Wednesday, September 30 September 30, 2026

Good morning. It's Wednesday, September 30th, 2026. I'm your MorningAI briefing, and today we've got a story that sounds like it came from a late-night comedy writer — but it's completely real.

Let's start at the top.

President Trump signed an executive order yesterday directing all federal agencies to stop using the phrase "artificial intelligence" and replace it with — and I'm not making this up — "Super Intelligence." Trump said the term AI sounds, quote, "fake," and that Super Intelligence better captures what's actually happening. The order affects all executive branch communications, official documents, and government-facing publications. Critics, including several congressional members, called it branding over policy. Supporters say it's a messaging win in the race against China. Either way, the federal government is now in the Super Intelligence business. Officially.

While Washington debates nomenclature, Delaware is making a much quieter but more consequential move. The state is pushing forward with a proposal to allow AI agents to legally run companies — as in, be the named operator of a corporate entity. Legal scholars are raising serious liability questions: if an AI agent goes rogue and breaches a contract or harms someone, who is responsible? The company that built it? The human who deployed it? Delaware says it's not putting the brakes on innovation. But this is the kind of regulatory frontier that will define how AI gets commercialized over the next decade. Every founder building with agents should be watching this closely.

On the model front, OpenAI quietly launched what's being called the Decisions API, running on GPT-6 Luna. It's designed for high-speed agentic decision-making — think responses in roughly 150 milliseconds. This is clearly positioned to compete with TypeSafe's Jev platform, which has built a fast following in the enterprise automation space. If the latency numbers hold up in production, this could shift how developers wire up real-time agent pipelines.

Over in hardware, a team at UT San Antonio is making noise with a chip called Genesis — designed to solve one of AI's most annoying problems: catastrophic forgetting. Current AI systems can learn new things, but often only by overwriting what they already know. Genesis takes a different architectural approach to let models accumulate knowledge incrementally without that regression. It's early-stage, but if it scales, it's a meaningful unlock for deploying long-lived AI agents in the real world.

Funding news: two deals worth calling out. EliseAI, the AI-powered property management platform, just closed a round that values it at four billion dollars — led by Andreessen Horowitz and Bessemer. This company automates leasing conversations, maintenance requests, and tenant communications for large multifamily operators. The valuation signals that vertical AI in real estate is hitting a serious growth phase. Second, Destro AI raised eight million in seed to build coordination software for heterogeneous warehouse robot fleets — different robots from different vendors, all talking to each other. Eight million sounds small, but the industrial robotics coordination problem is massive and almost entirely unsolved.

On the regulatory front, New York Governor Hochul convened her FutureWorks Commission to start mapping out policy recommendations on AI's impact on the workforce. Meanwhile, at a White House tech summit, Trump announced what he's calling a "morally binding" voluntary AI agreement with top tech leaders. No enforcement mechanism. Just moral pressure. Draw your own conclusions.

And one quick note from across the Atlantic: Google is appealing two EU directives requiring it to open parts of Android to rival AI services. This is part of a broader clash between Big Tech and EU competition law that's only going to escalate as AI becomes more embedded in mobile infrastructure.

Today's gem of an idea: with Delaware signaling that AI agents can operate companies, there's a real opening for a compliance and liability management platform built specifically for AI-run entities — a legal stack that tracks agent actions, creates audit trails, and helps human operators stay on the right side of liability. Think Stripe Atlas, but for deploying AI agents as legal business operators. The timing is now. The market need is clear. And nobody has built it yet.

That's your MorningAI briefing for Wednesday, September 30th. Stay sharp out there.