Skip to main content

Insight · June 18, 2026

Designing around
the machine.

Midjourney shipped a body scanner. The device was the easy part.

01 · What shipped

A working machine, not a concept.

You step into a shallow pool of warm water. A ring of half a million sensors sends sound through your body from every angle. About a minute later you have a 3D map of your insides down to a fraction of a millimeter. No radiation. No tube. No lying perfectly still.

It is ultrasound, carrying the resolution of an MRI. A full body scan in 60 seconds instead of an hour in a tube. They are calling it the Midjourney Scanner, and yes, it is the same Midjourney that makes the images.

60s

Full body scan

100x

Faster than an MRI

10x

Cheaper

9

People who built it

The film

A technical dive inside the Midjourney Scanner · Midjourney Medical →

02 · The category they walked into

A market that was supposed to be impossible to enter.

Medical imaging belongs to a handful of giants. The moat was supposed to be the physics, the capital, the regulators, the decades of installed base and service contracts. It is the kind of category a sane founder is told to never touch.

A team of nine, at a company known for generating pictures, shipped a working machine into it anyway. Not a slide. Not a render. A device you can stand inside.

The reframe

“They did not build a hospital machine. They built a spa. The scan is a side effect of a place you would want to be anyway.”

03 · The move

They redesigned the room, not the device.

The incumbents spent decades optimizing the machine. Stronger magnets, tighter tolerances, faster reconstruction. Every gain lived inside the same anxious experience. A cold tube. A loud room. A reason to be afraid before you arrive.

Midjourney left the device alone and rebuilt everything around it. Warm water instead of a tube. A place you would choose to visit instead of a place you are sent. The scan stopped being the point and became the thing that happens while you relax.

That is the whole story compressed into one decision. The spec sheet was never the product. The experience was, and the team that owns the experience owns the category.

A glowing gold cross section of the body with each muscle and organ labeled by name, the output of the Ultrasonic CT method
Ultrasonic CT · every structure named down to a fraction of a millimeter · Midjourney

04 · Why the image company

It is the same problem wearing a different coat.

Generating an image and reconstructing a body are the same discipline underneath. Both take incomplete, noisy signal and resolve it into a coherent picture. Midjourney spent years getting very good at exactly that, then pointed the competence at sound waves instead of pixels.

This is not a company doing something random. It is a company following its core skill into an adjacent market that nobody thought it was allowed to enter. The adjacency was always there. Most teams just never look up from the category they are told they belong to.

05 · Nine people

The headcount a hard category now requires.

A decade ago a project like this needed hundreds of people and a war chest. The number that should stop you is nine. The tooling collapsed the team you need to attempt something this ambitious, and a small group with sharp taste now reaches what used to require an institution.

This is the part worth internalizing. Small teams are not a constraint to apologize for. With the right leverage they are the advantage. Fewer people, more taste, faster decisions, and nobody to slow the one idea that matters.

Closing

The machine was the easy part.

The hard part was everything around it. The water, the room, the reason to come back. That is the work, and it is the work most teams leave undone.

Donny Smith

Written by · June 18, 2026

· ECD, Founder, Bttr.

Over the past 15 years, he has led creative teams and contributed to products used by millions of people worldwide, working with companies including Apple, Opendoor, JP Morgan, GE Aerospace, Pepsi, and Alterra Mountain Company.

LinkedIn

Imagery and figures · Midjourney Medical · A New Era of Midjourney →

Share this perspective

More insights

Adjacent perspectives.

How to Evaluate a RAG System

8 min read

How to Evaluate a RAG System

Evaluating a RAG system means scoring its two stages on their own, because a retrieval augmented generation answer is only as good as the passages behind it, and a wrong answer has two possible causes a single number cannot tell apart. Retrieval is measured before the model writes anything. The LlamaIndex evaluation guide scores a retriever with hit rate, the share of questions where the right passage lands in the top results, and mean reciprocal rank, which rewards ranking it high. Generation is measured given what was retrieved. Ragas, the open evaluation toolkit now maintained by VibrantLabs, splits the answer into claims and scores faithfulness as the share the retrieved context supports, then scores answer relevance by generating questions from the answer and comparing them to the real one. Read as two pairs, the four metrics locate the fault: high recall with low faithfulness is a generation problem, low recall with high faithfulness is a retrieval one. Anthropic showed the payoff in September 2024, cutting a retrieval failure rate measured by recall at 20 from 5.7 percent to 1.9 percent by adding context, keyword search, and reranking. And it is not a one time check. Models drift, documents go stale, so the teams that trust their systems run these scores on live traffic the way they run unit tests.

What Is a Reasoning Model

8 min read

What Is a Reasoning Model

A reasoning model is a large language model trained to work through a problem in steps before it answers, rather than reply in one pass. It spends extra computation at the moment you ask, called test-time compute, to write out a chain of thought and check its own work. Anthropic describes its extended thinking as letting the very same model give itself more time and effort rather than switching to a separate model, and reported that Claude 3.7 Sonnet reached 84.8% on the hard GPQA science set, with a 96.5% physics subscore, using parallel sampling at a 64K token thinking budget. The habit is trained, mostly by reward. DeepSeek showed in January 2025 that its R1 Zero model learned to reason purely through reinforcement learning with no supervised fine-tuning first, developing self checking and long chains of thought on its own, and shipped DeepSeek R1 under the MIT license at performance comparable to OpenAI o1. The cost is real, so the newest models add a dial. Claude Opus 4.5 hit its best score on SWE-bench Verified using 76% fewer output tokens at a medium effort. The skill is matching how hard the model thinks to how hard the task is.

What Is Fine-Tuning

8 min read

What Is Fine-Tuning

Fine-tuning is training a pre-trained model further on your own labeled examples so a behavior is set into its weights. It changes how a model acts, its tone, its format, its grip on a narrow task, rather than what facts it knows. Google frames the move as taking a general-purpose model and adapting it to your domain on a curated dataset of your own examples. The gains are measurable. When Anthropic opened fine-tuning for Claude 3 Haiku in Amazon Bedrock, a team fine-tuned it to moderate comments and lifted classification accuracy from 81.5% to 99.6% while cutting tokens per query by 85%. Lighter methods such as LoRA keep the base model frozen and train a small set of new weights beside it. The order that saves money is a plain prompt first, retrieval when the model lacks facts, fine-tuning only when it lacks the right behavior. A fact that changes is a retrieval problem. A behavior that drifts is a training problem.

Industries We Serve

Aerospace & DefenseBiotechnologyMedical & HealthcareManufacturingFinancial ServicesConsumer ProductsEnterprise Software

New Business

Start a project

Headquarters

North America

© 2026 Bttr. All rights reserved.