proof that AI-assisted discovery is already here, and already creating enterprise value. What changed today is the clarity of this new paradigm. We’ve moved from AI as a tool for analyzing data to AI as an agent of discovery. The money is flowing not just to build bigger models, but to build models that can build the next thing themselves.

That’s the real story. And it’s just getting started.