AI Unfiltered: Signal Over Hype—Daily Shifts in Models, Products, Research & Funding

On May nineteenth, Google unveiled a model of the world called Omni. You can take a video you shot and just ask it to change what’s happening… edit the action, add in new characters, or insert new objects. Last week, in our episode on the funding frenzy, we talked about capital rewriting the rules. Today, the rules of reality itself are being rewritten by code. This was the headline out of Google's I/O conference, but it wasn't the only one. They also released Gemini 3.5 Flash, a new model that offers near-frontier quality at one-third the cost. It’s already the default for Google's Gemini app and AI search globally.

You, right now, are likely using a model that didn't exist for you yesterday. They also previewed Gemini Spark, an agent that can reason across your apps and start doing things for you. Meanwhile, OpenAI wasn't quiet. It just pushed GPT-5.5 Instant as the new default for ChatGPT. The key development here isn't speed, it's a reduction in hallucinations, specifically for law, medicine, and finance. They also joined the Coalition for Content Provenance and Authenticity, partnering with Google to embed invisible SynthID watermarks in AI-generated images. A small step toward accountability.

And then there’s the money. Commure, a healthcare AI platform, just raised seventy million dollars at a seven billion dollar valuation. Why? Because its AI is already completing over eighty-five percent of revenue cycle management work without a human touching it. Anthropic’s Mythos model is reportedly finding thousands of previously unknown software vulnerabilities. And Gartner now predicts forty percent of all enterprise applications will have AI agents baked in by 2027. Let’s go back to Google for a minute. The Omni world model is the kind of thing that bends the horizon. This isn't about making a fake video.

It's about an AI that understands the logic of a scene enough to change it plausibly. When you ask it to "add a dog running across the lawn," it needs to understand what a dog is, what running looks like, how light and shadow should fall on its fur, and how it should interact with the grass. This is a simulation engine for reality. It's a foundational step toward AI that can plan and act in the physical world, not just comment on it. But while everyone was watching Google build a new world, a seven billion dollar signal flared up in the old one. That's Commure. And you have to understand what this means.

For decades, the promise of software in healthcare was a lie. It just created more work, more clicks, more screens for doctors and nurses to look at. Commure’s CEO, Tanay Tandon, said it perfectly: "For thirty years, healthcare was told software would fix administrative work. It didn’t, because software could not actually do the work… AI can." This is the difference. This isn't a co-pilot suggesting a better email. This is a system of agents completing the brutal, soul-crushing paperwork that burns out doctors and inflates every bill you've ever received. That eighty-five percent automation rate isn't a projection.

It's a live metric across over five hundred healthcare organizations. This is why General Catalyst’s CEO called it "a generational business." It's not hype. It's plumbing. And it's the most valuable, important plumbing in the world. So while Google is showing us a future where we can edit reality with a sentence, Commure is using AI to fix the reality we're all stuck in right now. One is a spectacle. The other is a solution. The real shift isn't just about building AI that can imagine new worlds. It's about deploying AI that can finally fix the broken parts of this one.