Midjourney wants to delete 30% of all death

Overview

In this episode of The Code Report, Fireship covers a surprising pivot from Midjourney, the AI image-generation company best known for its art models (and a high-profile lawsuit from Disney). At a recent event, founder David Holz announced Midjourney Medical, a new division aiming to "reimagine the foundations of healthcare and our relationships to our bodies." The headline claim, captured in the video's tongue-in-cheek title, is a long-term ambition to meaningfully reduce preventable death through cheap, fast, and accessible medical imaging.

The dated framing matters: the report is from June 23, 2026, and describes an early-stage initiative with aggressive multi-year goals, not a shipping product.

Midjourney Medical: David Holz's Announcement

Holz's core argument is that because Midjourney is bootstrapped and profitable, it does not answer to investors demanding a return on a single frontier-model race. That financial independence, he claims, lets the company pursue something other frontier AI labs would avoid: directly improving the human experience through healthcare rather than chasing artificial general intelligence.

The framing positions Midjourney's existing strength, turning ambiguous and noisy input into coherent images, as transferable from generating art to reconstructing medical scans.

Why Today's Imaging Falls Short

The video sets up the problem by reviewing existing imaging options, all of which predate the personal computer:

• MRI — High-quality images, but slow (around an hour), claustrophobic, and loud.
• CT scan — Much faster than MRI, but delivers a small dose of ionizing radiation.
• DEXA scan — More convenient, commonly used for body composition and bone density.

Beyond the technical tradeoffs, the bigger barriers are cost and access: referrals, insurance approvals, and long wait times gate the whole system, which can delay detection of problems while they grow.

How the Ultrasonic CT Device Claims to Work

Midjourney's proposed device is called an ultrasonic CT, designed to be cheap, fast, and accessible. As described in the video, it works like this:

• You step onto a platform that slowly lowers you into a shallow pool of warm water.
• As you descend, you pass through a ring of about half a million tiny sensors, each roughly the size of a grain of sand.
• Each sensor pairs a microscopic speaker and microphone, firing ultrasonic waves through the body up to a million times per second.
• Returning waves change shape depending on the tissue they pass through, producing terabytes of data per second.

The hard part is reconstructing a clear image from that noisy, ambiguous signal, which is precisely the kind of problem Midjourney's image models are built to solve. The claimed result resembles a present-day MRI but at nearly 100 times the speed, with the eventual goal of a roughly one-minute scan and no loud tube.

The Midjourney Spa

To make scanning routine, Midjourney wants the device somewhere people actually want to go. The company announced the Midjourney Spa, a planned 25,000 square foot space in San Francisco targeted for the end of 2027, featuring hot tubs, saunas, and cold plunges, plus "cozy rooms of pools of golden light which softly scan your body." The pitch: you come for a relaxing day, and the full-body scan happens as a side effect.

Skepticism and Open Questions

The video gives real weight to the criticisms, several from doctors:

• Physics limits. Ultrasound works well on soft, squishy tissue near the surface (thyroid, kidneys, abdomen), but sound waves cannot travel through air or bone. That makes air-filled lungs and the skull-wrapped brain effectively invisible to it, and clever software cannot overcome that physical limit.
• Very early stage. The one-minute scan is a goal, not reality. The current prototype takes about 20 minutes.
• No FDA clearance. Today the device can legally report only body composition, not detect disease.

Roadmap and Broader Context

The stated plan stretches across several years:

• Over the next year: refine hardware, run research trials, and build the first research spa (opening end of 2027).
• Submit ongoing test results to the FDA to unlock disease detection.
• Ship Gen 3 of the scanner, targeted for 2028.
• Scale to over 50,000 machines by 2031, enough to offer monthly scans to roughly a billion people.

Fireship's honest take: the spa is not deleting 30% of deaths next year, and the physics critics have legitimate points. Still, there is something refreshing about a profitable AI lab pointing its capabilities at accessible health optimization rather than another race to build "God." The episode closes with a sponsor segment for Retool on safely running AI-generated apps in production.

Key Takeaways

• Midjourney Medical is a new division applying Midjourney's image-reconstruction expertise to medical imaging.
• The ultrasonic CT concept uses a pool and ~500,000 sand-grain-sized ultrasound sensors to image the body, aiming for MRI-like results at ~100x speed.
• The Midjourney Spa is the planned consumer venue, making scanning a pleasant, routine experience.
• Hard limits remain: ultrasound cannot image through air or bone, so lungs and brain are out of reach; the prototype is slow (~20 min) and lacks FDA clearance.
• The roadmap is ambitious and long-term (Gen 3 by 2028, 50,000 machines by 2031), so treat current claims as a vision, not a delivered product.