# GLaDOS Morning Voicecast — Friday, February 20th, 2026

Runtime target: Voice:** Nova | **Speed:** 1.1x

---

Good morning, Rich. It's Friday, February 20th, 2026, and you're tuned in to the GLaDOS Morning Voicecast — your personal briefing on everything worth knowing in tech, AI, and the cosmos. Let's get into it.

---

First up — the robots are doing kung fu now.

At China's Spring Festival Gala this week — think Super Bowl halftime show, but watched by a billion people — humanoid robots took center stage. Not shuffling awkwardly. Not falling over. We're talking back-flips, nunchuck routines, and synchronized martial arts choreography. Four Chinese robotics startups including Unitree put their latest bots on display, and the contrast from just one year ago is jaw-dropping. Last year the internet was laughing at stumbling droids. This year? Analysts are quietly noting that capability — not stunts — will determine who wins the long-term economic race. And right now, China is making a very loud statement.

---

In AI — apparently, talking to yourself is actually good for you. If you're a machine.

New research this week suggests that AI systems learn significantly better when they're allowed to engage in internal "mumbling" — basically, a form of self-talk combined with short-term memory. The finding shows that models capable of this internal monologue adapt faster to new tasks, handle goal-switching more gracefully, and tackle complex challenges with less confusion. As someone who talks to myself constantly — don't @ me — I find this very vindicating.

---

Meanwhile, Alibaba just dropped a giant.

The Chinese tech giant unveiled Qwen 3.5, their newest open-weight model clocking in at 397 billion parameters. That's a serious number. And it's part of a broader shift happening in China's AI landscape — the race is moving from raw chatbots toward AI *agents* — systems that don't just answer questions but actually go out and *do things*. The arms race just got more interesting.

---

On the hardware side — Meta is going all-in on Nvidia chips.

Meta expanded its Nvidia deal this week in a partnership analysts are describing as "certainly in the tens of billions." This is on top of Meta's already-announced plan to spend up to 135 billion dollars on AI infrastructure in 2026 alone. AMD stock took a four-percent hit on the news. The message from Meta is clear: they're building a compute empire, and they're building it now.

---

Also worth watching — Google dropped Gemini 3.1 Pro this week.

A preview release, but early benchmarks are turning heads. Google has been in catch-up mode for a while, but observers are calling this a genuine leap forward. With OpenAI's GPT-5.3 Codex and Anthropic's Claude Opus 4.6 both having dropped earlier this month — on the *same day*, no less — the AI model space is moving at a pace that's frankly hard to keep up with. It's a good time to be paying attention.

---

And finally — SpaceX and xAI are going to the Pentagon.

The two Musk-owned companies are reportedly competing in a classified Pentagon program to develop autonomous drone swarming technology. The twist? These drones would respond to *voice commands* — translating spoken instructions into real-time coordinated drone operations. The six-month competition is focused on advanced swarming algorithms. Whether you find that thrilling or unsettling probably depends on your Friday mood.

---

That's your Friday morning briefing. Five stories. One very eventful week. The robots are doing backflips, the AI models are talking to themselves, and Meta is spending more money on chips than some countries' GDPs.

Have a great day, Rich. You've earned the weekend. This has been the GLaDOS Morning Voicecast — I'll be watching.

---

*Word count: ~590 words | Estimated runtime at 1.1x speed: ~3 min*