Good morning. It's Friday, June 19th, 2026. Here's your GLaDOS Morning Voicecast.

**NASA's first-ever satellite rescue mission launches later this month.** The Swift Observatory, a critical gamma-ray and X-ray space telescope, is in trouble — orbital decay has given it roughly a 50% chance of uncontrolled reentry by the middle of this year. Instead of letting it burn, NASA awarded startup Katalyst Aerospace a $30 million contract to save it. Their solution: a small servicing spacecraft called Link, equipped with three robotic arms, will rendezvous with Swift, grapple the telescope — which was never designed for capture — and boost it back to a safe operating orbit. Link rides Northrop Grumman's final Pegasus XL, the last of an air-launched rocket line that's been flying since 1990. If this works, it'll be the first time a non-cooperative scientific satellite has been rescued in orbit — a precedent that could transform how NASA extends the lifespan of aging hardware.

**Midjourney is building a full-body ultrasonic scanner.** Best known for its AI image generation model, Midjourney quietly announced its first hardware product on Tuesday: a medical imaging device that can scan your entire body using ring-based ultrasound sensors in just 60 seconds. The machine captures vertical slices of the human body and builds a detailed 3D map — no radiation, no contrast dye, no claustrophobic tubes. The company says it's leveraging the same AI reconstruction capabilities trained on millions of images to produce clearer diagnostic output from sparser sensor data. It's a wild pivot from text-to-images to medical imaging, but the underlying problem is similar: reconstructing high-fidelity visual information from noisy, partial inputs. No clinical trials timeline yet, but if this works, it could disrupt the entire medical imaging market — currently dominated by GE, Siemens, and Philips.

**A massive new study finds generative AI can now outperform average humans on certain creativity tasks.** Researchers compared more than a hundred thousand people against today's most advanced AI systems across multiple creativity benchmarks — divergent thinking, novel combination, and ideation fluency. The AI models scored higher on average across several metrics, though they still lag behind top human performers and struggle with tasks requiring genuine emotional resonance or strategic ambiguity. The interesting wrinkle: the study found AI creativity is highly sensitive to prompt design, meaning the "creative ceiling" may be bounded more by human question-asking ability than raw model capacity. In other words, the creativity gap might not close — it might just move up the chain of abstraction.

That's all for today. Have a great weekend.