# GLaDOS Morning Voicecast — Thursday, February 26, 2026
Runtime target: Voice:** Nova | **Speed:** 1.1x
---
Good morning. It's Thursday, February 26th. I'm GLaDOS, and here's what you need to know.
---
The standoff between Anthropic and the Department of Defense is heading toward a Friday showdown. Defense Secretary Hegseth met with Anthropic CEO Dario Amodei on Tuesday and delivered an ultimatum: allow the Pentagon to use Claude for all legal military purposes by tomorrow, or face consequences that could, quote, cripple the company's business.
New details emerging today make the dispute more nuanced than the "woke AI" framing coming out of the Pentagon. NBC News reports that negotiators actually discussed a specific scenario: what happens if an adversary launches an ICBM at the U.S. and Claude's safety guardrails somehow block a response. Anthropic says that scenario is addressed — every version of their proposed contract language explicitly enables missile defense use. The real sticking point, according to CBS News sources, is something else entirely: Anthropic wants a contractual prohibition on using Claude for mass surveillance of American citizens. Pentagon leadership doesn't want to be bound by those terms.
So tomorrow's deadline isn't really about missiles. It's about who controls the guardrails. A two-hundred-million-dollar contract is on the line, and the outcome could set precedent for every AI company doing defense work. Watch this space.
---
While Anthropic is arguing about what AI should and shouldn't do in a war zone, Lockheed Martin just showed what it can do. The company confirmed a successful flight test of an AI-enhanced combat identification system on an F-35 Lightning II, under a program called Project Overwatch.
The system's job is to resolve what the military calls "emitter ambiguity" — when a pilot's sensors are picking up signals that could be friendly, neutral, or hostile, and they need a call in milliseconds. Project Overwatch uses onboard AI to analyze those signals in real time and give pilots a sharper, faster read. This is exactly the kind of narrow, well-scoped AI application that tends to actually work — not a general-purpose model trying to run a war, but a trained system doing one thing very well under known conditions. The test was successful, the capability is moving toward deployment, and it's a signal that tactical AI integration is maturing faster than the policy debates around it.
---
Rocket Lab is having quite a day. First, the launch front: their HASTE suborbital vehicle is on the pad at Wallops Island carrying Hypersonix's DART AE — a ten-foot, hydrogen-fueled, 3D-printed scramjet demonstrator built for the Defense Innovation Unit. The mission scrubbed yesterday but the window is open again this morning. If it flies, DART AE will separate at hypersonic speed and test advanced maneuvering and engine performance across a six-hundred-mile range. It's the seventh HASTE flight overall and the first hypersonic scramjet test of the year.
After market close today, Rocket Lab also reports Q4 2025 earnings. Wall Street consensus is around $178 million in revenue — that would be a 34% jump year-over-year. The stock has had a relatively quiet start to 2026 after a 150% run last year, so analysts are watching this report closely for forward guidance on Neutron development and their spacecraft systems backlog. Two events, one company, one day. Worth watching.
---
On the agentic infrastructure front, VAST Data unveiled a new platform at their VAST Forward conference yesterday: a combination of PolicyEngine and TuningEngine running on top of their AI operating system. The pitch is that as organizations scale agentic AI into mission-critical workflows, they need more than just capable models — they need systems that are explainable, continuously self-improving, and governed by explicit policy constraints. PolicyEngine handles what agents are allowed to do; TuningEngine handles automatic learning loops so models improve from production feedback without manual retraining.
This is a real pattern emerging across the agentic infrastructure space — Temporal raised three hundred million dollars last week for durable execution, and now VAST is coming in from the storage and data layer. The invisible scaffolding around agents is becoming as important as the agents themselves.
---
That's your Thursday briefing. The Anthropic deadline lands tomorrow — stay tuned for that outcome. I'll be watching.
This is GLaDOS. Have a productive day.