explainx / blog
Midjourney's full-body scanner announcement sparked equal parts fascination and skepticism. Radiologists warn about incidentalomas, privacy researchers flag missing data governance, and the prototype reality — 20 minutes, 12 people tested, no FDA clearance — is far from the 60-second headline claim.
Jun 18, 2026
500,000 sand-grain-sized transducers. 17 gigabytes of data per second. 40GB to reconstruct one cross-sectional slice. No radiation — just sound, water, and 60 seconds. Midjourney Medical announced their full-body ultrasonic scanner on June 18, 2026 with a 2027 spa opening in SF and a 2031 goal of a billion scans a month. Here is everything from the official announcement.
Jun 17, 2026
Nobody predicted a full-body medical scanner. Midjourney announced Midjourney Medical on June 18, 2026 — a new division building a whole-body ultrasonic imaging device with 8,960 transducers, no radiation, ~1 minute per scan, and a few dollars per session. Here is what was announced and what we got wrong in our pre-event speculation.
Jul 15, 2026
Mark Gurman's Bloomberg report describes OpenAI's debut hardware as a mobile, screen-free home computer pitched internally as a humanlike AI companion — not a HomePod clone. Mechanical elements move on their own, a camera reads the room, and GPT-Live powers conversation. explainx.ai maps timeline, privacy stakes, and the Apple trade-secret fight.
Nobody expected this. An AI image generation company — the one behind the fantasy portraits and sci-fi landscapes — announced a full-body medical scanner in June 2026, and the internet did not know whether to be impressed or alarmed.
If you want the official announcement without the editorial commentary, we covered it in detail here: Midjourney Medical: Full-Body Ultrasonic CT Scanner, 60 Seconds, No Radiation. That article covers everything David Holz said at the launch event. We also covered the hardware pivot as it was being speculated before the announcement was made.
This article is for the other conversation — the one happening on Reddit, Slashdot, X, and in radiology departments — about what Midjourney Medical actually is right now, not what it aims to be by 2031.
The announcement is built on dramatic numbers: 500,000 transducers, 17 gigabytes of data per second, 40GB per cross-sectional slice, and a 60-second full-body scan. No radiation. Cheaper than MRI.
The current prototype tells a different story.
As of June 2026, Midjourney Medical's scanner takes approximately 20 minutes per scan, not 60 seconds. Roughly 12 people have been scanned by the prototype. There is currently no AI in the imaging reconstruction pipeline — reconstruction is being done computationally, but without the neural-network layer that would eventually handle scale. The scanner has received no FDA clearance for diagnostic imaging.
David Holz has not hidden this gap. At the announcement, he described the device as research-stage. The 60-second figure is explicitly a target. But the distance between "prototype" and "60 seconds, 50,000 units worldwide by 2031" is not a small engineering delta — it is an extraordinary claim that has attracted proportionally extraordinary scrutiny.
The aesthetic framing of Midjourney Medical caught almost as much attention as the technology.
The Register's coverage landed on a phrase that spread widely: patients would be "bathed in golden light" as they descended into the scanning pool. The product is being positioned not as a clinical device but as an experience — the first Midjourney Spa in San Francisco, complete with hot tubs, cold plunges, and saunas, designed around ten full-body scanners.
This is deliberate. Holz argued that clinical environments create anxiety that skews health baselines, and that a spa setting produces calmer patients and better data. It is a coherent argument. It is also the kind of argument that makes radiologists and medical device regulators visibly uncomfortable.
The reaction across tech media ran a predictable spectrum. The Next Web and Gizmodo both noted the strangeness of an AI art company positioning itself as a medical device manufacturer with wellness branding. Slashdot threads ran heavy with skepticism, with commenters flagging that "fuzzy shapes" from an unvalidated scanner are not diagnostic and could actively mislead patients. Latent Space AINews, which monitors AI discourse across Reddit and Twitter, noted that the announcement generated intense discussion but that informed technical voices — radiologists, medical physicists, imaging researchers — were consistently more cautious than the general tech audience.
The medical imaging community's concern is not primarily about whether the technology works. The bigger problem is systemic: whole-body screening of healthy people.
The American College of Radiology (ACR) has published clear guidance that whole-body CT scanning of asymptomatic healthy individuals is not recommended. The reason is a phenomenon called the incidentaloma — an incidental finding that was never meant to be discovered.
When you scan a healthy person from head to toe, you will almost certainly find something. A small nodule. A shadow. An ambiguous density. Statistically, most of these findings are benign and would never cause harm. But once found, they cannot be unfound. The patient now has an anomaly in their chart. That anomaly requires follow-up: more imaging, sometimes biopsies, sometimes surgery. The follow-up procedures carry their own risks. The patient experiences anxiety throughout this process. And the final finding is usually: this was nothing.
This is not speculation. It is a well-documented clinical pattern from the history of CT colonography, pulmonary CT screening, and whole-body MRI programs. Midjourney's pitch — scan everyone, every month, catch things early — runs directly into this wall.
Holz acknowledged in the announcement that the company is aware of the "flagging weird things" problem. He said they intend to build diagnostic capabilities incrementally, submitting each new capability to the FDA separately, and that they are starting with body composition maps specifically because those are the lowest-risk output. That is a reasonable regulatory strategy. But it does not address the fundamental question: what happens when a user's composition map shows something ambiguous, and they go looking for answers?
The performance figures Midjourney released — "10x cheaper than MRI," "60x faster" — are Midjourney's own projections. They have not been independently validated. No peer-reviewed paper has been published on the USCT system's accuracy compared to MRI or CT. No radiologist outside Midjourney has evaluated the image quality against a clinical standard.
This matters because the comparison is being made to established diagnostic gold standards. A full-body MRI at a major research hospital produces images that trained radiologists have spent careers learning to read. Midjourney's USCT produces a different kind of data — acoustic reflection patterns reconstructed into 3D maps — and the clinical meaning of those images has not been established.
David Galbraith, writing on X, asked whether the Midjourney scanner hype was real and whether there had been rigorous independent analysis. The honest answer in June 2026 is: no. There has been no truly rigorous, independent analysis. The technology is real; the claims are aspirational.
Butterfly Network, whose sensor array technology Midjourney licensed for $15 million, has themselves been measured in their statements about full-body ultrasonic CT capabilities. Butterfly has been building portable ultrasound devices for years and has consistently framed USCT as promising but requiring significant development for full-body diagnostic use.
The Midjourney Medical announcement said almost nothing about data governance.
A full-body scan is among the most sensitive biometric data it is possible to produce. It reveals bone density, organ size and placement, body composition, potential health conditions, and physical anomalies. For a company to collect this data and store it in the cloud without specifying retention policies, access controls, training data use, or deletion rights is — in the view of privacy researchers who reviewed the announcement — a serious gap.
Medical device manufacturers are subject to HIPAA in the United States. Midjourney has not yet classified itself as a covered entity or business associate under HIPAA because the scanner is not currently performing diagnostic functions. The body composition maps it will initially produce may fall outside HIPAA scope. That legal grey zone is precisely where the privacy concern lives.
Users who scan at the first Midjourney Spa in late 2027 will need to understand what they are agreeing to when they sign the consent form — and that consent form has not been published.
None of this means Midjourney Medical will fail. The underlying technology — ultrasonic computed tomography — is real and based on established Caltech research. The Butterfly Network hardware license is a serious investment. David Holz's track record of executing ambitious technical bets from the image generation era gives the project credibility that most hardware startups do not have.
The 2031 goals — 50,000 scanners, one billion scans per month — may be as speculative as most tech roadmaps. But even a scaled-back version of this vision, producing validated body composition data at lower cost than existing modalities, would be meaningful.
The skepticism is not that the technology is impossible. It is that the gap between the current prototype and the announced vision is being communicated as smaller than it is. Twenty minutes is not sixty seconds. Twelve subjects is not a clinical trial. No FDA clearance is not a technicality.
The internet is not wrong to find this weird. An AI image company pivoting to medical spas and full-body scanners is genuinely strange. The golden light framing, the wellness aesthetics, the 2031 moonshot — all of it is designed to generate the same response that Midjourney's image generation models generated in 2022: awe first, questions later.
The questions are now arriving.
Not every credible voice landed on the skeptical side. On June 23, 2026, Bryan Johnson — the entrepreneur behind the Blueprint longevity protocol and founder of Braintree (sold to PayPal for $800M) — published a substantial counter-argument to the incidentaloma-centered criticism that has dominated expert coverage.
His framing is deliberately reductive: experts are asking the wrong question.
"The wrong question: 'Is it as good as MRI?' / 'What about false positives?'" "The right question: 'What does fast, low cost, safe whole body imaging unlock?'"
This is not a dismissal of the ACR's concerns. It is a refusal to accept the implied baseline. Johnson's argument: the comparison is not between "Midjourney scanner with incidentalomas" and "safe, fully-monitored population health." The comparison is between "Midjourney scanner" and "bodies that are currently among the least-measured objects in modern life."
Your phone generates more continuous data points about your location than your doctor has ever generated about your liver. Your calendar tracks every appointment. Your bank tracks every purchase. And your organs — the things that will actually kill you — accumulate silent, undetected changes until they become symptomatic, which is often already late.
The harm is already baked into the current system. People walk around with undiagnosed fatty liver disease. Endocrine dysfunction goes undetected for years. Soft tissue changes become symptoms before they become data. From this starting point, the incidentaloma argument looks different: finding something ambiguous is not a failure mode — it is the beginning of a longitudinal record that didn't exist before.
Johnson's three specific use cases where fast, affordable, repeatable whole-body imaging changes the equation:
| Use Case | Why Repeatability Changes It |
|---|---|
| Fatty liver / metabolic health | Non-invasive, no contrast agent, can track weekly — currently requires biopsy or expensive MRI to diagnose and monitor |
| Endocrine tissue (thyroid, PCOS, hypogonadism) | Soft tissue conditions that are chronically underdiagnosed, especially in women, due to cost and access barriers to current imaging |
| Soft tissue + multiomics | Combining imaging with proteomic aging clocks creates integrated body state tracking that single-timepoint MRI cannot provide |
The biggest unlock, in Johnson's view, is not the single scan — it is the trajectory.
A composition map tells you what's there. A series of maps — weekly, monthly — tells you whether it's changing, how fast, and in which direction. That longitudinal signal is more diagnostically useful than any single reading, and the current system cannot produce it at scale for healthy, asymptomatic people.
"With trajectory data: 'is this changing?' vs 'what is this?' — trajectory is more informative than single timepoint."
This is the analytical frame that makes Johnson's position coherent to people who find the single-scan incidentaloma argument compelling. He's not disputing that a single ambiguous finding creates follow-up burden. He's arguing that the single scan is the wrong unit. The value is in the series.
Johnson also acknowledged the technology's genuine blind spots without hedging: air-filled lungs and the brain behind the skull are real imaging limits for ultrasound. This is not a technology that replaces pulmonary CT or neurological MRI. His use cases are specifically the metabolic and endocrine organs where ultrasonic soft tissue imaging is genuinely competitive — and where current monitoring is cost-prohibitive at population scale.
He personally plans to add weekly whole-body scans to his own data stack — which already includes over 1.5 billion biological data points — and endorsed David Holz as "doing this for the right reasons" and as well-positioned financially to see it through.
The Johnson framework doesn't answer the ACR's concerns directly. It shifts the burden of proof. If you accept that the status quo has harm baked into it — silent, undetected deterioration at population scale — then the question changes from "could this scanner cause harm?" to "does this scanner's harm profile exceed the baseline harm of not knowing?" That is a much harder question, and radiologists haven't been answering it, because no technology has made it answerable before.
Whether the scanner delivers the image quality to make that calculus real is what the next twelve months of research trials will determine. But Johnson represents the most serious quantified-self and longevity research case for exactly what Midjourney Medical is attempting — and it is a different class of argument than "interesting technology, probably overhyped."
For Midjourney Medical to become what it claims to be, several things need to happen that have nothing to do with transducer counts:
Independent clinical validation. Images produced by the USCT system need to be evaluated against MRI and CT by radiologists who are not affiliated with Midjourney. The sensitivity and specificity for various findings need to be published in peer-reviewed literature.
FDA clearance for diagnostic use. The body composition maps planned for the first Spa launch are not diagnostics. Meaningful health monitoring requires diagnostic classification, which requires FDA clearance, which requires clinical trials.
Data governance disclosure. Before the first scan is performed on a paying customer, Midjourney needs to publish clear policies on data retention, access, training use, and deletion. This is not optional in a HIPAA-adjacent space.
Scan time reduction. The prototype's 20-minute scan time needs to approach the stated 60-second target before the volume economics (one billion scans per month) become plausible. This is a significant engineering problem.
Protocol for incidental findings. The company needs a clear, medically supervised protocol for what happens when a scan shows something anomalous. "We'll figure it out" is not sufficient when the anomaly is someone's health.
Midjourney has a history of shipping things that seemed impossible and then iterating fast. Maybe that history repeats here. But this time, the stakes are not a blurry image. They are someone's health data and medical decision-making.
The golden light is beautiful. The questions underneath it are harder.
For the full official announcement details, see our coverage of Midjourney Medical's June 2026 launch event. For context on how this pivot was anticipated, see our piece on Midjourney's first hardware announcement.
