Hard Fork
Hosted by Casey Newton and Kevin Roose
The companies changing our phones, our media and our meat.
13,560 words · 6 speakers
Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code. A system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T dot com.
I, um, I've been having a rough week, I have to tell you.
What happened?
Well, there's this joke that I want to use that I can't use. It's driving me crazy.
What's the joke?
Okay, so as we record this, as you probably know, Chairman Xi Jinping of China is in San Francisco.
Yes.
He's having a meeting with President Biden. It's this really big deal. So I thought, what could be better while Chairman Xi is in San Francisco than to change my Grindr bio to Chairman Xi?
Like, just imagine, you're scrolling through the grid. You want to see who's cute in the neighborhood. And then you come across the official Communist Party portrait of Xi Jinping. And the bio just says, "Visiting for work. Looking for cuddles." I, if I saw this, it would make my whole life. And so I want to give that somebody that'll set experience. But I can't because the platform has a rule against impersonation.
No.
Yes. And I was like, as much as I like this joke, I am not willing to give up my Grindr account for it.
Well, and isn't Grindr now owned by a Chinese company?
Well, it was, but the US got very concerned about this because there's a lot of what I guess we could call sensitive data being exchanged on Grindr. And they legitimately worried that it would have like national security implication if you had a bunch of US military service members running around the world with Grindr on their phones while it was owned by a Chinese company. So they had to spin it out.
Well, I'm sorry that your Grindr gag didn't work. But I am happy that you're now definitely on our no-fly list for attempting it.
I'm Kevin Roos, a tech columnist from The New York Times.
I'm Casey Newton from Platformer.
And this is Hard Fork. This week on the show, there's an AI pin that has the tech industry talking, and I tried it. Then YouTube opens its doors to deep fakes. And finally, SyFy Foods CEO Josh March serves us a little Thanksgiving dinner that was grown in a lab. We had to sign a waiver.
We did.
Casey, a couple of weeks ago on the show, we talked about these AI wearables that have been announced over the past few months. And recently, we got a big update on one of them. Just last week, the Humane AI pin was launched and demoed, and you actually got to see it up close, I understand.
That's right. It was the sequel to the Inhumane AI pin, which committed many fewer human rights violations, and I appreciated that. But no, listen, imagine you have $700, and you want to answer for yourself the question, what if Siri was good? OK, you might be interested in the Humane AI pin. This thing, it's a little computer, and it has a little magnet, and you snap it together, and you wear it on your shirt or your jacket. And I know what your next question is going to be, which is, well, what the heck does it do?
What the heck does it do?
Well, I'll tell you. So it's got a camera, can take pictures, OK? If you want to talk to it, you tap it. You can ask it questions that you might ask an AI. It has a speaker, what they call a "personic" speaker. That's a combination of personal and sonic. And this speaker will talk to you. You could say, hey, what's going on with the weather? Or give me some ideas for a recipe. Or is there a grocery store around here? And now you don't have to get your phone out of your pocket. You just tap your little pin, and you move on with your day. Oh, and one more thing, Kevin, because I know what you were going to say. You were going to say, Casey, did they build a laser projector into this thing so that you could project a user interface onto your palm in lieu of a screen? You were about to ask me that.
I was. It's true.
Let me tell you. They built a laser projection system, okay? So for the first time in your life, I know every time you're looking at a screen on your smartphone, you're saying, why isn't this just projected onto my hand?
Exactly.
The Humane people answer that question.
Okay, here are the other things we know about it. It costs $699.
It's available in three, what they call colorways. Why this company chooses not to use the word color and opts for colorways instead, I'll never know.
Colorways is Silicon Valley speak for colors, okay? We have a unique culture here and our own language.
So you also have to buy a data plan that costs $24 a month on the T-Mobile network. And it gives you a new phone number and you can use it to send and receive text messages. You're a green bubble if you use the Humane AI PIN.
Green, which is the color of life, Kevin.
So you get this PIN. It is sort of being billed as not necessarily a smartphone replacement, but something that may over time come to do more of the things that you would currently use your smartphone for.
Yeah, I think when these folks started at Humane, they were very curious about what would be next generation hardware. People have very different answers to this question, right? The Meta people think it's going to be some kind of helmet or glasses that you put on your face. The Snap people think it's definitely glasses. The Humane people said, we're going to build a PIN, but everyone is trying to figure out, well, is there kind of something beyond the simple smartphone? And maybe that thing doesn't have a screen.
Right, so that's one sort of reason for this device to exist is that I think everyone in Silicon Valley is thinking about what comes after the smartphone? What is the sort of logical next step? And a lot of companies are also thinking about, well, these AI tools, they're very powerful and very cool and very potentially useful, but using them on a smartphone just feels a little bit anachronistic. Maybe there's some device that should be custom-built for this. And so this is, I would say, the first major release that we've seen that tries to answer that question of, like, what would a device that was built for AI look like?
Yeah, and I think it also tries to answer the question of, like, if AI gets good enough, maybe you don't need a screen anymore. Maybe you don't really need apps anymore, right? Maybe it is just a purely conversational interface that does whatever you want it to do.
So I was not invited to the launch of the Humane AI PIN.
What did you do to those people, Jesus?
And but I did watch the launch video. Very strange launch video, I gotta say. It takes place in this like empty office with nothing on the walls. It's like two people who sort of look like they're doing Steve Jobs cosplay.
It's called minimalism, Kevin. Okay, the style has a name.
And I would say it's like a kind of low energy presentation of what they say will be like a world changing device, which I was a little confused about. But aside from the aesthetics of the video, there were a few things that I did wanna like flag as being potentially cool, because I don't wanna just crap on this idea out of the gate.
Well, you kind of do.
Well, I have some questions about it, but I will say there are a couple of things that struck me as pretty cool. One is, the device itself looks cool. It's like a very sleek sort of iPhone looking device, which makes sense, because a lot of the people at Humane came from Apple. And I thought that a couple of the features were cool. One is, it can summarize your text messages to you over voice, which is always something. I don't know if you've ever tried to use Siri or something while you're driving to like send and receive text messages, but it is infuriating.
Yeah, absolutely.
Because it reads, it will read like not only the text messages that you want, but like also the like, the security code that you got when you tried to log into your bank. It'll read that to you. You know, it'll read like all the emoji like tap backs to your messages. It just doesn't really do a good job of getting you the information that you're looking for in a succinct way, but.
Yeah, it doesn't seem like the people who built the feature use it, is I think how I would describe it.
So but this this humane AI pin, you can ask it, you know, catch me up. This was something that they showed off in the demo video, and it will sort of summarize using AI all of the texts and emails that you've gotten since you last asked.
Yeah, and by the way, I will say this is a very good pitch for the sort of early adopter tech crowd, because what could be more flattering to your own ego than the idea that you're receiving so many messages that you actually need an executive summary of the messages that you received? It's like, yeah, let me, let me dig into the details. It's like, oh, well, your friend shared another meme, and that was your executive summary of your messages.
Totally. So what other cool feature I thought was this instant translation feature, which was in this demo video, where they basically. Show two people talking, one of them is talking in Spanish and the other one is wearing one of these humane AI pins and they can just sort of tap it and it will instantaneously take that Spanish that someone is talking and convert it into English and speak it out loud to you in English in sort of something that resembles their voice.
Yeah, then that could be a fun thing to take around the world and meet new people and talk to people you might not otherwise be able to talk to.
Totally, like ordering at a restaurant in a foreign country, you could just tap your little pin and it could sort of translate what you want.
There you go.
So those are some of the cool things. I do have some questions about this device, though, because I would say the demo that it sounds like you and some other reporters got was fairly limited. Like, people weren't actually allowed to try this on for very long. My colleague, Aaron Griffith, was able to try it for like 10 minutes, which is always kind of a red flag when a company releases something and is like, we can use it, we can show you, but you can't use it yourselves. I always kind of wonder, like, what are they hiding?
Yeah, yeah. Let's just say that reporters were never asked to try the Theranos for themselves before that whole thing went down.
So here's one of my questions. Is the experience of using it or seeing it used, did you feel like this is actually a step forward for computing? Or did it just kind of feel like a new gadget that was kind of cool, but maybe not all that much more useful than your smartphone?
Yeah, I think there are sort of two different ways of using this thing that I have very different feelings about. There is the laser projector thing, right? And when I was watching the humane employees show this to me, there was part of it that was just quite fascinating because they have developed these gestures, like to zoom out to the home screen, you pull your hand back from the projector. And to do this picking, you make these gestures, like pinching your fingers together. And it's like watching someone who has learned a new form of sign language to interact with a computer. And there is just kind of a weird spectacle in that. At the same time, I looked at that and I thought, this thing does not do enough to get me to learn a new language yet, you
know? So the projector stuff, it's unbelievably cool just from a technology perspective. I don't think it is going to be the easiest way to do anything that they showed me, right? But then there's a second question, which is... Do I want to wear AI? And I think there are cases where the answer is probably definitely going to be yes. Right? We've seen in sci fi, a lot of the way that creative types have been predicting an AI future is that you wear an earbud and you're just sort of able to converse with that all day long. Her, the movie from 2013 by Spike Jonze, the most famous example of that. We talked on the show about Mrs. Davis, a show from this year that has the same kind of metaphor.
And I do think people are going to be doing enough computing during their day that the ability to just kind of like tap a thing on their ear and be like, hey, like, give me directions to the grocery store. Or, hey, you know, what emails have I gotten since I was in that meeting? That makes a lot of sense to me. Now, this is a brooch that you put on your chest, not a thing that you put in your ear. But 1.0 hardware is almost always bad. Right? Like, I, did you buy the first iPhone?
No, neither did I. Right? And yet, we could both agree, the iPhone was a really good idea and they got there. So to me, the interesting question is not like, should everyone go out and buy a $699 device with a $24 a month plan? Because the answer for most people is obviously going to be absolutely not. To me, the interesting question is like, well, is there a direction here? Is there a path to something? And does the path wind up being AI on your chest? Does it wind up being AI somewhere else? Or do we have it all wrong? And we really just are going to use smartphones forever? But if I had to make the bet right now, I would say that yes, there is something beyond the smartphone.
Yeah. Speaking of that path, I want to bring up one more thing about this company that I just find totally fascinating and entertaining, which is their origin story. This was a story that my colleagues, Aaron Griffith and Trip McGold wrote in the New York Times about Humane, and it is truly my favorite detail about this company, which is that it owes its existence to a Buddhist monk who goes by the name Brother Spirit. Did you read this story?
I did, I did. Was very interested to learn this. It didn't come up during the presentation, I'll say that.
So I'll just read this paragraph to you. Quote, "A Buddhist monk named Brother Spirit led them, the founders of Humane, to Humane. Mr. Chowdhury and Mrs. Bonjourno had developed concepts for two AI products, a women's health device and the PIN. Brother Spirit, whom they met through their acupuncturist, recommended that they share the ideas with his friend, Marc Benioff, the founder of Salesforce." A more San Francisco paragraph has never been written in the pages of a major newspaper, I would submit to you.
I love that. I mean, look, Marc Benioff, he's at the center of a lot of stories in San Francisco. You know, he's a man about town. He's making connections, he's wheeling and dealing.
Brother Spirit, I need like a 3,000 word profile of Brother Spirit and how he's become the tech advisor to Silicon Valley.
Let's get him on the show.
Yeah. So, since this launch, there has been a lot of hubbub. People are saying, this is a great idea. I'm very excited about this. Other people are saying, this will never work. Who's going to pay $700 plus $24 a month to wear a thing on your shirt that can't even do the things that your smartphone can? And then there was, there were some funny little details in the promotional material for the launch itself that turned out to be wrong in the ways that AI products sometimes do get things wrong. So, there was a part in the video where they showed off this AI pin being used to take the nutritional contents of a snack that the, one of the founders was preparing to eat. So, he taps his pin
and is holding some almonds in his hand and he says, how much protein is in these almonds? And the humane AI pin uses its camera to analyze it and says there are 15 grams of protein in these almonds. People who watched this video later pointed out that it would take about 60 almonds to get 15 grams of protein, so many more almonds than the small handful that was shown in this video. So, if you are relying on the humane AI pin for your nutrition facts, you may be getting things wrong.
You might die of malnutrition.
Another discrepancy with the video, they showed off the AI pin being asked where the best place to watch the next solar eclipse would be and suggested watching it from Australia. Turns out, the upcoming solar eclipse is actually not going to be visible from Australia.
Well, I thought maybe the pin was just having fun with you because imagine you book a trip to Australia to get down there and then the pin is like, "Waka waka!" But you know, something similar happened with the Google Bard launch, where there was like a screenshot that contained this factual error. So yes, I think if we've learned nothing else from these AI launches, it's that you really want to get a fact checker for your promotional materials.
Totally. So aside from these sort of small details and some of the skepticism around this product category in general, like, did this strike you as an AI wearable that you would actually want to wear?
No, I wouldn't. But at the same time, I do want to try to bring peace to Silicon Valley. Because as you said, there are really two camps here. There is the camp of folks, particularly in the sort of hardware reviewer crowd, that are just extremely skeptical, saying, "Nobody wants this. This thing is going to flop. It's the next Juicero." And then you have people that I would just describe as like technologists, people who work on product, who are engineers. And they're looking at this and they're like, "There is an insane amount of cool technology in there." So I just want to say, like, both of these people are actually right. It is both true that most people should not buy this device, and it's true that there's some incredible technology that they built in there. Now, what does this mean for the humane company? I don't know. Hardware is a very, very difficult thing to get right. They're going to need to hope that they sell enough
or are going to be able to raise enough additional capital that they can make a version two and three and four and hope they're able to like find that product market fit with what people actually want. But in terms of like, how good a start are they off to? Hmm, I don't know. I gotta say, probably a C minus. I would have to give this whole thing.
I will say, like, what this demo and hearing about this AI pin really made me wish for was not an AI wearable, it was a better Siri. I would give so much money for a Siri that could actually do the things that I want Siri to do, that could do this kind of instant translation, that could summarize my text messages and not just read every little emoji out to me. That is what I want from Siri. But I will also say, like, I think there is something to this idea of the screen. Right? Screens get a bad rap.
Mostly from you.
Mostly from me. But screens have one thing going for them, which is that they are a very dense way to consume information. I don't know if you've ever tried to order something from an Alexa device. Have you ever tried to do this?
I would never even try, honestly.
So I haven't tried in a long time, but that's because when I did try, back when this feature first came out, you know, they would say you can reorder dog food through your Alexa device and you would try doing this and what it would spit back was a list that would take about two minutes to read. It would say, okay, we've got Purina kibble. It's a 30 pound bag for $46. We've also got Iams kibble and it's got lamb flavor and chicken flavor and it's, it comes in a 28 pound bag. And it was like five minutes later, you have this list of things that you can order from. Whereas if I'm doing this on my screen, I can consume all of that information very quickly, select the dog food that I want and move on with my life. Audio is just not a very information dense medium. And
so yes, I think there will be times when you want to talk to an AI and have an AI talk back to you, but I do not think that that is how we will go about doing things like ordering products or buying plane tickets or anything like that, because it's just, it's just so slow.
You're totally right. And I do think the question of what if Siri were good is a good one. I mean, I think another very potential outcome here is that Apple buys Humane. Like after the first round of sales, don't go too hot, but they built all this cool technology that is often a time where hardware startups look around and say, well, who might this hardware be useful to?
Totally. Okay. That's the Humane AI pin. We will keep tabs on this category and this story going forward. And Humane, if you want to send us some demos to use, we will, we will try them out.
Until then, we'll put a pin in it.
Hey.
When we come back, YouTube opens its doors to deep pics.
Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T.com.
Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically integrated AI cloud platform, delivering data centers, compute and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling and predictable pricing. Iron, AI cloud for the next big thing.
It's time to innovate at the speed of AI without compromising control. That's OutSystems. Take a look at what happened with one global retailer during a period of massive expansion. They used OutSystems to build AI agents to automate invoicing and translation processes, shifting 95% of their daily store operations to a new digital app, saving thousands of hours of manual work. OutSystems gives you the flexibility to deploy how you want and the governance to do it safely. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.
All right, Casey, speaking of AI, we had some other news this week about AI and content moderation. You had a newsletter this week about YouTube and some policy changes that they have made to deal with AI-generated content on their platform.
That's right. And if you're already saying, I don't know, Kevin, that sounds a little boring. Here's how I would frame it. If you have spent the past year wondering, when is the flood of deep fakes truly going to arrive and make it difficult for me in many cases to tell what is true and false? Well, we are getting ever closer to that day, my friend.
Totally. So let's just recap what happened with YouTube this week and then talk about what it means.
So on Tuesday, YouTube put up a blog post outlining what it calls its "approach to responsible AI innovation." And in practice, what they did was give people permission to post a lot of what I like to call synthetic media, what are often called deep fakes. These are videos that have been created using generative AI or their video that maybe you shot it on your camera, but you use some sort of AI tool to. To manipulate it.
Right. And a lot of platforms have been coming out with these kind of generative AI policies, trying to wrestle with what do we do when people start making fake videos or posts that are manipulated in some way using AI to claim that something happened that actually didn't happen. So we're starting to see platforms starting to grapple with this. But what YouTube did is not just throw open the doors to deepfakes, right? Because they actually did add some disclosure requirements and require creators to label videos that show what they call realistic, altered, or synthetic content. So what did they actually announce? And then let's talk about what it sort of means between the lines.
Yeah, well, so as you point out, there are these disclosure requirements. So if you are going to upload a synthetic video and it looks realistic, whatever that means, we don't have details on that, you are going to need to tell YouTube when you're uploading the video. And YouTube is then going to put one of two labels on it. If you're doing something that's sort of silly and fun, like maybe you use an AI tool to create like a dog chasing a cat and it's very cute and no one cares about the societal implications of that, then it's just going to sort of be in the metadata. When you sort of click into a detailed view of video, it'll say, hey, this was made using an AI tool. If you are doing something a little bit more sensitive, and again, we don't have a lot of details
about what this is going to be, but I don't know, maybe you make synthetic video about an election and it feels a little bit edgier, then there's going to be an overlay on top of the video. So as people watch it, they're going to see a little box that says, hey, this is altered or synthetic content.
And is this sort of working on the honor system? Like YouTube is trusting that creators will check this little box when they do upload stuff that was generated using AI?
Pretty much. You know, there are no reliable tools for detecting, in many cases, what videos were made using generative AI. And so YouTube is going to ask people to just be honest about that. Now, at the same time, they have AI systems of their own. And I'm sure that over time, those will get better at detecting what was made with AI.
So there's this little overlay if your video is deemed to be realistic or maybe deals with some sensitive topic. What else did they announce in this blog post?
Well, look, maybe you're listening to this and you have the assumption that if I woke up one day and said, I'm gonna make a deep fake of Kevin and the deep fake is gonna say, I'm a big dumb dumb and I'm bad at podcasting. And I made that.
That's just the affirmation that I say to the mirror every morning. How'd you get that video?
So let's say.
Did you hack into my Dropbox?
So let's say I make this video and I go to upload it to YouTube, that that would just sort of automatically be against the rules, right? Maybe YouTube might even have an automated system to say like, I don't know, did Casey have Kevin's permission to do this? That is not what they're going to do. They're going to let me upload that. And then if you, Kevin, do not like the video, you can go into YouTube and you can file a request under YouTube's privacy policy to say, hey, I don't like this, take it down. And then YouTube will consider a variety of factors, including whether you are a public figure. And then they will decide, or their automated systems will decide, whether they want to honor your request. But the answer might be no, and the I'm a dumb dumb video might stay up on YouTube.
And it now has four million views and counting.
We're going viral, baby. I just bought a house with the AdSense money.
So if a deep fake video is made of me and uploaded on YouTube, I do have some recourse. I can ask for it to be taken down and YouTube will consider that. Do you think that's YouTube trying to sort of get ahead of some problems that it sees coming? Or do you think this is already happening?
Well, I'm sure that in some cases it's already happening, although I don't think it's happening that widely. But I think YouTube sort of arrived at a fork in the road, where they had to decide, do we want to be really restrictive about what we enable people to post using generative AI? Or do we want to be more permissive and say, hey, go nuts? And we'll just sort of remove bad things on a case by case basis. And to be honest with you, I sort of assumed they would err on the more restrictive side of things, just given all of the backlash to social platforms in general over the past half decade or so, the sort of misinformation they've enabled, the hate speech, the cyberbullying, the harassment. But instead YouTube just said, you know, we're gonna trust you guys and you have to tell us if you made it using generative AI. But otherwise, our policies are mostly gonna stand as they are.
Hmm. And why do you think they did take that permissive approach?
I mean, I have not been able to. To get them to give me a straight answer to that question. I think they would say that they have always operated in a tradition that tries to enable the maximum amount of speech. I think there are some good reasons to do that. If you believe that we should be able to do parody and satire of our world leaders and upload those videos to YouTube, and you want to make a parody or satire using generative AI, then YouTube is enabling you to do that. At the same time, I would just note, if I were trying to design the cheapest way for YouTube to do content moderation in a world of generative AI, this is the system I would pick. Because the onus for reporting bad things, it is not on the platform. It's not on me who made the video calling you a dum-dum. It is now on you, the possible victim of my deep fake,
to go in and just hope that the little form that you fill out reaches somebody who agrees with you.
And I guess my big question about this is like, how is it gonna work when people with a lot of clout with YouTube, when advertisers, when big YouTube creators, when celebrities start getting deep faked into videos that they didn't appear in and that they're not happy with? Is that just gonna be sort of treated on a case by case basis? Or will they actually lobby YouTube to change its policies? What do you think?
Well, it's a great, open question. Do you remember the video of Nancy Pelosi in which she appeared to be slurring her words?
This was a big controversy from 2019. And this was not actually a deep fake. Someone had just taken a video of Nancy Pelosi speaking, and they slowed it down a little bit to make it seem as if she were slurring her words and maybe potentially drunk. And this was a huge controversy. Nancy Pelosi was super mad about this. They were lobbying Facebook to take it down. It was a big thing inside of Facebook. Ultimately, Facebook decided to leave that video up. But again, we're so early in these days that most famous people have not yet had the experience of seeing themselves deep faked. And I have to imagine that pretty soon, a lot of them are going to see these deep fakes, and some healthy percentage of them are gonna hate it.
Right, they're gonna be super mad. And they're not gonna be sort of placated by like a little overlay on the video, even if the creators are honest and forthright and check the box when they're uploading their deep fakes.
And of course, Kevin, there is another twist, which is that there is a different system. If you are a musical artist.
Really?
Right, yeah. So we talked about this a little bit on previous episodes when we were talking about AI music stuff. But what YouTube made clear with this most recent blog post is that there is going to be a different process for its music partners. We know that YouTube is heavily dependent on its relationships with the major labels for all sorts of things. Music videos and music in general is a huge percentage of the listening and the watching that YouTube gets, right? So YouTube does not want to lose these deals. It also really wants to create new creative tools so that you or I could go into YouTube or some new app and say, hey, I want to make a Radiohead song with Drake's voice and that would just be available to me. I think they'll probably try to set it up so that the artist gets some sort of revenue if what I create winds up making any money, right?
But so as part of these negotiations and as part of this new blog post, YouTube has gone out and said that if you're a record label and you represent artists participating in these early AI music experiences, you're going to get a different request form that you can fill out saying, hey, my name is Drake, I'm a famous recording artist, I've noticed a bunch of people are making songs using my voice, I don't like it, take them down. And I assume that those requests are going to get a lot more attention than Joe Schmo when he files his request saying, hey, I don't like the deep fake.
Right. So this policy, I guess it feels like maybe a 1.0 policy that's going to sort of get revised over time. Because right now, the technology for creating video deep fakes is just not quite there yet. Like you can create little short clips, but it's just not really, you can't type in like, make me a, you know, a full-length music video of Drake singing a Radiohead song. Like that is still beyond the sort of reach of these AI systems. But I think they're probably thinking like, it's not going to last. These tools will get better. And pretty soon, it will be possible to make something like a full-length music video. And so it feels to me like this is sort of them wanting to put something out, but knowing that it's going to change and evolve over time. Is that your read on it?
It is. And by the way, that's good. You know, I should also say these policies aren't even taking effect yet. They're going to take effect in the coming months, which means that YouTube is giving itself time to refine them based on the feedback that it gets. So, so that's good. Like I'm all in favor of that. It's good in general to see platforms trying to get out in front of these issues and not just catch up once the crisis has already metastasized.
Yeah, I also think it's probably them trying to get ahead of potential deepfakes having to do with the 2024 election in the US. We know that the platforms, they all sort of orient their trust and safety work around US elections because those are the times when the stakes are the highest for them and when politicians are really paying attention to what's going viral on some of these platforms. So I understand the impulse to want to do something. You had an interesting point in your newsletter, though, which is basically about how this is content moderation that is happening at sort of a different level than we are used to. Explain that and sort of walk us through your logic there.
Yeah, so in the first era of social media, content moderation rarely happened at the level of the tool. And what I mean by that is if you open Adobe Photoshop and you wanted to draw a figure of a naked human, Adobe is not going to pop up a little wizard that says, hey, looks like you're trying to draw nudity. Knock it off, right? But if you go to OpenAI's Dolly or many of the other generative AI tools that are broadly available and you say, show me a naked man, they're going to say, absolutely not. So we're in this era where the creative tools that we have to make generative AI are being quite restrictive and they're saying, you can't, even if this is just for your own personal use on your own laptop and you have no intention of sharing it with anybody, we're not going to let you do that.
And then you come to these platforms, which in a previous era were responsible for doing the moderation, right? If I made that naked image in Photoshop and I went to upload it to Facebook, Facebook would say, you got to take that down. Now we're in this world where if I make a deepfake of you, Kevin, and I go to upload that to the website, even if the creative tools that are broadly available would not let me make that image, if I'm somehow able to make it, YouTube, Facebook, they say, for the most part, go nuts. So it feels like we're in a little bit of a topsy-turvy world and I think we should have a conversation. Do we actually want these creative tools to refuse us in the act of creation? Or would we rather that the platforms take the responsibility to say, you know what, do whatever you want on your own laptop, but if you want to host it on our servers, then you have to follow some rules?
I think the platforms are very eager to pass the responsibility for content moderation to another set of companies. They feel like they have been doing this sort of- Sicifian task of rolling this content moderation boulder up this hill for many years, thanklessly and at great expense and embarrassment to themselves. I think they love the idea of that responsibility now falling to OpenAI or Midjourney or another one of these AI tools.
Yeah, I think that that is totally true. I think that ultimately, though, it is the generative AI companies that are, in one sense, being very responsible. But on the other hand, like, I do, I do want us to have a conversation about that. Right. I think like, in another era, this is like a banner that I could imagine the Electronic Frontier Foundation picking up and saying like, what you do on your own devices is your business and companies should broadly enable that. And what we really ought to be concerned about is how the media that you're creating gets shared and distributed. But we don't live in that world because the platforms are saying, absolutely, share and distribute this kind of content. Now, the platforms would say, Casey, we still have all of the same rules, right? You cannot post hate speech if you make it with generative AI. You cannot bully people using generative AI.
And I hear that, but I'm looking at these policies as nascent as they are, and it just seems like they're leaving a lot of wiggle room for people to cause mischief and mayhem. So I would just say, let's keep our eyes on this as these video making tools get better.
Totally. All right. When we come back, we talk with a man who is trying to bring lab grown meat to the masses. And we get a taste.
A taste of the future.
Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically integrated AI cloud platform, delivering data centers, compute and software for AI training and inference, underpinned by its expansive portfolio of land and grid connected power in renewable rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iron, AI Cloud for the next big thing.
Companies are pouring money into AI, but few leaders can explain how those AI investments work with the systems you already have. You know, the ones that actually run your business? Without Systems, you don't have to start over with AI. You can choose the projects where AI will have the most impact. And OutSystems gives you the platform to make it agentic on your terms with complete control every step of the way. So you can trust AI with the mission-critical systems that run your business. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.
AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.
So Casey, we're coming up on Thanksgiving and the holiday season when a lot of people gather with their families and eat a bunch of meat.
Mm-hmm.
And this has been coming up for me recently because I have family members who are non-meat eaters. They're vegetarians and vegans. And so every year, I'm confronted with the same horrible question that gnaws on my conscience all year, which is, why do I still eat meat?
Yeah, it's one of those things where if you sort of just do any level of research, it becomes clear to you that if nothing else, you should probably be eating less meat as an American than you're already doing.
Yeah, I like to call myself an intellectually convinced vegetarian because I know on a rational level that eating meat is indefensible. I know that our kids and grandkids are going to be horrified that we ate meat.
And yet, what we have to say to them is, but it was delicious. And that's the thing that so often gets left out of these conversations.
It's true. I mean, I don't have to tell you, there are all kinds of reasons that meat is not something that we should be eating as much of as we do. It has environmental sustainability risks. It's cruel to animals when you think about factory farming, and it has public health risks, too. And so in the lead up to Thanksgiving, these arguments about meat are always rattling around in my head. And so this week, I wanted to actually do something about it and have a conversation that squares directly on this issue of meat.
You've looked into the question of can we sort of have our beef and eat it too, as it were?
Exactly. So I've been fascinated for years with this issue of cultivated meat, as they're now calling it. It used to be called lab grown meat.
Although I liked it better because then at least I knew where it came from. It came from the lab.
So cultivated meat, if you're not familiar with it, is just a term that applies to meat that is grown from the cells of animals, but it is not grown on an animal. There is no living being that has to be slaughtered to make this meat. Instead, it is grown in a lab, in these bioreactor tanks. And this has been something that tech companies and food science companies have been working toward for years. And I would say that it has been an area where the progress so far has been disappointing.
Yeah, for decades now, scientists have considered this a holy grail of the food world. If they could cheaply create meat that didn't require slaughtering animals, that would have a lot of obvious benefits. But we have unfortunately just seen failure after failure in this space.
Totally. There's been so much hype around this industry. In the early 2010s, there were enough sort of alternative meat projects getting underway that people gave it a name. They called it schmeat. Do you remember this?
As in S C H M E T, schmeat. I love the word schmeat.
So schmeat did not stick, but the efforts to make cultivated meat did. The first lab grown burger was unveiled at a taste test in London in 2013. According to my colleagues at the New York Times at the time, the meat itself was a bit dry. And a bigger barrier was that it cost about $325,000 for a single burger.
It's so hard to fit that into the average family's budget.
So obviously, the promise and the hope of these companies was that over time, the cost of making cultivated meat would come down radically. Eventually, it would be possible to mass produce it, and it would be sold to customers at a price that approximates, or even maybe is cheaper than the meat that we would all buy in the grocery store today.
A beautiful dream for schmeat.
And a dream that so far has not panned out. Today, there are estimated to be about 100 companies working on cultivated meat around the globe. And in June, just this year, the US became the second country in the world to approve the sale of cultivated meat for two different companies. But the dream of mass-produced cultivated meat that is sort of plentiful and cheap enough for people to use it as a substitute for slaughtered meat is still a ways off. So today, we're going to talk about why it has been so hard for cultivated meat to hit the mainstream. And we're going to be talking about it with Josh March. Josh is the CEO and co-founder of SciFi Foods, which is a startup that is making cultivated meat burgers by combining beef that it grows in a lab
with plant-based ingredients. He is very optimistic that this future of cultivated meat is coming. And he actually was very helpful in helping us understand why it has taken so long for this damn cultivated meat to arrive on our plates.
Yeah, well, I can't wait to schmeat him.
OK, that's enough for the word schmeat. I'm putting a moratorium on that. Let's bring in Josh.
Josh March, welcome to Hard Fork.
Thank you.
Hi, Josh. So one thing that's different about your approach to lab-grown meat, or I guess cultivated meat is like the industry term, because it sounds less weird than lab-grown or something like that. So one thing that differentiates your approach from a lot of other startups in this area is that you are combining lab-grown or cultivated meat with more plant-based ingredients. Your burgers are about 90% plant-based and only about 10% of the cultivated beef. So why did you choose that approach?
Yeah, I mean, look, ultimately, we're working towards a future where we can create any kind of meat products 100% cultivated from cells.
That future of doing that in a really large scale and affordable way is still a bit of a distance away. The reality is like the technology does not exist to do that at scale at low cost today.
Right.
You know, I'm very skeptical of anyone who claims otherwise.
I will put some cards on the table here and say that like cultivated meat has been a category that has been, I would say like disappointing for me. I do this column every year called the Good Tech Awards where I like single out a bunch of startups that I think are doing good things for the world. One year I featured a bunch of cultivated meat startups and I talked to them and they were all telling me, you know, our products are about to be on store shelves, you're about to be able to eat this in restaurants, get ready, cultivated meat is coming. And then it just didn't, right? It's too expensive. There turn out to be production problems. There are all these companies that have gotten, you know, raked over the coals for, you know, exaggerating their production details or claiming that they could make stuff for cheaper than they were actually making it.
So what is the sort of advance that you guys have come on to that you think is going to allow you to actually bring this stuff to market and not just be like another cycle of empty promises?
Well, first of all, you're completely right. And, you know, that's really the reason I started the company. I mean, after spending a few years on the sidelines of the industry, I felt there was a lot of very bold claims and a lot of very like hand waviness that costs were magically going to come down. And the more I learned about the technology and the science involved, the more I realized that that's not necessarily true. There's some real biological, physical limitations and you need a pretty clear plan for how you're going to get an affordable product. And it was clear to us from the get go that first of all, the only products that would be viable today would be blended products where the majority is still plant based and you're really just using the cells essentially as a flavoring ingredient. That solves a lot of the technical as well as the cost challenges.
You know, when we created the very first Sci-Fi Burger. Three, four years ago now, it costs about $20,000 to create.
That's more than you'll pay at any restaurant in America, except for Salt Bay's restaurant. That actually is cheaper than the Salt Bay burger.
And why is it expensive? Fundamentally, like you take a cell from an animal, that cell has all kinds of behaviors and characteristics that are really optimized for growing, you know, in a cow, not optimized for growing in a big steel tank. And so you have to change that cell's behavior and adapt them to grow in a cheaper and more scalable way. And it was obvious to us, based on really the background of our team, that the fastest and most reliable way to do that would be to use the power of genetic engineering, really enabled by tools like CRISPR, which have kind of really changed the game when it comes to being able to make tiny edits to animal cells. And that that was really the route to shift that behavior and get to a scalable and affordable product.
I mean, it's just dawning on me like how incredibly complicated this is. You know, like I mostly write about software. You have to solve an insane number of technical problems just to grow a little bit of this beef.
Yeah, and you know, what was really important to us and one of our kind of guiding principles as a company was that we only wanted one miracle, right? So I think a lot of startups fail, especially in bio manufacturing, when you're relying on like multiple miracles. You need to like engineer new cell lines, you need to create new bioreactors, you need to like come up with cheaper downstream processing, right? And it's like, it's too many things. And we felt all of that would be way too risky. And so we knew that if we could just use CRISPR to engineer our cells to have certain kind of core KPIs in terms of how they grow, we could make a 10% product at an affordable price that would be profitable to us at scale. But we really tried to reduce the amount of risks and miracles down as much as possible.
What was the one miracle that you needed? I'm not sure I caught that. Like, what was the miracle that you're betting on?
The one miracle was basically, could we use CRISPR to engineer our cell lines to shift their behavior so they can grow really effectively?
It seems like CRISPR and meat should have a sort of, you know, bond. It doesn't like a nice...
That's basically what we're doing, yeah.
So Josh, I'm gonna ask a question that I already know the answer to because I am obviously a very studied food scientist and know all of the things that you're talking about. But for any of our listeners who might have questions, could you just walk me through like how you actually make lab-grown meat?
Yeah, yeah, absolutely. So, you know, we first, a number of years ago, took a tiny bit of muscle biopsy from a living cow. A vet did it, not us. And it was just a tiny bit of muscle. The cow like went back running around its field afterwards. We then take that to our lab, we isolate the individual cells. We then have done years of R and D to adapt and engineer those cells to get them really comfortable growing in.
This is the CRISPR part.
That's the CRISPR part.
So CRISPR, from my basic understanding as an expert food scientist, is a way to kind of snip material out of certain genes and insert it into others. It's like a gene editing tool.
It's gene editing, but what's great about CRISPR is that it allows you to make very targeted tiny changes. Like, literally, most of the things we do is we delete a single base pair of DNA, and that just stops like a protein being expressed. We're not putting any foreign DNA. We're not taking like DNA from salmon and putting it into beef. All we're really doing is just shifting the expression of proteins in beef.
So what are you doing with CRISPR on these beef cells taken from the cow?
So as an example, when you take a cell from a cow muscle, right, that cell only wants to grow when it's attached to other cells or a surface. So you can grow it in the lab in a petri dish, right? But it's like a single layer of cells. It's a tiny amount, very, very expensive. But what you really want to do for manufacturing is you want it to be growing just floating in liquid in a big steel tank. You want it to look kind of like you're fermenting wine or bacteria yeast, that kind of thing.
Just a giant floating mass of beef and water. Sounds very appetizing.
Yeah.
Almost like a soup.
Tasty, really tasty beef soup.
It's a sous vide.
Yeah, exactly. But if you just take like a normal cow muscle cell and put it into that condition, it doesn't like those conditions, it will die. Right? So we figured out how we could use CRISPR to make a few different tiny changes that in combination would make the cell comfortable growing in liquid, in suspension, it's called.
Interesting. So then you have these cells that can grow in liquid.
You let them grow.
Yeah. So basically, what is a bioreactor, right? A bioreactor is basically just a big steel tank that is extremely clean, perfectly sanitized, not a single bacteria in there until you put the cells in. And then it allows you just to carefully control the conditions. It allows you to add different nutrients and feeds. You're feeding the cells sugars and amino acids and vitamins, all the stuff that, you know, life needs to exist. And it's just the right conditions so that the cells keep growing. And so essentially, once we've done those years of R and D, we then have a cell line that grows really well. And each time we want to manufacture it, you know, we start from a tiny vial of cells, and we start expanding it in larger and larger vessels, until eventually it's in a big production bioreactor, and it keeps, you know, doubling every day or so. And then eventually, after
like a week in the big production bioreactor, we can harvest it, centrifuge it down, and we have tasty beef cells.
Got it. And how much meat can you currently produce this way in like a week?
Out of our biggest bioreactor, we could produce like 2,500 grams or so a week.
So 2 kilograms?
Yeah, a couple of kilograms a week.
Great math, Kevin.
We love math challenges on the show. So that is not a lot of meat.
No, but we have a very small pilot plant here in the Bay Area that we just finished construction of. You know, this is not a large scale commercial facility yet, but we will be commercializing out of the pilot plant, which we're excited about.
And when you do commercialize, when you're at scale, what do you think a pound of lab grown meat will cost you to produce? Like, what is it going to take? My understanding is even though the prices have come down, it's still very expensive to make cultivated meat in these bioreactors. I think a lot of people see price as the main barrier here. Regular beef, if you go to the supermarket, you're paying, you know, five, seven, maybe 10 bucks a pound for that. What does it cost you to produce a pound of meat this way currently? And then what do you need to get the price down where it's maybe equal to or almost equal to the price for regular beef?
We like to think of everything as like cost per burger. You know, I mean, quarter pound of burger, obviously. Out of our pilot plant, we just finished construction of, we'll be launching commercially out of that end of next year. And we think we'll be at about $30 per quarter pound of burger. And that's the 90% plant, 10% cells out of the pilot plant. From a larger scale commercial facility, our goal is to get it down to a dollar a burger. And we know exactly what we need to do. It's mainly just around the performance of the cells. And we know that that kind of performance is possible from what people have done with other cells and other species. But we think it's very biologically possible. And it's just a couple of years more of development to get us down to about the dollar a burger cost. Now, that's our cost. So we'd be selling about $2 burgers. That's your wholesale price of about $8 a pound.
And then what if any sort of regulatory hurdles are there? Because my understanding is that the sort of regulators are being very slow to kind of give approval to cultivated meat to be sold in grocery stores and at restaurants. So what needs to happen for this to be sort of legal and sort of widespread?
You know, I think we're actually in a really great place in the US in terms of how forward thinking and supportive the regulators are. The FDA and the USDA have this joint regulatory framework where the FDA regulate the cell lines and the upstream process of growing the cells. Once you harvest it, the USDA regulate the harvest process and like the facility, they basically treat it like you've slaughtered an animal at that point when you harvest the cells. And you have to have a pretty developed process, right? You have to demonstrate that you can consistently grow the same cell line, you know, to a reasonable scale. And you have to be able to do a lot of like food safety tests and show that there's no contamination. You have to submit safety dossiers and the regulators want to look at the final product
that you're creating and check, you know, you're saying it's beef. Is it actually, is it fats and all that kind of stuff similar to beef? And so you have to submit quite a lot of information. Once you've actually submitted it all, though, generally the FDA will give you an approval letter in nine to 12 months. They've already done that for two companies. And then the USDA will come approve the facility. Honestly, I don't think the delay has been on the regulatory side. The delay has mainly been companies getting to the stage where they're ready to submit.
I want to ask about the culture piece of this too, because I I'm very optimistic that you all will eventually come up with something that, you know, Casey or I or other sort of like coastal, you know, are
Snobs.
Yeah, coastal snobs will eat and feel virtuous about. But I want to read you a tweet that I saw. I guess we're calling them posts now from Ronnie Jackson, who is President Trump's former doctor, who is now in Congress in Texas.
And one of your favorite X accounts to follow.
Yeah. He said, and I quote, "I will never eat one of those fake burgers made in a lab. Eat too many and you'll turn into a socialist Democrat. Real beef for me."
You know what? That's funny is eating real beef turned me into a socialist Democrat. So I think it just has different effects on different people.
So, but I think there's a serious question, which is like, it is clear to me and to a lot of other people that the barrier to a culture where we all eat cultivated meat is not just the technology or the cost. There's also something sort of intrinsically linked to kind of a feeling that people have where things that are grown in labs are weird and beef that comes from cows is natural. And so why would you swap out the thing that is natural for the thing that is weird? So how do you market something like this? Or how do you think about positioning this so that, you know, someone like, you know, Ronnie Jackson, maybe not him specifically, but so that, you know,
No, let's focus on Ronnie Jackson. If he can get Ronnie Jackson, we're winning.
Yeah, what's your Ronnie Jackson strategy?
Yeah, well, first of all, I think this is Really important topic, more than almost any other food and especially red meat has a lot of like emotional symbolic kind of connotations. And so I do think you have to be quite careful. And I think this is one of the challenges that plant based meat has had by just being like meat can be plants. And people are like, no, it can't, you know. And so I think, again, that's part of the reason for like taking a different tack from a branding. And like the fact that we can be more pro meat, I think is really important. And I think there is this message that like, hey, yeah. If you can go and hunt all your meat, great. But like, let's be honest, most of us do are like hunter gathering in the grocery store, right? And if you're doing a hunter gathering in the grocery store, we'd rather like be able to create meat in this like new, awesome, like fun, exciting way without factory farming and all like the other crap that comes with it.
And I think it has to be fun. You know, we think about this a lot, like we, you know, even though a lot of my motivation may be like, you know, climate and animal welfare. Like, that's not most people's motivation when it comes to eating a burger. And in fact, even for most people who intellectually agree with those things. When it comes to like ordering a burger and like on a Friday night, like they're not thinking about those things. It's a very like lizard brain, you know, emotional decision to get the tasty burger. And so we think it's really important to just be fun. This is part of the reason for like our branding, you know, called it sci fi foods. Because we think that people are going to think this is sci fi and weird. We don't think we can avoid that.
So lean into it.
Many lovely journalists still call this lab grown meat, as we've heard on this podcast today already.
Shame on them.
So we don't think we can avoid it, right? And we know the meat lobby is going to come after it hard for being Franken meat. So we have to take something that people are going to think is sci fi and make it safe and fun. And we think we can do that by kind of leaning in, a bit of a wink, and just getting people excited that this is a cool new way to make real meat.
Yeah, I mean, I remember years ago, I went to what at the time seemed like it was one of the hottest new restaurants in New York, which was called superiority burger. Do you remember superiority burger?
So it was a place where you could get a plant based burgers and everybody was like, you have to have this burger. It is incredibly good. And so I was visiting New York and I went and I had the plant based burger and I really enjoyed the experience, but I think 90% of it was just the taste of cheese and ketchup. You know, but I wonder, it's like, if that presents an opportunity to you, like for the most part, people are not eating just plain beef patties, right? They're smothering it in things that taste very good. And so maybe it doesn't matter as much, you know, if it isn't an identical experience to eating a beef burger.
Yes. But one of the challenges with a lot of the kind of plant-based burgers is that you just have this kind of off-taste of like plants and like serially kind of people in the industry, they call it serially off notes and planty off notes.
We hate serially off notes.
Yeah, people do. It's kind of amazing when they want a burger, you know? And that's actually one of the biggest things that our product doesn't have.
Well, then maybe it's time to taste the darn product. What do you say, Kevin?
Let's taste the darn product. You have brought some of your burgers to give us to try.
Well, actually, we wanted to do Thanksgiving themed.
That's right.
So instead of doing burgers and smothering and ketchup, we have meatballs with a cranberry glaze.
Oh, sounds great to me.
Sounds awesome.
You can't talk about meat for this long and not be a little hungry.
Yeah. All right, let's, let's get some of this meat in here.
All right.
Amazing. I'll sit.
Yes, thank you, chef.
Thanks, Jeff.
Thank you, chef.
Well, this is exciting. Never done this format before, so this is a special for you guys.
Nice. Great. Well, thank you for helping us get into the Thanksgiving spirit.
So we are here and in front of me are three meatballs that are made out of your cultivated beef plant mixture. And there's like a lovely looking cranberry glaze on here. And I'm very excited to dig in. But Casey, before we dig in, we actually have to do something that I have never had to do before I eat a meal before, which is to sign a waiver. So Josh, what is this waiver that you are having us sign?
So this basically just says this is a novel food. It's not yet approved for commercial sale.
And we've done our own internal safety assessments. We believe this food to be completely safe. But because this is not approved for commercial sale, essentially you have to do this at your own risk. And so that's basically what the waiver says.
No, I don't have time to read this whole thing. Can you just tell me, am I going to die from this?
Definitely not.
Okay. Because I'm prepared to die for, for the podcast, but I would just like to know.
I've been eating this now for multiple years regularly, and we've done over 100 tastings and have zero adverse effects.
Okay, wow. Okay, so I have to, okay, we're going to sign the form here.
Now, we let Jesus take the wheel.
Yeah. Let Josh take the wheel.
All right. So we have our forks here.
Yeah, and Kevin, you actually did mention when you were describing this, each of these meatballs, which are quite appetizing looking, are sitting in a cloud of what I believe is mashed potatoes.
Oh, that looks great.
Very good.
And there's a little like, a little sage leaf on top.
Should we try it?
Let's go, let's try it.
There goes nothing.
Mmm.
Hmm.
It's good.
Mm-hmm. It is good.
And there is a beefiness to it.
It didn't hit me right away, but I chewed it a little bit and then I got the beef.
The texture is nice and beef-like.
Which I appreciate.
Which you love it when people say that about you.
So this, this might be, well, I was gonna say this is the most science that has ever gone into anything that I've eaten, but then I thought, I eat Cheetos. And we all know.
Nothing in America has been engineered more than a Cheeto.
Yeah, nothing in, nothing in a Cheeto came out of the ground. They don't grow those Cheetos in fields, let me tell you.
Cheetos were not part of God's plan for this world.
Now, here's my question. So, again, I think there's, there's good beefy flavor in here. I think that if I did a blind taste test and you handed me a pure beef meatball and this meatball, I think I would enjoy both, but I would be able to tell the difference.
What do you think would be the difference in taste if this were 100% cultivated beef versus what I'm eating right now?
Yeah, then you wouldn't be able to taste the difference at all.
Yeah, and we do.
Because practically, there is no real difference, right?
No, it's the same, it's the same cells, the same fats, the same things that create that flavor. So what we find is that 10% makes a big difference. You would taste the difference clearly between plant-based as well.
It is like, like I said, if you do a blind taste without beef, you'd think it's beef.
If you compare it directly to like 100% conventional beef, you're like, oh, I can tell the difference a little bit. The more we add. The smaller that difference becomes. And there's other things we can do to improve the flavor.
So if we were 20% cultivated meat or 30%, it would taste more like beef, but it would also be significantly
But it's also much more expensive. Yeah. And so for us, it's just like, hey, this is a journey that we're on. And we think we can be completely transparent with consumers, like this is the 10% products. And at some point when technology evolves, we'll be able to go to the 20 and the 30 and the 40, and eventually 100.
Right. And do you see a path to work on meats other than beef? Like, is there, will there be chicken and lamb and?
Yeah, 100%. You know, beef for us is really the holy grail. When you think about people wanting to eat different meat products, like it's mainly red meat. And from a climate change perspective, certainly beef is just like the biggest contributor by a long way. And ground beef also has the highest price per pound of like any ground meat product. And it's one of the biggest markets from a revenue perspective of any meat product. So beef's the really best place to start. But our same platform and approach can apply to any, any species. And we intend to keep going from here.
Well, Casey, do you feel like a socialist Democrat yet?
I don't know. I just ate a plate of meatballs worth $90,000.
No, no, no.
That makes me a capitalist.
Many years ago.
What would these cost today if you were selling them in a store?
Uh, probably like a hundred bucks.
Okay.
It's a nice lunch.
That's actually about the cheapest lunch you can get in the financial district of San Francisco.
That's true. That's true. I like these. They don't taste exactly like meat to me, but they do taste better than a lot of the alt meats that I've tried. Casey, can you imagine eating this like in a year or two as part of your normal diet?
I mean, yes, like I can. Um, I think if, I'm thinking about, you know, I have, I have friends who are vegetarian. Maybe we go to a vegetarian restaurant. I see this is on the menu. Like, I can imagine ordering it. Yeah.
You're in the clean plate club over there. I gotta, I gotta step it up.
Something I realized as I was eating was that I was actually quite hungry. The podcasting takes a lot out of people. People don't know that. People think, oh, it must be so easy. Just sit there, run your mouth. No, you're burning a lot of calories.
Well, Josh, it's a very cool demo. I, I enjoyed my meatballs. And I think Casey did too, judging by the fact that we all finished all three of them.
Yeah, we did. We were talking about two demos on the show. This was the only one I could eat, and so I did actually prefer it.
Josh, what are you serving at Thanksgiving this year?
You know, I have a big group Thanksgiving with a load of friends and I'll be allocated something. Normally I do the green bean casserole.
That's good.
Well, this would work in a pinch if the turkey burns in the oven or something.
Yeah, exactly.
Thank you so much for coming. Really good to talk to you.
Thanks, Josh. Thank you.
Today's AI builders need infrastructure that scales with their ambitions. Iron is a vertically integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. Iron's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. Iron, AI cloud for the next big thing.
Building and scaling AI today can feel pretty chaotic, mostly because teams are juggling too many fragmented tools at once. That's where Outsystems comes in. Outsystems gives you one unified agentic systems platform to manage your entire portfolio of agents, apps, and workflows. You get to innovate at the speed of AI, but with the security, governance, and control your enterprise actually needs. And the results are already there. A major manufacturer used Outsystems to build a multi-agent system that analyzes blueprints, and a global bank cut their development time by 60%. Own your agentic future today at outsystems.com/nyt.
Casey has some feedback for you about our YouTube channel.
Oh, great. Glad I'm sitting down.
So I was watching one of our videos, which are looking great, by the way. Love our channel. People should go subscribe.
Absolutely.
You know, it's a good time.
Great time.
This swiveling in your chair has to stop. I'm gonna disable the swivel in your chair.
Listen, they've already disabled everything else. This chair used to have wheels on it.
It's true.
And so I used to get to kind of move around. And now I'm locked into place. I don't get to have my little stand for my laptop anymore. All I have is a little swivel.
I was watching our video and I was watching you. And I was watching you like just rock gently back and forth. And I was, I was feeling like I was on a ship.
Just the subtle swaying of your body back and forth.
What if I told you that it's like a fundamental element of my creative process? Like, you know how Cyclops has to wear the visor to prevent himself from shooting people with lasers all the time? That is what I'm doing while I'm swiveling.
What are you preventing?
I am preventing myself from interrupting you even more than I already do. So that is what I'm doing.
Well, if that is the reason, then I guess I'll accept it. And our YouTube audience will just have to deal.
Thanks, YouTube.
Sorry for the swivels.
It's OK.
Hard Fork is produced by Rachel Cohn, Davis Land, and Emily Lang. We're edited by Jen Poyant. This episode was fact-checked by Caitlin Love. Today's show was engineered by Daniel Ramirez. Original music by Diane Wong, Rowan Nemezdo, and Dan Powell. Our audience editor is Nell Galogly. Video production by Ryan Manning and Dylan Bergerson. Special thanks to Paula Schuman, Pooing Tam, Kate LaPresti, and Jeffrey Miranda. You can email us, as always, at [email protected].
Anybody got any lab-grown meat recipes?
We're listening.
Building and scaling AI today can feel pretty chaotic, mostly because teams are juggling too many fragmented tools at once. That's where Outsystem... Comes in. Outsystems gives you one unified Agentic Systems platform to manage your entire portfolio of agents, apps, and workflows. You get to innovate at the speed of AI, but with the security, governance, and control your enterprise actually needs. And the results are already there. A major manufacturer used Outsystems to build a multi-agent system that analyzes blueprints, and a global bank cut their development time by 60%. Own your Agentic future today at outsystems.com/nyt.
Page updated October 12, 2026 · transcript and summary are machine-generated
