The first headline I have is from Bloomberg, and it's titled, Almost $1 Billion Later, the US Still Can't Make a Medical Glove.

A medical glove seems like it would be an easy thing to make. Is it an easy thing to make, Alex?

Unfortunately, it is not.

Only about 1% of nitrile gloves used in the US are made domestically. Most come from Malaysia, with China supplying the key raw materials. This became a big problem during the pandemic. Obviously, materials like that were in really high demand, and not having any domestic source made it really challenging. So there was a federal push, $1 billion worth in 2020, started under Trump and continued under Biden, but it's produced zero gloves. None of the 6 companies that received money are making gloves or the raw material for them today.

Wow. So is there any domestic glove manufacturing operation in the US?

There are a lot of challenges as to why it's been hard to scale up glove manufacturing. First, it's the raw material. It's really chemically intensive to produce. It lacks a lot of existing US supply chain and infrastructure. And then the gloves themselves are really complex compared to, say, masks or gowns, so it's a much larger lift. Ultimately, too, when the gloves are made by American companies, they cost more. And since a lot of these organizations that buy gloves, hospitals, the government, they're buying in bulk, they don't really want to pay more for the same thing.

Yeah, these are disposable objects that someone in, say, a healthcare setting might be using dozens of every single day.

Exactly. Some smaller players like American Armor Gloves can source their material from Italy and South Korea, and they're benefiting from demand spikes tied to the Iran war, especially given petrochemical cost fears. So there's a couple bright spots, but it looks like the government isn't planning to fund these companies any further. They're just pushing procurement rules to try to require agencies to buy American-made gloves when available.

We've talked about those types of procurement rules in the context of, say, railroad equipment. They have a mixed track record, and it becomes tempting to skirt them a little bit by moving the bulk of your supply chain overseas, but perhaps doing a final assembly step in the US.

I think ultimately this glove manufacturing initiative, and its lack of success despite receiving $1 billion, really shows how reshoring is so much harder than people think. And a lot of these failed initiatives erode the credibility to continue doing it in the future.

I still remember buying a box of nitrile gloves at some point during the pandemic, and they had a little American flag on them. I was like, all right, American-made nitrile gloves. And then under the American flag, it said distributed from the United States. And in different text elsewhere in the box, it said made in Malaysia. Well, it might be a little while before we see any gloves that are truly made in the US at scale. Switching gears to the food world, there's just been a major disruption to the supply of milk.

Are you familiar with Fairlife?

I think I've seen it in the grocery store. It's an attractively designed bottle. It looks like a national brand of milk.

I have a lot of friends that exclusively are into Fairlife.

Really? What is it?

It's ultra-filtered milk. They have a lot of different protein drinks. It's supposed to be, I think, better in terms of the lactose.

So do your friends just drink it as a beverage? They reach into the fridge and pour themselves a glass of Fairlife milk?

This is something I learned recently when I visited one of my friends. He drinks a lot of milk.

No kidding. Is he 5?

That's what I said. But I guess it's protein.

Yeah, sure. People want protein. This is a major trend in the food world, everyone wants high protein food.

And he's like, you're not buying Fairlife milk, I only buy Fairlife milk. But anyway, I thought of him when I saw this headline from Inc. Coca-Cola bought Fairlife a few years ago, and their retail value has grown from $10 million in 2014 to $4 billion last year.

Whoa.

It's a huge brand. It's growing, high profile. My friend is not the only Fairlife fan, obviously, but they have now experienced a ransomware attack. All of their US production is suspended.

Wow. What happened?

A hacker group infiltrated Fairlife's production facilities. We've seen this before. We talked previously about the hack at Jaguar Land Rover. These kinds of manufacturing facilities are really attractive for these kinds of hacks because they are so disruptive. It can halt production really fast. It has a lot of cascading supply chain impacts on retailers and consumers very quickly. The ransomware attackers know that the companies they're targeting are highly motivated to get their systems back off the ground.

They're also sometimes unprepared for this type of thing, because we talk a lot on this podcast about digitizing manufacturing and adopting digital technologies. In some ways, manufacturers have been a little slow to adopt some cutting-edge digital technologies, but in many ways, manufacturers are highly digitized, and they got into basic ideas of digitization pretty early, which means that a lot of factories and food processing plants are full of older equipment that is online, connected to some sort of process control system, whether that's SCADA or just an ERP network running in your factory, and that creates a pretty large surface area for attacking. If you're running a giant stack of software and hardware that is a blend of technology from 30 years ago up to today, that's exactly the kind of environment that attackers have an easy time going against.

We'll see how this gets resolved. The JLR hack took over a month to handle, which was extremely disruptive for the entire UK supply chain.

Yeah, the more complex your system is, the longer it takes to unwind this.

A funnier, smaller problem Fairlife had was they were actually sued earlier this year because the plaintiff claimed that the cows the company uses don't actually live a fair life but are mistreated.

And also it has a happy cow cartoon on the front, which is relevant to the case as well. Maybe if the cartoon showed a sadder cow, they wouldn't have gotten sued.

An empathetic cow.

Yeah, exactly. Just a neutral cow is all you need. Well, let's change gears and talk about AI. This is always in the news, but recently, in the last few weeks, we've seen the emergence of some new frontier models like Fable and ChatGPT-5.6 Sol that have blown through a lot of the benchmarks. And in quick succession, the emergence of open-source models and lower-cost models that perform just about as well.

Christopher Mims at The Wall Street Journal has a piece, AI's Wider Availability Is Good for China, Not Great for OpenAI and Anthropic. OpenAI and Anthropic are gearing up to have these blockbuster IPOs. But at the same time, these Chinese models are emerging that are significantly cheaper and pretty good. Silicon Valley is pretty familiar with the fact that many people will settle for good enough if they can save a lot of money.

That's right. And it comes at a moment when a lot of companies are starting to realize that they spend a lot of money on AI. 6 to 12 months ago, the zeitgeist was that you needed to get all of your employees using AI. And just in the last couple of months, more companies have realized that they're spending a ton of money on inefficient AI usage. Now they're looking for ways to trim that back and become more efficient with what they're doing.

I think a lot of companies initially were rewarding employees that use the most AI, and now they're realizing we're not saving any money on employment that way.

That's right, exactly. A monthly AI bill for a software developer who's using it inefficiently could be about as expensive as it is to have a software developer in the first place.

Another article looking at the role of China in AI is by Cade Metz at The New York Times. It's Why Silicon Valley Can't Stop Looking Over Its Shoulder at China. It references Xi Jinping's recent speech, where he talks a lot about the importance of open source models. Do you know why the open source model is so disruptive compared to what OpenAI and Anthropic are doing?

OpenAI and Anthropic are charging for the usage of their models. With these open-source models, users have a choice. They can pay to use it on infrastructure that's hosted by the model provider, or they can download a version of the model and run it on their own equipment. It might require some reasonably expensive computing equipment, but then it's much cheaper to run it on an ongoing basis, and you're not being billed for all the tokens. Of course, open-source software makes it easier not only to copy certain aspects of it, but also to go through a process called distillation, where you repeatedly query one AI model to reverse engineer how it approaches problems and then incorporate that into your own AI model. So from a national security perspective, a lot of US policymakers and the leading frontier AI labs would say we don't want to open source these models and make it easier for China to catch up with us. Of course, Xi Jinping would like it very much for China to be able to catch up in that way. Another way of looking at it is in terms of shutting out competition. Anthropic and OpenAI have this way of calling on the federal government to more fully regulate AI. They want to bring down AI regulation in ways that are favorable to them. And it makes me think of the end of the tobacco advertising era, when it was proposed that we ban tobacco advertising. Philip Morris was actually supportive of that initiative in some ways, because it owned several of the leading tobacco brands like Marlboro. So if you shut off all tobacco advertising, you freeze the structure of the market and the market shares in place. You're not able to advertise new entrants. You're not able to disrupt the market by reaching consumers in a new way. So you lock in the dominance of Marlboro and all the other incumbent brands by banning advertising. And in the same way, if you regulate the development of new AI models, you effectively lock in Anthropic and OpenAI and Google and a handful of others that have the really leading frontier models.

Cade Metz talks about how, right now, America is still in the lead. It's really their race to lose, but the developments from China are really starting to bridge that gap.

Yeah. And it causes some panic in the market here. Stock prices of a lot of the relevant AI companies in the US drop every time there's one of these announcements where a Chinese open-source model has gotten really, really close. And I think it's also put a lot of pressure on the US frontier AI labs to bring their costs down. A few weeks ago, we saw Anthropic release Fable 5, a breakthrough model and top performer. And just today, the day we're recording, Anthropic announced Opus 5, an evolution of its Opus model that equals or exceeds Fable 5's performance on a lot of benchmarks at half the cost. So it seems like they're both responding to cost pressure from some of their customers, and perhaps they're even distilling their own Fable 5 model down into Opus 5.

The Opus-Fable pricing change is interesting because we see how these companies in the US have to respond to pricing pressure from developments in China. And they also deal with a much different regulatory environment, as you were mentioning. Last month, Anthropic shut down its 2 most powerful systems after a really surprising government demand to bar foreign nationals, including employees of these companies, from accessing the models. And OpenAI has also said that the administration has interfered with product releases. So that's another way in which the approach of the US versus China towards these technologies really differs significantly.

Yeah, the US federal policy right now is both very supportive of American frontier AI labs maintaining their dominance and continuing to develop incredible new models, but it's also quite anxious about seeing those models escape into the rest of the world, and it's eager to wall off this American capability a little bit. So the fact that these Chinese models are able to almost catch up calls into question whether you actually can wall these things off.

And I think the unpredictability of the government response also makes it hard for these companies to plan for their future and develop things in a streamlined fashion.

Exactly. Speaking of things that they're developing, if you think about the core functionality of these chat-based agents reaching new heights and actually converging, the capabilities of all these labs plus the Chinese models are reaching the same level. Something you might think to do is expand the capabilities of your AI model into new domains, such as the physical domain, perhaps controlling robots or incorporating hardware that can bring the model out into the world. That's something we talk a lot about at Lumafield, because we have this very unique set of physical world data in the form of CT scans, and there's a lot you can do with AI to exploit that and find patterns in the world that you would never be able to uncover otherwise. AI is very good at that, especially when you have a unique dataset. We'll have more to talk about on that front later this summer, but you can see the frontier labs having the same thought. OpenAI has been widely understood to be working on some hardware products for a long time, and we have a little more clarity on what those are.

Bloomberg's Mark Gurman has reported that OpenAI's first consumer hardware product will be a portable, screen-free smart speaker. It's being pitched internally as a human-like AI companion that lives in the home. It can control all your appliances, play media. It also reportedly has some mechanical elements that let it move on its own around the house.

It sounds like an Amazon Echo speaker.

Have you ever seen the Amazon Astro?

No. What is that?

It's a cute little robot with a tablet face that was supposed to roll around your house and do Alexa stuff, read stories to your children, and things like that. It was never super successful. I think it's still nominally on Amazon, but it has to be invite only. You can't just buy it.

Interesting. What stood between it and success, do you think?

I think people just don't want a surveillance robot following them around their house. Amazon did pursue doing the Astro as a security product, and it didn't. So they did the thing that everyone thought the device was.

They just made it more explicit.

Yeah.

That's interesting. It makes me think of this device I heard about a long time ago, an alarm clock that sits on your nightstand, and as soon as the alarm goes off, some wheels on it fling it off your nightstand and it starts running around the room.

I have that clock, actually.

You have that?

The developer of that clock was my sister's intern.

No kidding. That's insane. Do you use it?

I used it a lot in high school. It's been around for a while.

And it's for people who hit snooze too much?

Because then it runs away, so you have to go get it.

Does it just fill your waking moments with panic?

A little bit. It's like, I have to grab it right now.

Yeah, exactly. I would be so worried. You have this thing clunking around under your bed, and you're fishing around, groggy, trying to wake up. It is something that would wake me up for sure. I've thought about getting it, but I haven't gone quite that far.

This is the device that's speculated they're going to release. It's anticipated in the second half of this year, but they also have a real piece of hardware that we've seen around the office. It's a keypad built for Codex.

Interesting. So OpenAI's very first piece of hardware is not in fact an AI-enabled...

Correct. It's the Codex Micro. It's a $230 mechanical keypad developed with peripheral maker Work Louder. It's a rotary dial that can adjust the AI's reasoning level, and it has a joystick to trigger coding workflows. It's a physical way to interact with your AI.

Speaking of hardware, the next article I have is a visual story from the Financial Times, The Century-Old Device Choking the World's AI Push. It's a fascinating deep dive into transformers. When I think of transformers, I mostly think of when I was a senior in college. I was living in downtown Providence in the dorm, and one day there was this huge boom and all the power went out, and you could see all the office workers in downtown Providence going out onto the street. It turned out that a squirrel had gotten into the transformer and exploded. It cut all of the power to the city.

Wow. At least it was a fast end for the squirrel, probably. Those transformers are also full of terrible chemicals, I think, and when they explode, it is not good.

That was the last time I really thought about transformers, the complete power outage and seeing everyone totally lost with what to do.

Yeah, this article is interesting because it suggests that the transformers we still rely on in a lot of places are the classical transformers you might remember from high school or college physics class, with an iron core and copper wound around it. The ratio of copper on one side to the other is the ratio of the voltage transformation.

That is exactly what's driving the shortage. These transformers are essential to manage the massive power that AI data centers use. That's a huge headline, how much power they use. Everyone's talking about the impact on the grid. But the transformers are still largely made in a very manual winding process. There's a material constraint around having enough material, but there's also a really skilled labor element that's incredibly hard to replace.

Plus, these are very capital-intensive manufacturing processes, and the manufacturers of the transformers are concerned that this AI data center buildout boom is a fad and that it's going to disappear.

Making a transformer can take 3 to 6 months per unit, and it can span across 80,000 different possible designs. They're investing hundreds of millions to expand capacity because they're booked out 4 to 5 years into the future. But they're also hesitant, wondering if this is really going to keep going.

Yeah, this is a recurring theme on this podcast, the difficulty of managing manufacturing in markets where demand might be really spiky. We talk about it in government procurement too, where the US wants to buy a lot of new missiles and the manufacturers don't necessarily want to increase their missile manufacturing capacity by a factor of 10, because that might leave you with a lot of very expensive capacity in later years that you can't amortize.

There are a few companies looking into long-term fixes. Startups like Heron Power are focusing on solid-state transformers. Those use power electronics instead of the traditional steel and copper designs. But all of this has a really long lead time as well.

It'll be interesting to see if we can get to the point where solid-state transformers can totally overtake the traditional ones. I had just assumed that a lot of transformers were now solid state. Solid-state transformers are everywhere in consumer electronics and lower-power applications. The transformer bricks, the power supplies we attach to our laptops and phones, are all solid state, which is a huge improvement over the giant wall warts we used to have 20 years ago. These transformers are now tiny, reliable, and adaptive to the cables you put in. It hadn't occurred to me that this technology hadn't really made its way into large-scale power distribution applications.

Definitely. They're fascinating to look at. Everyone should go take a look at the photos in this article. They're incredible.

Speaking of AI infrastructure and what's inside those data centers, The Verge has a great article about GPUs.

Justine Calma has a piece, Who's Afraid of the Big Bad GPU? She makes the argument that the energy debate around AI is interesting because GPUs are actually everywhere. We're thinking a lot about data centers and whether they're a valid use of this much energy and technology, but GPUs are also in gaming and in your phones. I think back to when people were mining a lot of Bitcoin. I assume people are still doing that. But AI is being uniquely scrutinized for its value and environmental impact in a way that other uses of the same technology are not.

That's right. And it's something you see also when people talk about AI data center consumption of water. When you state the absolute number of gallons a facility uses, it's a very big number. But when you compare it to many other familiar uses, it's a really tiny number. Compared to, say, a golf course, an AI data center might use considerably less water. The biggest users of water are agricultural applications, which are orders of magnitude bigger users of water than data centers and other household uses. So it helps to put these things in perspective and zoom out, in this case with GPUs, to look at other ways they're used.

It's interesting when the environment gets used as a reason not to do something, when there are so many things that are terrible for the environment that aren't generally scrutinized by the same people evaluating data centers in this lens. It's not to say they shouldn't be doing that, but it almost seems like a convenient story for when you're not getting enough personal value out of it to scapegoat.

That's right. They are striking facilities. You've visited some hyperscaler data centers?

No, they were just my customers, but I didn't get to go.

Oh, okay.

I haven't. I wish.

I haven't been in one, but a few weeks ago I was in southeast Wisconsin and drove past a brand new Microsoft data center in Mount Pleasant, Wisconsin, in the final stages of construction, already with a phase 2 under construction next to it. The scale really is very striking. I hadn't absorbed before seeing it how big they look from the outside.

It's like an ominous presence on the landscape.

At one level, it is a majestic monument of human engineering that resembles the scale of a highway or a giant bridge. But it is different enough from other things in the landscape that you can understand why people dislike them for aesthetic reasons. But I also think a lot of the objections to data centers are aesthetic.

They are disproportionately situated near low-income communities and communities of color. There's a valid environmental justice angle in terms of where these things are placed and how they're built. But the fundamental questions about their overall energy use are maybe a bit more muddled.

That's right. A lot of people are worried about their energy use driving up regional electricity prices. But even the evidence there is mixed. Virginia has one of the largest, if not the largest, concentration of data centers in the US, and electricity rates in Virginia are below average for the US. So if you really thought data centers were going to push up everyone's electrical rates across the board, it's not clear that happens. Virginia also probably has much easier regulations around developing new power sources. Really, it's a matter of keeping electricity supply and demand balanced, and that's what keeps electricity rates down. I think in a lot of other states, it might be easier to build data centers than it is to build power generation, and that is a recipe for power rates increasing.

I think the core question that comes through in this piece isn't whether these data centers use resources, because obviously they do, as everything does, but whether the specific growth of these technologies and the consumer value justify those costs, whatever they may be, and whether some of the manufacturers in the space, whether it's transformers or chipmakers or all the other many players, should be bearing more accountability.

We've been talking a lot about AI, but the last headline I have, I know you love trains, Jon.

What are you trying to say about me, Alex?

What is that game you always say you play on airplanes when you need to incinerate time? Is that the train game?

Oh, yeah, yeah, yeah. Well, when I'm on very long flights, I sometimes enjoy playing a game called Transport Fever, in which you have to operate an effective transportation network, principally trains, but they also offer other modes like ships, trucks, and airplanes. You have to make sure steel mills are well supplied with a balance of iron ore and coal, and that tool shops are supplied with a good supply of lumber, which comes from a lumber mill, which comes from a forest, because you have to supply the lumber mill with forest products. Anyway, it is a good way to, as you say, incinerate time.

I thought this story was really interesting. It's from the Financial Times. It's called How a Billionaire's Dream for a Florida Railway Went South. It talks about this really unusual effort by Wes Edens, a Wall Street billionaire, to make a privatized railroad in South Florida.

Boy, Brightline is a fascinating company, and in the history of American passenger rail, a real outlier.

There's been a lot of coverage of this train. I think the Miami Herald calls it Florida's death train. So that gives you a sense of where this is going.

Brightline is a fascinating company in the context of American passenger rail, because Amtrak was formed in the 1970s as all the privately owned railroads were hemorrhaging money on their passenger service and trying to abandon it. The federal government basically took over their passenger services and took their capital equipment. Technically, the railroads own Amtrak, but it's operated with subsidies from the government. In reality, Amtrak will never turn a profit, and it will never be relevant that the railroads own its equity. But since then, the railroads that bought into Amtrak were prohibited from operating their own interstate passenger service in competition with Amtrak. Between that and the fact that the railroads would really rather not be operating passenger service at all, you really haven't seen anyone come along and operate a new mainline passenger service until Brightline came along. It opened the first segment of its line in Florida in 2018, and the full line, from Miami to Orlando, in 2023. It's a real standout in the world of American railroads, which are otherwise very effective, profitable, and efficient movers of freight, but have no interest in moving passengers. Brightline has taken a very unique approach.

I really love Wes Edens' dream. You have a lot of billionaires in the US that have their pet projects, like going into space or autonomous vehicles. Here is one that is looking at public transportation, maybe not super public in a way, but trains, infrastructure to move millions of people across the state.

That's right. Instead of being a billionaire who wants to live in 2300, he's a billionaire who wants to live in an alternate universe of 2026, where the US made different decisions about rail transportation in the middle of the 20th century.

So everyone in Florida could live a little more comfortably.

Yeah, exactly. A little more like Japan.

These trains really stand out because they're beautiful, clean, and new. You can get champagne and elevated meal service with the higher-end tickets. The experience of riding the train is universally understood to be really positive. Unfortunately, ridership hasn't quite been where they were projecting. They were hoping to have 8 million riders annually by this year. It looks like this year it'll be closer to 4 million. That's why right now they're in a bit of a financial crisis. They're trying to restructure. They have nearly $6 billion in debt, with interest payments due extremely soon.

It's hard to think of a more capital-intensive business than a railroad, especially a high-speed railroad.

Brightline also has its own safety crisis that really distinguishes it from the public railroads we were talking about earlier.

Yeah. So what's going on with its safety issues?

Since they started running in December 2017, they've had 185 fatalities, at least 148 of which are believed not to be suicides. That's 41 deaths in 2025 while running 32 trains a day. For reference, a really busy train like the Long Island Railroad had 6 fatalities running 947 trains a day.

Big difference. And overseas, high-speed rail is extremely safe. I think the Japanese Shinkansen system is notably a very low fatality system.

There have been some jokes about the Florida Man meme. There's a great quote from the article, though. But if the people of Florida were uniquely stupid in a way that made them more susceptible to being hit by trains, you would expect them to be hit uncommonly often by all trains. This is not the case.

Interesting. So what other trains are there in Florida?

There's also, for example, an Amtrak line that has a significantly lower fatality rate.

So what is it that separates Brightline from Amtrak? What could explain that difference?

There are a few theories. The first is that for many years, these same tracks have only been running freight trains. The freight trains are infrequent and also very slow. So there are a lot of unmarked crossings, and a lot of these crossings are at grade, which means street level. In a lot of places, you see the train tracks are elevated, which makes people pay more attention to the fact that there are train tracks. Here they're flat everywhere. There are some examples of people who were hit while crossing well-worn footpaths, because the train bisects their neighborhood and they just don't see the train coming. They're not familiar with there being such a super fast train coming down this way.

High-speed rail is such an expensive form of infrastructure to build, and a big contributor to that expense is the need to build viaducts that carry the lines over developed areas. You see it in California right now. It's one of the things totally draining the California high-speed rail project of its money, the need to build these enormous viaducts that ensure there are no conflicts between trains and cars or trains and pedestrians. In this case, Brightline is a value-engineered approach to high-speed rail.

Right.

As you said, it's existing train tracks that were owned by the Florida East Coast Railroad to begin with. These are the train tracks that in a lot of ways drove the development of Florida's cities. It was one of the first railroads in Florida, and it was involved in a lot of property development in South Florida too. So the train tracks necessarily go through the centers of a lot of very dense areas. Rather than build entirely new viaducts, which would've cost tens of billions of dollars, going off the cost of the project in California, the Florida East Coast Railroad and Brightline were able to introduce rail service slightly lower than the threshold for high-speed rail. Under 125 miles an hour, the federal government allows you to have tracks that are not raised above grade. You could say this is the outcome.

Yeah.

This is what you get if you want almost high-speed rail at a 10th or a 100th of the cost of what it would take to build real high-speed rail with totally separated grades.

People just seem to not have the time to see the train coming.

That's right. The accounts of people being hit by these trains are incredibly sad and very troubling, and they've contributed to a sense among a lot of people in Florida that this is a uniquely dangerous presence in their communities.

The Brightline East Line feels very distinctly Florida. Their ability to essentially own the tracks is extremely unique. They're looking to expand into Brightline West, which will go between Las Vegas and LA. I think they were hoping to have it done in time for the Olympics. It looks like there's no way that's going to happen, but that'll be a very different endeavor, even on the financing side, because you're not able to have that same kind of structure over the assets.

That's right. In the case of Brightline West, I believe that's a completely new rail line being built alongside Interstate 15, which goes from Los Angeles to Las Vegas. And they're also skipping out on what is really the most expensive part of building a railroad, building it through a city. Brightline West is planned to go from an empty lot south of Las Vegas, near the airport, to Palmdale, California, which is a considerable drive outside of Los Angeles, toward the north. It's also very expensive and difficult to cross the mountain ranges that ring Los Angeles. So it's in some ways a more ambitious project, in that they're building this railroad from scratch. In other ways, it's a less ambitious project, in that they're going from the outer exurban ring of one city to the outer exurban ring of another city.

I hope they're able to figure things out and survive and avoid bankruptcy, because it would be great to have more high-speed trains in the US.

Exactly. I would love to see more high-speed trains in the US. I'm also just fascinated by the place Brightline occupies in the economic history of railroads. There's no other railroad operator that's advanced passenger rail in a privately owned context as much as Brightline has.