This podcast is supported by the Icahn School of Medicine at Mount Sinai, an international leader in research, education, and patient care. The Medical and Graduate School is part of the Mount Sinai Health System, one of the largest academic medical systems in New York City. Ranked among the top recipients of NIH funding, researchers at Mount Sinai have made breakthrough discoveries advancing the health of patients. Here, clinicians and scientists push the boundaries in cardiology, cancer, immunology, neuroscience, genomics, geriatrics, environmental medicine, and artificial intelligence. The Icahn School of Medicine at Mount Sinai. We find a way. This is the Science Podcast for February 12, 2026. I'm Sarah Crespi. First this week, more than half of all dogs that go through service animal training don't make it to graduation. In this segment, producer Kevin McLean journeys with online news editor David Grimm to Canine Companions, one of the largest organizations in the U.S. for training working dogs. At the facility, they meet puppies in preparation and learn about the behavioral testing and genetics that could be used to improve service animal schooling. Next on the show, researcher Keisha Lloyd-Day talks about observing the birth of a stellar black hole in the nearby Andromeda galaxy. He recounts how his team looked for this elusive event and describes what we can learn from observing it in the decades to come. Years ago, I ran into my friend's mom while she was out walking her dog, or a dog. It wasn't actually hers. This incredibly calm lab was well beyond the floppiest puppy stage, but young enough that some excitement or jumping around wouldn't have been unexpected. I don't remember if I commented on how relaxed the dog was, but I do remember her saying that the dog wasn't going to make it. Not in a morbid sense, just that the dog was being trained as a service animal and this one just wasn't going to pass. I was so surprised that this sweet, well-behaved dog wasn't good enough for whatever it was being trained for, but it turns out this is a pretty common story. Despite all the time, energy, and money that goes into training new service animals, more than half fail. This week in science, online news editor David Grimm wrote a feature about how researchers are working on improving success rates to build a better working dog. When I heard he was going to be out here in California, I jumped at the chance to see Dave in action and to see some adorable puppy tests. Hi Dave, welcome back to the Science podcast. Hey, Kevin. This was such an interesting story. You know, I was really excited for a bunch of different reasons. One, that some of your reporting for this took you out my way. That's right. Just to start, we're talking about service animals, not emotional support animals, right? What are some of the jobs that these animals are being trained for? We're talking about animals that are specifically trained to help people with a variety of disabilities, seeing eye dogs, dogs that help people with epilepsy or autism or even PTSD. We are not talking about emotional sport animals, as you said, Kevin, or military or police dogs. Those are sort of a different category of working dog. The sort of demand for these animals has been going up recently, right? Is the training keeping up with that as well? No, I mean, and that's the big problem. I mean, there's millions of people around the world with disabilities that could be helped by these service dogs. And yet there's only about 35,000 service dogs in the world. And there's only a couple of hundred organizations that train these animals. And so at best, these organizations are maybe pumping out a couple thousand dogs a year and disabilities are growing. The rates of hearing loss are going up. The rates of other disabilities are going up. And so really, you know, demand is far outstripping supply at this point. How did you get started on this story? Is this something you'd been tracking for a while or where'd this come from? I think I'd been working on this on and off for maybe three or four years. I'm sort of the cat and dog guy at science. I had these Google scholar alerts for cat and dog stories. And I believe I'd come across an article with an intriguing headline, something about finding better ways to use science to create working dogs. I mean, I knew about working dogs. I knew about scientific research on cats and dogs, but I didn't realize that scientists were actually exploring this field of sort of how to improve the working dog pipeline. What are some of the ways that researchers are trying to improve working dogs? So Kevin, when you and I met in Santa Rosa, that was at Canine Companions, which is one of the largest organizations in the world that trains service dogs. I remember when we first walked into one of the kennel rooms with some of the adult dogs, and there were probably 12 dogs in there. And, you know, I've like walked into an animal shelter, any number of places where like a bunch of animals are, and you just expect to hear kind of this like cacophony of barking and noise. And all you really heard was like, there were some people cleaning, there was a little bit of music going, a dog occasionally would maybe bark once or twice, but it was just like these dogs were just, they just behaviorally are much different than a lot of the dogs that I'm around. 100%. We really decided many years ago when we really started to breed our own dogs as opposed to just getting dogs that were donated for using shelter dogs. Brenda Kennedy's the chief veterinary and research officer at Canine Companions. We started off getting a bit of a tour around the facility with her. We need to develop our own program. And so we chose Labrador Retrievers and Golden Retrievers and Crosses of the Two. So that's the first thing. She's been there for a while, so it was really interesting to hear a little of the history of the breeding program. But what are they focusing on now? They want to know. is it worth putting this dog through training? And training is a two-year process that can cost upwards of $50,000 per dog. And so it's a huge resource investment. These dogs have to be bred and then they're born and they're sent to volunteer puppy raisers for about a year and they come back to the school for more advanced training. And then about half of these dogs don't end up graduating to spend all of that investment. And so schools wanna know two things. They wanna know, can we determine when this dog's a puppy, whether it's even worth investing all of this time and resources. But also, is there a way to figure out what job this dog would be best at? Would this dog be better as an animal that's going to help a child maybe with autism navigate the world? Or is this dog going to be better for maybe a veteran with PTSD or for somebody who has a hearing disability? These are very different skill sets. And right now, it's kind of a bit of a guesstimation. These organizations conduct these behavioral assessments. They have people rank the dog on impulse control and, you know, social manners and stuff. And they try to come up with metrics based on those things. But again, even with all that, the success rates are only about 50%. We're trying something called cognitive test batteries. It's a more scientifically rigorous approach. And that's why I was out at Canine Companions just to see why I saw a few puppies, but the puppy I read about is this yellow, I think it was a golden lab mix. This is Twizzler. And all these puppies have interesting names because these organizations have bred hundreds of thousands of dogs and you want to sort of be able to keep everybody clear. And in fact, there's so many dogs that even have the same names that you'll find dogs like Tracy the Sixth because there've been, you know, five of the Tracys in the program. But anyways, with Twizzler, you know, I think he was about eight weeks old. So he was just starting his testing. This is more like a scientific experiment. A couple things I saw were they would put this clear cylinder in the middle of a room. It was sort of like a clear trash can on its side, so there's just one opening. And they put some treats in the middle. Okay so there treats in the container He sees them go in They let him go Then what are they looking for What are his options here Does he run straight towards the trash can and crash into the side like a lot of puppies do, which shows a lack of impulse control, which is bad? Or is he able to quickly learn that, like, okay, there's one opening here, and no matter how they turn around the trash can, I'm always going to be looking for that opening. And that's something that Twizzer did, so that was really good. Something else he did was confronting what's called the impossible task, which has been used for a long time in animal behavioral studies. And basically the researchers will put some treats in like a Tupperware container and seal the lid on. And a lot of animals will just sort of keep on trying to get at that treat. And dogs sort of famously give up very quickly and look at humans for help, which is sort of chalked up to their intensive co-evolution with us. Whistler, do you want that treatment to help you? Good boy. So that was actually a really good demonstration of the other strategy. Yeah, so 4.8 seconds. They've sort of learned over 10,000, 15,000 years that we are partners. That's what Twizzler did. Twizzler sort of looks up at Emily Bray, who was the scientist who was running this part of the study. And he looks at her for five seconds, which is a long time. It's a communicative gesture. It's sort of Twizzler saying, like, I need your help with this. Places like Canine Companions are really looking for that communication with dogs. They want dogs to be really tuned in to sort of the mental state of their handlers when they're being trained or when they're out in the real world with the people that they're trying to be helping, that there is this sort of two-way communication happening. And all this stuff, especially with the cognitive test batteries, it's really in its early stages. But the goal is, can scientists come up with a handful of tests that are very objective? The scoring is very precise. These can sort of be repeated at a school anywhere in the world because everybody's using sort of the same setup and the same metrics. And can those sort of spit out two scores, one like a master score that says like, this puppy is worth training. This puppy should continue in the program. But maybe a more granular score that would say like, not only is this puppy worth training, but the test results are showing us that this puppy would be especially a good hearing dog or PTSD dog. And so we're not there yet, but that is sort of the promise of cognitive test batteries. How far along are they in this process with the canine companions? They've got scores for puppies. Have they seen how any of the adults have done so far? Well, Emily Bray, she's analyzed about a thousand dogs so far. You know, you've got to follow these dogs for at least two years to figure out, like, are the dogs we tested as puppies? Are they actually graduating? Are they able to do the jobs they were trained for? Once they go out with their handlers, and this is sometimes the most devastating, these dogs will sometimes graduate, and either for behavioral reasons, like impulse control, health reasons like elbow or hip dysplasia, they get placed with somebody who's sometimes been waiting years for a dog. And then the dogs fall apart at that point. You know, they'll go out into like a busy street and they'll get really stressed out or they'll develop these sort of fears and anxieties that didn't come up earlier in the program. And so the goal is like, how far can we follow these dogs to figure out how well these scores are impacting not only their ability to graduate, but their ability to stay as a long-term working dog. And so this stuff is still being crunched right now. These scores aren't right around the corner, but again, that's the hope that maybe in a few years, these organizations will start to have these much more reliable scores that they can use for these animals. You've been following this story for quite a while. What surprised you as you were actually visiting these facilities and reporting it out? Well, I was not surprised about how cute the puppies were. They were very cute. And that's, I mean, that's honestly, you know, one of the best parts of this job is to be able to go and see puppies and interact with puppies and even the adult dog. And I was just amazed to be able to see these dogs pull people in wheelchairs across the room, know what it means to open a medication drawer, to wake a person up in the middle of the night. You know, I saw these trials when an alarm, not only when an alarm blares, but to know, okay, like it's a doorbell. So I'm going to lead my trainer to the door. It's a fire alarm. I don't want to lead them towards that sound because if that sound is a fire, then I should be leading the opposite to be able to tell the difference between the different sounds. And when I was at Guiding Eyes, they actually put a blindfold on me. Right. So you visited another site where I sadly couldn't join you out in New York. This is Guiding Eyes for the Blind, which you mentioned is one of the largest seeing eye dog schools in the US. What was that like? This didn't make it into the story, unfortunately, but they put a blindfold on me and they had a seeing eye dog take me around this obstacle course. And for me to be able to sort of have my eyes closed and put 100% of my trust in this dog, and you know, there's tree branches I could have smacked into, there's curbs I could have slipped off of, and they put 100% of my trust into a dog. It was really scary. These animals were just incredible. The human life that they're really taking on, they're really 100% responsible for the safety of this person. And for me in this case, it was just really remarkably able to experience that firsthand. They're paying really close attention to the pedigree of each animal and everything too. I like, are they thinking that if the parents perform well in all these things, then the puppies will also as well? Or is that part of the process? That is a sort of a different approach called estimated breeding values. To your point, Kevin, it's more pedigree-based. One of the issues with the cognitive test batteries is you already have these puppies. So the puppy's already born. And at that point, we're trying to figure out, okay, well, is this puppy going to be a good service animal? And what Guiding Eyes is trying to do is like, well, can we optimize the breeding of these animals by knowing details about the pedigree and the genetics of the parents and other close relatives to know, like, should we even be breeding these two dogs because our data maybe shows that breeding this male and female tends to produce puppies that are not going to be successful. So that's the goal of estimated breeding values. It's really cool because it was ported over. The livestock industry has been using estimated breeding values. There's these statistical calculations that are really based on you have maybe this behavioral and health data from an animal and all maybe the hundreds or thousands of animals that are related to it. And it sort of spits out this score that tells you what are the chances that this parent is going to pass on a particular trait to its offspring? For example, is this cow going to be a good milk producer or a meat producer? So it's a very sort of economically focused. It's only been maybe in the past 20 years or so that scientists have been trying to port this over to dogs to figure out, can we use these estimated breeding values or EBVs as they're known to figure out like, well, is this dog going to be very resilient? You know, is this dog going to be able to go out into the world? Especially seeing eye dogs, it's such a stressful job because not only do you have to listen to all these commands from your owner, but you have to be paying attention to all of these environmental cues. Your handler may be telling you to cross the street and these dogs have to be able to be disobedient in some cases and say like, no, we can't cross the street right now because there's cars coming. So there's all of these different things that these dogs have to sort of take into account. It's very stressful. That's also another reason that a lot of dogs either fail out of these programs or even sometimes get placed with handlers and then don't make it very far as seeing eye dogs. And so the question is, you know, can we find parents that are going to be more likely to sire puppies that are going to be more resilient, be more able to take commands so that, again, we can improve the efficiency of the pipeline? I know you mentioned the canine companions are sort of early stages with the behavior side, but how has it been going so far with this estimated breeding value? So Guttin' Eyes has been using EBVs for about 20 years now, and they've actually already had some success with it. Jane Russenberger, who's their former head of breeding over there, has been working with two health conditions hip and elbow dysplasia This can cause a lot of stiffness in these animals It a big reason these animals drop out And she just using EBVs alone she was able to reduce the incidence of dogs with help and other species being born to almost nothing And when she first started at Guiding Eyes, like only one in five puppies that were born became seeing eye dogs, which is really low. And you imagine all the wasted resources. And now it's about halves, which isn't great, but it's a lot better than it used to be, thanks to EBVs. But the real goal is to get the EBVs even stronger because right now EBVs are sort of based on somewhat subjective behavioral assessments. They're based on pedigrees, which can be a little bit mushy. And so the next step with EBVs is what's called genomically enhanced EBVs or GEBVs. And this is basically trying to infuse genetic information to the equation as well. You actually have genomes and alleles and really sort of hardcore genetic data. And already the idea is that by using GEBVs, you can sort of double the accuracy of EBVs. Something that might take you multiple litters to figure out there's a problem can be done potentially a lot less litters. And another big promise of GEBVs is there are some conditions that can take years to pop up. So something like hemangiosarcoma, which is a really deadly blood cancer in dogs. I think it kills about 90% of dogs within a year of it being diagnosed. It's a huge problem because you've gone through all this training and this dog develops this disease, but it happens so late in dogs life, you're never going to pick it up when you're training this dog. But now you have this genetic information. And so one of the ideas is for a lot of these late onset conditions, GEBVs could really help predict these much earlier. And again, ideally before the dog is even born. So you're not breeding a mom and dad that have a predisposition to this condition. There's sort of this behavioral side. there's this genetic side or the breeding side. Are people combining both of those as well? Yeah. I mean, I think in an ideal world, you don't just focus on one or the other that you could potentially combine both. Now, these things are themselves very resource intensive, you know, getting genetic information on dogs, just dogs in the general population. So you can even figure out what genes in the first place you should be focusing on. And that's one of the big efforts you wrote about in your story called the Working Dog Project. Right. Which is trying to gather tens of thousands of genomes from dogs all around the world. These can sometimes be service analysts, but also just pets as well and trying to get the genetic information that you would even need to really be able to fine tune these GEBVs. If you can get that, if you can really sort of optimize these cognitive test batteries so that they're really accurate and we know which test to use. Because at the end of the day, the goal is, can we have the best puppies being born? Can we make sure that once these puppies are born, that we're investing only in the puppies that are most likely to graduate? And also, you know, can we figure out which jobs these dogs can be best at? And all of these scores, you know, whether it's cognitive test batteries or GEBVs can potentially help feed into that. And you may never get to the point where we can say with 100% accuracy, like this dog is going to be an amazing seeing eye dog. But even if these approaches can increase the accuracy 10 or 20 percent, that is still a lot less time and effort that these schools have to put in to dogs that aren't going to make it. And they can put that time and effort and money into dogs that are more likely to make it. Well, thank you so much, David. This is so interesting to talk about. Thanks, Kevin. David Grimm is the online news editor at Science. You can find a link to the story we discussed at science.org slash podcasts. Hi, Science Podcast listeners. This is Kevin McLean. I'm one of the producers on the show. I just wanted to hop in here before we get started to ask you to consider subscribing to News from Science. Every week on the podcast, we bring you one of the stories that the News from Science team has published, but there's so much more than what we can cover on our show here. For only about 50 cents a week, the money from subscriptions goes directly to supporting nonprofit science journalism, reporting on science policy, investigations, international news, and the latest breakthroughs from all around the world of science. Support non-profit science journalism with your subscription at science.org slash news. You have to scroll down and click subscribe on the right side. That's science.org slash news. We are continuing the experiment this week and sharing more items from around the website. We're talking news, research, careers. Speaking of careers, I wanted to call out that in this week's career section, we have a working life essay by Anne Trollard. Trollard's a staff scientist, and she recounts the problems she's encountered getting recognition for the work that she does. She's often a key player in analyzing data, in implementing projects, but she sometimes gets overlooked as a member of the team who doesn't have a PhD or hold a faculty position. It's really interesting to think about and worth a read. Next worth a mention is a science paper by Jared Westbrook and colleagues. They talk about the genomic approaches to accelerating American chestnut restoration. If you didn't know, American chestnuts are virtually extinct due to the widespread impact of blight. This paper looks at crossbreeding programs with Asian chestnuts and found some sound strategies for breeding resistance into the American trees. Last up, I have something from the news site. This is a story on a puzzling vaccine side effect by Gretchen Vogel and Kai Kupferschmidt. These vectors, called adenoviruses, have been used in vaccines like Johnson & Johnson's COVID vaccine successfully, but there have been some side effects like blood clots. They write about a new study that identifies the part of the virus that may be responsible for these reactions. Okay, that's it for the roundup, but please do stay tuned for our next segment on how to spot a massive star becoming a black hole. When stars die, they collapse on themselves, and the shock from that collapse can either cause an explosive event, like a supernova, or if there's not enough energy for the star to shake free from its outer layers, material falls back in on the star, creating a black hole. We know these steps, we know there are black holes around, small ones, big ones, but we've never actually seen a detailed step-by-step formation of a black hole from a star. This week in Science, though, Keisha Loy Day and colleagues report on what looks like just that, the formation of a black hole. Hi, Keisha Loy. Welcome to the Science Podcast. Hi. Thank you for having me. This is not an easy observation to make, right? We're looking for basically a tiny point of light, a star that goes away in a specific way. How did you go about looking for a star turning into a black hole? It's remarkably difficult, as you said, to find an individual star disappear in a galaxy that typically has billions of stars in it. So instead, what we leveraged was this longstanding theoretical prediction that told us that even when a star fails to explode and implodes into a black hole, There is a small amount of energy deposited into the star's envelope that should cause it to brighten by a tiny bit, making it brighter for a few months than what it was originally. And we were looking for these tiny blips of light that might accompany these black hole formation episodes. And these blips are most effectively searched for with infrared light. And that was the novelty of our search. We are looking for them in infrared light. So you're looking across years, decades of observations and billions of stars, and you found this in the nearest spiral galaxy to us, Andromeda, sometimes known as M31. Did you have an idea about how common such an event would be for stars in that galaxy? Is this a rare thing to happen or we just don't have a lot of these types of observations that we need to see a black hole forming? Long story short, we know very little about how black holes form, which stars form black holes, or even as questions are simple as how frequently do they form. So when we went out to do this project we were not even looking for this specifically Our goal was to do more of a census of understanding how do stars change in infrared light And this just happened to be one of the most remarkable objects that we came across Because you can see different things happening with infrared observations, but this little, I guess you'd say it's like a rise in infrared. And then guess what? It just starts to slowly decline until you can't see it really in optical at all, right? Absolutely. And the remarkable part was that, you know, you can imagine how shocking it might be for someone looking at data. You see there is this star that brightened infrared light. And suddenly you realize that, wait, it's actually gone. There's no evidence of it anywhere now, even with the deepest images from the Hubble Space Telescope. Okay. Can we do a little timeline of the formation of the stellar black hole? What time points do we have and what was seen at these different time points? All of this was done with archival data and that it all happened in the past. What we see in the data tells us that somewhere around 2015, 2016, there was this star that was very well known in the Andromeda galaxy for decades. It seemed to brighten in infrared light by maybe 50% or so. That's what was flagged by our analysis when we were looking for these variable infrared stars. And then over the next three years or so, it brightened and it faded, came back to its original brightness and then it kept fading below its original brightness in infrared light. But at the same time, what it did in the optical was even more remarkable in that it actually faded by a factor of 10,000 in brightness over the course of a few years. Wow. That's absolutely abnormal for stars. Today, what we know is that the source is absolutely undetectable in optical light. It's extremely faint. There is a faint infrared glow at about a tenth of what the star used to be. So physically, what we believe this is associated with is that as this star was dying, it's sort of in this dying gasp of trying to, in its final moments of its life, it ejects its very outer envelope. And this envelope travels out, becomes cool and shines in infrared light. And all the while, the star inside has actually died. No more fusion. No more fusion. Do you think that if we keep looking at it, it will plateau at a certain point? Or will it just keep fading in all the spectra? Nominally, we would expect it to keep fading over time now because it's already been about 10 years since the collapse. The fading rate becomes slower and slower over time. So we would expect maybe to keep fading over now decades, possibly centuries, until it essentially disappears into darkness. That's kind of your definition of a black hole. That's so cool. But there are possibilities that it might plateau, you know, because we know fairly little about how black holes behave in these very messy environments. And it is possible that the process of gas falling into the black hole hits a barrier somewhere. And all of that will be like research that it will be done for decades to come on this source. So you used a bunch of data sets to kind of corroborate this across the different spectra. and it fit the profile that you went into saying, okay, well, we know a black hole is going to do this in the infrared and in this and the optical. But what about the details? Like, did you learn new things about the process of black hole formation from seeing this in Andromeda? Absolutely. I think there's so little we know about how stars turn into black holes that, you know, when we first found this object, you know, this was only a hypothesis initially. Hey, maybe this is that. But then what we realized was that turns out that there was a previous star like this that was reported about 10 years ago, which had disappeared in optical, except it was 10 times further away. So the data on that was not as good quality. What we realized was that this star looked exactly like the other one. Even in that case, people had proposed that maybe it turned into a black hole, but it was a bit trickier to claim that because of the lack of good data. But here we were like, we had such good data that not only could we show the similarities between the objects, but also sort of construct a theoretical model that incorporates all of the physics we know today to beautifully explain why the star behaved the way it did. One thing that's kind of mysterious, I don't really understand, is why when a star collapses, it could turn into a supernova or it could turn into a black hole. Does what you see here kind of explain how that decision, I'll put in quotes, is made? I think what we see here actually makes that decision even more complex than we thought in the past. Back in the day, common wisdom was that the most massive stars, stars that are, let's say, 20 times more massive than the sun, all of them, they implode into black holes, whereas the lower mass massive stars, they might explode as supernovae. What we are seeing here is clearly a star that's nowhere as massive. It's something like 13 solar masses. If you asked an astronomer 10 years ago, would this star explode? They would be like, absolutely, yes, this must explode. And what we are seeing now is clearly a star of this mass that disappeared, never exploded. And what we show is that the best explanation is that it turns into a black hole. So it really changes in many ways the paradigms of astrophysics that we've adhered to for the last, I'd say, three decades. Are you sure you didn't just miss the bright flash? It didn't explode. It was too quick for you to catch. One of the neat things about this object is that it's in our nearest galaxy. If this object went supernova, it would have been a naked eye object. That's how bright it would have been. The fact that we did not see it really means that even with, obviously, lots of sensitive telescopes observing from the ground, that it really wasn't there. We've basically ruled that out. So now that you have kind of this more detailed information about how black holes form, is it going to help us find more of these like formation processes to help us look for them? Or are we going to just keep staring at this spot to learn more? Are we doing both of those things? I think a combination of both indeed. I think one of the really nice things about this object being so nearby is that this source, which is currently barely detected in infrared light with the James Webb Space Telescope, this is going to be detectable at our current sensitivity for decades to come. So not only did we find this, but we now have this opportunity to keep looking at how this star evolves into this something like a naked black hole over the course of decades, maybe centuries. So that's one avenue for it. The other part is that because it's so difficult, it's not common for us to look for stars that disappear. When a star explodes, it outshines its entire galaxy for a few weeks. It's very easy to find exploding stars. You can see them in galaxies millions of light years away. Supernovas are so flashy. This really teaches us that we should be looking for the stars that disappear, not only the stars that brighten. That is very cool. Does this tell us anything about supermassive black holes, about their formation or what's going on with them? The picture is that supermassive black holes were seeded by these stellar mass black holes very early in the universe. So imagine you have the first stellar mass black holes that might form at maybe five times the mass of the sun, 10 times the mass of the sun. And over time, they keep accreting gas onto themselves. They keep merging into other black holes. And over time, they grow into these supermassive scales. So the fact that we are finally getting to even scratching the surface of understanding how stellar mass black holes form is a step in that entire chain, because we really know very little in that entire process. First snowflake in the snowball, right? Kishaloy, thank you so much for talking with me. Thank you for having me. Keisha Lloyd-Day is an associate professor at Columbia University and associate research scientist at the Flatiron Institute. You can find a link to the science paper we discussed at science.org slash podcast. And that concludes this edition of the Science Podcast. If you have any comments or suggestions, write to us at sciencepodcast at aaas.org. So, for example, if you like our roundup or don't like our roundup, drop us a line. To find us on podcasting apps, search for Science Magazine or listen on our website, science.org slash podcast. This show was edited by me, Sarah Crespi, and Kevin McLean. We have production help from Podigy. Our music is by Jeffrey Cook and Wen Kui Wen. On behalf of Science and its publisher, AAAS, thanks for joining us.
