Welcome to Go/No-Go. I'm Jon Bruner.

And I'm Alex Hao.

This is a podcast about manufacturing, engineering, design, and the calls that can make or break great products. It's brought to you by Lumafield, an AI-powered manufacturing intelligence platform that uses X-rays and cloud-based software to give manufacturers and engineers complete visibility into the products that they make. This is one of the news segments of this podcast where we talk about recent things going on in the world of manufacturing, engineering, and product development. And we're starting off with a really interesting new entrant to the medical device space — one that none of us saw coming.

We make scanners, we love scanners, and apparently Midjourney is also making scanners.

We welcome them to the world of scanners. This is very exciting.

This is a headline from The Verge: Midjourney Enters Medical Imaging. We all know and love Midjourney for their AI image generation tools. But now they are launching Midjourney Medical and the Midjourney Scanner, which is a full-body ultrasound-on-chip device.

This caused a little bit of confusion when we were reading it at first because they use the term CT in some of their materials to describe what they're doing. Generally in a medical setting, CT refers to X-ray CT — like Lumafield's technology, but the medical version — where you use X-ray images taken from different angles and reconstruct them into a 3D model that your physician can slice through to see cross sections. CT stands for computed tomography, and it doesn't inherently have to be X-ray CT. I think that's the liberty they're taking in calling this CT.

It's a really fascinating solution. Basically their vision is a kind of vat of water with an array of about 40 Butterfly Network ultrasound-on-chip modules inside. The ultrasound waves are emitted from the chips, go through the water, pass through your body, and then they reconstruct a model from the imaging.

It seems to be positioned as something you would get done on a very regular basis — just go down to the Midjourney Spa. They've talked about Midjourney Spas having a variety of health-related facilities: gyms, wellness centers, and this ultrasound CT technology. It's positioned as a diagnostic you'd use very regularly, similar to the full-body MRI scans that sometimes get marketed as a consumer product, or the DEXA scan — an X-ray-based scan people use recreationally to determine body composition. This is positioned similarly: something you get on a regular basis just to keep on top of what's happening inside you.

Part of their value proposition is claiming you can have a whole-body scan done in as little as 60 seconds, whereas a normal full-body MRI takes between 60 and 90 minutes of having to stay completely still.

A couple of interesting things about this. One is that medical statisticians take a pretty dim view of preemptive full-body scanning, because a lot of medical diagnostics have a much higher false positive rate than the actual prevalence of uncommon disorders in the population. There are tests with a 5% false positive rate that you'd use to look for a disease or condition that appears in 0.01% of the population. Which means most of the positives you get are actually false positives. A real positive could be life-saving, of course, but false positives in a medical context can be a huge problem if they cause unnecessary surgery or a bunch of additional tests that might introduce radiation. It's a more controversial product within the medical community than a lot of people appreciate.

That's really the criticism of recreational CT that we see advertised — you expose yourself to something that could create more harm than the potential benefit on the other side. That's actually a benefit of ultrasound-based imaging: it's not using ionizing radiation. The device isn't cleared yet for diagnostic use, which is also why they're keeping it in the spa context. The big question with the Midjourney solution is whether it actually works. There have been explorations of this use of ultrasound with water in the past, but you need super clean water with no bubbles. Hair can interfere. Fat doesn't necessarily image well this way. You'll notice all the models in their promotional material are very thin. There are a lot of questions about whether you can really get a viable image out of this process.

Interesting.

There are also questions about water purity, how you ensure no bubbles — someone asked about that and they said they have some kind of vibration system to remove bubbles from the water.

This is going to be a fascinating project to watch. It also highlights that Midjourney itself is a really interesting company. Unlike the other frontier AI labs, it's completely bootstrapped, very closely held, and has little if any outside investment. This suggests they're profitable — that they have real cash flow from their core image generation business that they're able to put into other ventures. I wonder: once OpenAI and Anthropic become profitable, will we see them go in all of these other directions and take on interesting side projects?

People love to try to get into the medical space. You see that with Amazon and One Medical. When you have a lot of money, it seems like people want to get into healthcare.

It's one of the fields that is just begging, as techies like to say, to be fixed. There are of course millions of people who are trying to make medical diagnostics work from a background in medical diagnostics. But it is exciting to see something so radically new come onto the medical diagnostic scene. It's an industry that tends to be very concerned with certification, licensure, and coming from the right background to launch a medical device. This is really bold in a way that feels exciting.

Continuing in the vein of medical devices: the FDA and Dexcom have shared that they experienced a theft of scrapped product. Two lots of their G7 sensors were marked as scrap and sent for destruction. They were stolen during the destruction process and resold by third parties.

This is another story about why it's important to go with an authorized distributor for critical products. The products Dexcom ships to its authorized distributors are entirely safe and go through very stringent quality processes. What happened here is that some products that did not meet Dexcom's standards wound up in the gray market supply chain. An unauthorized distributor that services a lot of mom-and-pop and corner-store type pharmacies managed to get a hold of the products and sell them to retailers. If you stick with an authorized distributor, this is very unlikely to happen. But it's a scary thing to think about from the end user's perspective and makes you consider what kind of pharmacy you're buying from.

It's also interesting to think about how old-fashioned theft and piracy remains such a supply chain consideration. A few months ago there were headlines about $400,000 worth of lobsters stolen in Massachusetts.

I didn't see this.

Someone stole a truck of lobsters. They don't really know what happened — it's a lobster, it doesn't have a QR code, so they're just long gone.

This is why we need serial numbers on lobsters. So we can track these things down.

But it's like — you think about all your quality control processes, making the best product, and then just moving the product introduces all of these avenues for very old-fashioned interventions.

I would love to watch a good heist movie about the theft of these Dexcom sensors. It seems like the kind of thing that would be in a 70s mafia film.

Our next story looks at the defense supply chain. The Wall Street Journal's Drew FitzGerald wrote a piece called "Why It Takes Years to Get a Patriot Missile to the Front Line."

This is a salient issue because US weapons stockpiles are on the low side right now — particularly after the conflict in Iran, but also from supplying Ukraine over the last several years and some other global conflicts. Naturally, if your stockpiles are getting low, you call your supplier and ask for more. It turns out it's not that easy for the Pentagon to call up Lockheed Martin and say they need more Patriot missiles.

Each missile relies on a supply chain of over 400 companies. Another problem is that many of the Tier 2 suppliers in this chain provide components to multiple missile supply chains. So even if you were able to ramp up production for Patriot missiles, that could actually impact the supply of other critical munitions the government needs.

It takes years to ramp up production. The order to ramp up has just recently been issued but isn't expected to fully materialize until 2030.

The Trump administration is really pushing conventional weapons suppliers to rebuild those stockpiles, in addition to investing in more modern warfare technologies. But a lot of these companies have hesitation because the spending can be erratic. Right now they're trying to use seven-year contracts to make suppliers feel more comfortable investing in expanding their factories and bolstering their supply chains.

Manufacturers have to think in terms of the capital outlays required to build an entirely new factory. At some level, a $2 billion purchase order must seem like a cause for celebration — surely they'd want to go out and build you a factory. But they have to amortize the cost of that factory across many years of sales. Without a guarantee that the purchasing program will continue, that factory could become a huge albatross. Spreading contracts across more years to guarantee a more stable flow of demand is a much better way to motivate manufacturers to increase capacity.

The other big problem is the desire to have as many components made in the US as possible. But some elements — like microelectronics — are extremely hard to source domestically. A lot of the chips in current devices are really old. As you run out of those stockpiles, the question becomes who can keep making them, and what about the next generation. That's a huge concern.

A few episodes ago we talked about passenger rail cars and why they're so expensive when the US buys them. A lot of that comes from Buy America provisions for transit equipment, which means there are just a couple of US suppliers doing the bare minimum volume to be practical, and very spiky demand — a local transit agency gets a federal grant, issues an order to Siemens to build rail cars in Sacramento, and it's just very difficult to operate a big manufacturing operation under those conditions. Plus it all has to be domestically sourced wherever possible. So you wind up relying on relatively small-scale manufacturers that specialize in country-of-origin constrained manufacturing rather than being truly competitive at a global scale.

Similarly, the US defense industrial base has shrunk significantly. There were 51 prime contractors in the early 1990s and there are only five today. It's going to take a long time to rebuild that network.

A lot of these weapons programs are decades old. We even have equipment like the B-52 bomber, introduced in 1955, that still flies for the US Air Force. That creates a lot of legacy compatibility challenges. We actually have some customers at Lumafield among the defense community who use our CT scanners to reverse engineer their own products — because those weapon systems predate CAD. And even if they didn't predate CAD, they would be from an era of CAD where you couldn't just load the files into SolidWorks or Catia or NX and start work. The reverse engineering challenge there is quite significant and is a big part of keeping these weapons programs going.

Speaking of the conflict in Iran — a recurring topic these days. There's a piece by Megan Snaith in the Financial Times about how stalled subsea cable projects are threatening Middle East digital ambitions.

What's going on there?

Many countries in the Middle East have been trying to diversify away from oil into a more digital future — huge investments in data centers, digital infrastructure, even linear cities. But when you have a data center, you need to connect it. And that's done through physical infrastructure, including subsea cables. Given all the conflict, the mines in the strait, multiple subsea cable projects are basically on pause. Even with a ceasefire now, there are a lot of concerns. Many of the ships that can lay this kind of cabling were deployed elsewhere during the conflict and it'll take months to get them back. Insurance agencies want to see an extended period of peace before they'll cover these high-tech deployments. They also have to survey the ocean floor to make sure there are no mines where they're trying to lay the cables. There are cables planned to connect Africa, Asia — especially India — and parts of Southern Europe that are all on pause right now.

That's fascinating. I wonder what it would be like to be a ship insurance underwriter in these conditions. What data streams are they looking at to figure out whether it's too risky to insure a ship in the Strait of Hormuz?

We should ask Shahram Mokhtari from iFixit. He used to work on things like that.

That's right. We'll have him back on another episode of Go/No-Go.

These cables have seen a lot of disruption over the past few years, including incidents of sabotage in Europe. There's a growing sense of the vulnerability of these systems to geopolitical instability.

I wonder at what point satellite communication might be able to overtake undersea cable communication. It seems a lot easier, but the throughput must not be comparable at all.

At my last company I actually worked on some of the lasers that go into this type of infrastructure. There's a lot of interest in space-based optical communications — optics are so much faster than other communication technologies. That's why you see data centers moving from copper to optical wherever they can. And with undersea cables you can increasingly add more and more bundles of fiber to move data at speeds that aren't possible any other way.

Plus undersea cables are relatively easy in that you don't have to buy easements from thousands of different property holders. You can just do a straight line. And I've read that the rumor about sharks attacking undersea cables is somewhat exaggerated.

I mean, these cables are huge. If you've seen pictures of the cross sections — they're really well protected. I don't think a shark is going to do anything to one of those.

I think the cable wins.

But when a cable does break, it costs millions of dollars a day to fix. You have to find it, bring it up, cut it out. It's a really fascinating piece of infrastructure that people don't appreciate — the cloud is actually a lot of it is underwater.

That's an awesome slogan for an undersea cable provider. The cloud is underwater. Moving on to land-based communication — have you seen the Snap AR glasses?

Yes. And I'm getting ads for them absolutely everywhere. Every app I open — Instagram, Financial Times, New York Times. Just constant Snap Spectacles ads.

How are they being positioned?

I didn't even realize at first that the models were celebrities because the frames are so thick. I thought they were just models. Then I clicked on the ad and it's like, oh, that's Jimmy Butler. I had no idea that was Jimmy Butler under there.

An interesting decision to pay for a celebrity's face if the face becomes obscured by the product.

They're really trying to position them as fashionable in the advertising. And there are memes about how Evan Spiegel completely undermined his entire marketing campaign by how uncool he looks wearing them in the wild.

He mostly just looked uncomfortable wearing them. They were sort of crushing his ear in some of the pictures. These glasses are enormous. For listeners who haven't seen them, they're chunky — there's a way of looking at them as chunky throwback glasses from the 60s. They are definitely statements.

For people who aren't familiar: Snap was selling the first generation AR glasses only to developers within their network. Now anyone can buy them. They're $2,195 with a $200 refundable deposit and should ship later this year.

That's pretty expensive. What are people going to do with them?

The technology itself is really impressive. The arms look ridiculous — they're so long and bulky — but there's no external puck or anything. It's fully integrated. Snap has been investing in these glasses for years. Back in 2021 they spent $500 million on WaveOptics, which makes waveguide technology that makes the lenses look clear while still having a large display. People I know who've tried them say the display is actually very impressive. Whether they're wearable day-to-day and what the actual killer applications are — I'm not sure yet.

I find I'm a poor judge of how these wearables will do in the market. When the Meta Ray-Bans came out a few years ago, I thought it was an interesting side project for their VR ambitions. But they've actually done quite well. And you have a pair, right?

I love my Meta Ray-Bans. They're so great — speakers, camera. My mom loves them. She got all of her friends at tennis into Meta Ray-Bans.

Do they record themselves playing tennis?

They don't, I guess — they'd probably get motion sick from the video. But my mom really likes the camera. I think the photo editing is just different from the iPhone, and she prefers it.

Do the Meta Ray-Bans have a display in the lens, or are they completely input devices?

There's a version that has a display, but I haven't tried those.

Sometimes when I've talked to people wearing AR glasses, I find it a little difficult to communicate with them because I always have the sense that they're reading something while pretending to make eye contact with me.

Snap actually has a pretty successful ads business, but it doesn't really reflect in their valuation. And because Evan's been so into these glasses for so long, investors keep writing letters like, please stop.

Snap's stock fell a bit after they were announced — some hesitation about the product design, and maybe concern that the company might remain focused on an expensive side quest rather than its core digital business.

And it doesn't even speak to their main users, which is teenagers. A huge part of their value is that it's a really unique advertising market. But those are not people spending two grand on smart glasses.

They're people like perhaps you in a few months.

I know. Well, what I'm definitely planning to buy this year is the foldable iPhone. I don't know what it's going to look like, but I know I'm going to buy it. I've been waiting for years.

There are a lot of rumors. Have you always wanted a folding phone?

I think so. Ever since I saw the first Samsung foldable I wanted one. But I can't leave the Apple ecosystem.

You're going to watch so many YouTube videos in full landscape mode.

I'll watch this podcast in full landscape mode.

It's broadcast in 4K so it's worth it. My final thought about Snap is actually quite positive. These Spectacles are a remarkable piece of hardware development, especially when you consider the difference in resources between Snap and the other companies doing cutting-edge consumer hardware. I think of Snap as a large profitable tech company, but if you compare it to Meta, Apple, and Google — which have market caps ranging from $1.4 trillion to $4.3 trillion — Snap is about 500 times smaller, with a market cap of just $7.7 billion. The resources available to Snap are much, much smaller. It's a much bigger bet for the company to go after a novel piece of hardware like this, and the progress they've made on things like waveguide lenses and other technology apparently in these glasses is quite impressive.

I'm excited to see them. Speaking of gadgets: Brian X. Chen at the New York Times has written a guide on the dos and don'ts of buying used tech gadgets. Memory component shortages are pushing up electronics prices — Apple just raised prices on a bunch of their devices, citing memory as a constraint. So he's published a guide on what makes sense to buy used and where to buy it so you don't get scammed.

Do you think it's going to become fashionable to buy old hardware?

We buy a lot of questionable gadgets here at Lumafield, so a lot of his tips make sense. He cites buying from the manufacturers — ideally they have an official refurb program — or buying from a big brand with return protection. He gives the example of buying something from Amazon and being able to return it because it arrived scratched. We're always buying random things on eBay, AliExpress, whatever. But if you're not looking for the newest thing and that's okay, you can save a really considerable percentage.

I wonder if there's room for a vintage electronics market, a little bit like the vintage car market. Lumafield's co-founder and head of product Andreas has a really old iPhone that he loves, from when Apple made a very small form factor phone. And I feel like that's kind of coming back around to being a cool novelty you can show off — this 10-year-old iPhone that still works and fits easily in a pocket.

I think that's true for some gaming consoles too. Even though there are emulators, some people really want the old machines.

They want to experience those games the way they did in childhood. I think there's room for it. We just have to make it a little trendier. Speaking of questionable hardware — you recently bought an interesting robot from Temu.

I bought a robot dog. In the advertising material it's meant to look like one of those Boston Dynamics dogs. But then you get it and it doesn't look anything like that. It's a little bit better though.

But it does have Bluetooth audio.

It does. It plays music. You can click the kung fu button and it plays weird music. And it can't just move forward and sideways — the wheels actually allow it to move fully sideways. It's terrifying. It really upset all the office dogs. They hate it.

Watching a little robot dog moonwalk around the office is an unnerving thing even for the humans.

Some of our team has been at the Automate conference in Chicago this week, and it sounds like Automate is just full of robots — humanoid and dog robots everywhere you look.

We need more labor in the manufacturing field. Labor shortages are one of the key barriers to the resurrection of manufacturing in the United States and a lot of developed countries where populations are stabilizing and people are becoming wealthier and less willing to work in manufacturing. Perhaps the answer is automation.

The Financial Times has a piece called "Robot China's Bid to Beat its Demographic Decline." Something you see at Automate — or even at random events for pizzazz — is that a lot of the robots we're seeing are Chinese. This is stemming from a state-level initiative driven by national concerns about labor shortages as the population ages. The working-age population in China, which they count as ages 15 to 64, peaked at a billion last decade but will fall to 300 million by 2100.

That's a total cratering.

Exactly. They're worried it will impact their ability to remain a global superpower. They installed the most industrial robots in the world and now also make most of the world's humanoids. They accounted for 90% of the roughly 13,000 to 16,000 humanoids shipped globally last year. This year they expect homegrown humanoids to rise to 50,000 units.

One of the advantages Chinese manufacturers have is the ability to tap into an enormous working-age population and pull together a workforce very quickly. If you're a really large customer of Foxconn or Pegatron, you can have 10,000 workers show up at your factory all at once. This is a vision for doing that, but with robots.

Right now it's not quite working. The robots are bad at tasks that are too easy — super repetitive tasks that a normal industrial robot handles well. But if a task is too complicated, the robots also can't do that. They're getting progressively better though. It's been really big in the services sector. In the hotel space, for example, some companies have reduced their staff-to-room ratio to 0.1 — a 100-room hotel with just 10 employees because the people bringing utensils and water are all robots.

I was in a restaurant in rural Texas about a year ago that had one of those busboy robots hovering around. Servers were taking dirty dishes and stacking them on top of this tray-table thing that was wheeling around. I asked my server whether she liked working with a robot, expecting to hear that it was unnerving or that she felt bad about the job loss. She said everyone there loves it. In a tight labor market, it's really hard to keep people in entry-level jobs like busing. Workers just don't show up sometimes. When that happens, the servers have to bus tables, which means they can serve fewer tables and get fewer tips. Robot busboys actually improve the productivity of the whole staff. Everyone in the restaurant agreed.

There's a lot of nervousness in China about how this is going to impact social cohesion. But I guess we just have to wait and see.

There may be no alternative if China wants to remain a manufacturing powerhouse.

Speaking of the importance of people in this robotic environment: The Verge has a headline about how Ford had to hire back former engineers to fix mistakes made by its automated systems.

What did they do in the first place?

They had basically tried to transfer institutional knowledge from experienced employees into different systems to automate parts of vehicle development. But it wasn't working out. Ford has really led the industry in the number of recalls and has seen their quality ratings slip for years. Recognizing that, they decided to rehire about 350 former senior employees to bring that talent back in and shift from a reactive approach — finding and fixing problems after the fact — to actually preventing them from happening in the first place.

Product development turns out to be very complex.

When people work somewhere for a really long time, they develop a lot of tacit knowledge. You can look at a situation, read it, and interpret it correctly. I think that wasn't fully appreciated at a senior level. People with that level of experience don't even know what they know because it's so natural to them. That's what Ford came around to recognize. They had these former employees come back to retrain the systems, mentor younger engineers, and improve data collection. And now they have a renewed appreciation for the value of that knowledge.

The last time we talked about Ford on this podcast was something along the same lines. We discussed their effort to develop a new, very low-cost stripped-down truck, created by a Skunk Works group of engineers who threw out all the old cumbersome design details they felt the main truck division was hamstrung by. It turned out that a bunch of those were in fact sensible design details, and they rolled them back in. In particular, I think the article focused on a protruding lip above the windows, which the Skunk Works group thought was an archaic hangover from earlier designs. It turned out that lip was critical to preventing the window from leaking. People intuitively know why you have that lip over the window. It never occurs to anyone to write it down. So the automated system never learns it.

And this initiative has really worked. Ford was just named number one in J.D. Power's initial quality ranking among mainstream automakers. They've been open about how they realized relying too much on automated systems was really hurting them.

Interesting. This isn't exactly the same thing, but a couple colleagues and I visited the Ford assembly plant at River Rouge in Michigan a couple of weeks ago.

How was it?

Very cool tour. Have you ever been?

No, I haven't.

It's a public tour — anyone can sign up. They let you into a little gallery and you walk along a catwalk over the assembly line. It's a much more manual process than you might imagine if you've seen a lot of automotive factories, where the dominant image is robots flinging things around and spot welding. This is the line that produces the F-150, which is now largely an aluminum body. There's a lot less welding than there used to be. What you're walking over is really the final assembly process: bolting things together, installing all the interior trim. There are some robots doing things like windshield and rear window installation, but the rest is very manual — installation of the center console, the ceiling liner, the seats. It's assisted manual assembly, using a lot of powered jigs and things like that. But fundamentally, a human is maneuvering the center console into the truck, and then two more humans are coming in with torque wrenches and tightening the six bolts that hold it in place.

That's awesome. I'm so glad you can still find factory tours for automotive.

There are so few now. But that is a great one. Anyone who's in Dearborn, Michigan should go see the Ford factory tour at River Rouge. With that, we're going to wrap up this episode of Go/No-Go. If you want to get in touch with Alex and me, email gng@lumafield.com — we love getting your emails. We've heard from a handful of people whose work we've actually discussed on this podcast, and we especially love that. Go/No-Go is brought to you by Lumafield, which makes an AI-driven manufacturing intelligence platform that gives engineers total insight into the products they're building and shipping. You can learn more at lumafield.com. For Go/No-Go, I'm Jon Bruner.

And I'm Alex Hao.

Go/No-Go is brought to you by Lumafield, which was founded to upgrade manufacturing. You can learn more at lumafield.com. Go/No-Go is produced by Austin Carder and edited by Brian Tran, with additional assistance from Eric Petralia.