A behind-the-scenes look at Midjourney’s medical scanner leaves many questions unanswered

Midjourney’s Medical Scanner: More Questions Than Answers for Developers

Last week, Midjourney released a nearly 20-minute behind-the-scenes video of its much-hyped “dunk-tank” ultrasound scanner. The tour, hosted by tech YouTuber and Midjourney engineer Marcin Plaza, offers the most detailed look yet at the hardware the AI startup hopes will “transform medicine” with cheap, radiation-free imaging. But for developers and builders watching closely, the video raises far more questions than it answers — and the gaps are instructive for anyone building at the intersection of AI and physical products.

The scanner itself is described frankly by Plaza as scores of ultrasound probes “hacked apart and slapped on a glorified hot tub with an elevator in it,” connected to off-the-shelf computers and Raspberry Pis. The team showed some assembly work and a scan of an imaging phantom (a standardized test object used to validate ultrasound performance). Yet the video largely glosses over the fundamental physics and imaging challenges that independent experts flagged when Midjourney first announced the scanner. Those experts told The Verge that Midjourney had shown little evidence it could overcome the well-known limits of ultrasound — a technology that has been refined for decades — or produce the kind of detailed, high-resolution images the company has suggested, at the speed and scale it promises.

Midjourney has repeatedly emphasized that the scanner will launch as a wellness product focused on body composition, rather than as a diagnostic medical device. That distinction is critical: wellness products do not require FDA clearance or clinical trials. In the video, head of medical Tom Calloway doubled down on this approach, saying the focus on body composition would let the company “speedrun” and open immediately once testing is complete. Yet the video’s tone and language still lean heavily on medical aspirations, asking what physicians could do with frequent scans taken over time. Calloway did not seem concerned about the apparent contradiction: “I don’t think there’s anything to really clarify,” he said, promising frequent blog posts with progress updates. CEO David Holz, meanwhile, explained that Midjourney’s lack of outside investors gives the company freedom to pursue the scanner. “No one can tell me not to do it,” he said.

The Developer’s Take: Why the Details Matter

For developers who build with or around AI tools, this story is a case study in the gap between ambitious claims and operational reality. Midjourney is best known for its image generation platform — a leader in the space that we cover extensively in our comparison of the best AI image generators — and it has also made strides in video generation, as outlined in our guide to the best AI video generators. The company’s pivot from software to hardware, and specifically to a medical-adjacent physical product, is a radical departure. It highlights several lessons for anyone considering similar moves.

First, the importance of validation at the physics level. Ultrasound imaging is not a new frontier. The technology has well-documented limitations in resolution, depth penetration, and artifact handling. Simply stacking more probes and applying AI post-processing does not automatically solve those constraints. Midjourney has not published peer-reviewed data, shown head-to-head comparisons with existing ultrasound systems, or released raw scan results that independent experts can evaluate. For developers used to iterating fast in software, this is a sobering reminder that hardware products — especially those involving diagnostic or near-diagnostic claims — require rigorous, transparent validation before they can be taken seriously by the technical community.

Second, the regulatory dodge is a red flag. Positioning a device as a “wellness” product to avoid FDA oversight is a well-worn path in consumer health tech. It works for things like smart scales and sleep trackers. But when a company uses medical language — showing physicians, discussing diagnostic potential, implying clinical utility — while explicitly avoiding the regulatory burden of proving safety and efficacy, it creates a trust deficit. Developers evaluating partnerships or integrations with such a platform should scrutinize whether the claims match the evidence. If the product cannot pass muster in a clinical trial, can it really deliver the “transformative” results it pitches?

Third, the “no investors” argument cuts both ways. CEO Holz correctly notes that without outside investors, Midjourney can pursue moonshots without short-term pressure. That freedom is valuable for blue-sky R&D. But it also means there is no external accountability mechanism — no board, no investors, no regulators demanding evidence before production. In a field like medical imaging, where mistakes can have serious consequences, internal checks and balances alone may not be sufficient. Developers looking to build on top of such a platform should demand transparency and independent review, not just company blogs.

What This Changes for Builders

For the developer community, the most immediate takeaway is to treat Midjourney’s scanner with healthy skepticism until independently verified data emerges. The video shows a working prototype — fans spinning, water circulating, a phantom being scanned — but it does not show the kind of clear, detailed imagery that would convince radiologists or engineers. The company has promised frequent progress updates, but developers should withhold judgment until raw scan outputs are made available for public or expert review.

More broadly, this episode underscores a pattern we are seeing across the AI hardware space: companies with strong software reputations rushing to build physical products without fully accounting for the different validation standards those products require. Software can be patched; hardware can be recalled. The stakes are higher, and the cycle is slower. Developers who are building applications that could depend on such devices — for example, wellness apps that interpret body composition data, or research tools that rely on frequent cheap scans — should factor this uncertainty into their roadmaps.

There is also an ethical dimension. If the scanner ultimately ships and consumers use it for wellness insights, but the underlying imaging quality is poor, users could make misinformed health decisions. Midjourney’s team may be well-intentioned, but intention does not replace evidence. The company would benefit from engaging with independent researchers, publishing their methodology, and addressing the physics questions head-on. Until then, the video is a fascinating glimpse of ambition — but not proof of capability.

Looking Ahead

Midjourney’s scanner will likely appear in spas and wellness centers before it reaches clinical settings, assuming it ships at all. The company’s cash runway (thanks to its image-generation business) gives it time to iterate. But the clock is ticking on credibility. Every month that passes without clear evidence of superior imaging is another month where the gap between promise and proof grows.

For now, the developer community should watch, learn, and insist on the same standards we apply to any technology: show us the data, let us test the outputs, and be honest about what the device can and cannot do. Midjourney has a track record of delivering impressive software. That does not automatically translate to hardware — and definitely not to medical hardware. The next few months will reveal whether the company can back up its vision with substance.

Source: The Verge. Details as reported; verify specifics at the source.