Body Scanners, Banned Models, and DeepMind's Space Game Obsession
TL;DR
Midjourney is building a body-scanner spa business, not just image models: David Holz unveiled a full-body ultrasound system built with Butterfly Network chips, said it is one of eight new Midjourney products, and talked about 50,000 machines plus roughly $40 billion to scale the vision.
The scanner pitch is 'MRI-like' convenience, not an MRI replacement: The demo showed a person lowered into water for a one-minute scan that can generate terabytes of data, but Ashley and Kylie stress that critics are right that ultrasound will not replace MRIs, CTs, or procedures like colonoscopies.
Anthropic's Fable ban looks like a precedent-setting power move: Ashley argues the government rushed the ban over jailbreak concerns, gave Anthropic about 90 minutes to respond, and then leaked aggressively to Axios and Politico, creating a dangerous model for punishing labs that do not align with federal preferences.
DeepMind's EVE Online deal is Demis Hassabis going back to his original AI thesis: By studying CCP Games, now Feneris, and its 20-year single-universe space economy offline, DeepMind gets a rich simulated world to test world models instead of relying only on brute-force scaling of LLMs.
OpenAI hiring Noam Shazeer is a serious architecture signal: The Transformer co-author and Character.AI founder leaving Google for OpenAI suggests the company wants fresh model ideas, not just more scale, especially after all the noise about scaling hitting a wall.
Snap's new AR glasses landed as an aesthetic and market mismatch: Kylie calls the $2,200 Specs huge and school-bully bait, while Ashley says glasses still have promise for calls and capture, just not in the form Evan Spiegel showed on CNBC.
The Breakdown
Midjourney's David Holz says he wants 50,000 body-scanning spa machines and maybe $40 billion to make daily full-body ultrasound as routine as an Apple Watch check, while Anthropic's Fable model got swept into what Ashley Vance calls the first real US government AI-model ban. The episode then swerves into DeepMind's new EVE Online obsession, a big OpenAI hire in Transformer co-author Noam Shazeer, and a brutal roast of Snap's enormous $2,200 glasses.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.