Pushkin.
I try to bring my lunch to the office, but when I can't get my act together, which is most of the time, I go out and get lunch at a salad bowl place, and I pay like fourteen dollars. This always makes me feel like a little bit of a chump, because fourteen dollars seems like a lot to pay for a salad. But the salad is healthy and
fast and tasty, so I buy it. There's of course, an issue here that goes beyond my own conflicted relationship to lunch, and that is this, a restaurant meal made with healthy, fresh ingredients costs more than a restaurant meal made with unhealthy processed ingredients. A salad costs more than a burger and fries. And it would be great not just for me, but for everybody if somebody could use
technology to make my overpriced salad a lot cheaper. So I was pretty excited the other day when someone showed me video of a machine that might help make that happen. This video you sent me is password protected, which is curious to me, Like, why is it a secret this video? Why don't you want to show the world that your thing works?
Because right now that version of the makeline is being tested at the Chipotle Cultivate Test Center.
Oh and so you.
Know, we haven't officially announced anything about this.
I mean, we have.
Announced that we're working together, right, but we haven't showed anything yet. O.
Great, we'll I should be doing that very shortly. So this is like, which is.
Why scoop it. You can't see the video yet, but we're gonna tell you about this video that you cannot yet see. I'm Jacob Goldstein. This is What's Your Problem? The show where I talk to people who are trying to make technological progress. My guest today is Stephen Klein. He's the co founder and CEO of Hyphen, a company that's developing an automated make line, basically an automated way to make the bold based meal that you get at fast casual restaurants like Chipotle or Sweet Green or whatever.
Chipotle is an investor in Hyphen, and one of the major projects the company is working on is an automated system that could make online orders at thousands of Chipotle's. Steven's problem is this, how do you make restaurant food made from fresh ingredients cheaper?
When my co founder Daniel and I started the company, we actually wanted to really be a vertically integrated restaurant, meaning that we wanted to effectively make the cost of healthy and delicious food the price of fast food. And we wanted to be as ubiquitous as fast food, right, And we thought that we could do that by automating the production of the food and then having it beyond wheels.
So we built a fully robotic food truck at about fifteen months with five people and two million dollars, and then COVID hit about two months later, and so where we were operating at this in the back of this robotic food truck, you know, they were effectively ghost hounds, you know, overnight, and so I still remember viscerally kind of walking the streets in aerie silence, deciding like what
are we gonna do. So we effectively retooled the system to put it in a make line, and that's what we're seeing today in that video.
So it actually started as a food truck.
Here's what you see in that password protected video. Stephen sent me this thing called the make line, which looks like what you see when you walk into a Sweet Green or a Chipotle or whatever along counter with containers of ingredients sunk into the counter. But this make line
is different than all other make lines. Each container here has an automated trapdoor or augur or something in the bottom to dispense the food, and underneath the counter, like where the cabinets would normally be, there is this automated system where a bowl drops down into this like clamp at the beginning of the line, and some lettuce or whatever automatically drops into the bowl from the bin above.
Then the clamp like pivots and hands the bowl off to another clamp a little further down the line, and the next ingredient is dropped into the bowl, and all on, all down the line until the salad is made, the bowl gets to the end of the line, and then finally this little elevator raises the bull up to the top of the counter.
Actually, the guy that designed the elevator was designing an elevator for one of the reads at disney Land, very similar design about you know, ten thousand times the size. But it was pretty funny that that was his first project.
When we kicked off.
When you watch the video, the whole process actually looks kind of simple, which, of course it is not. Steven says, one of the hardest parts, which you can't see at all in the video, was getting the little trapdoor system to allow just the right amount of every ingredient to drop down into the bowl every time. And the details of why that's a hard problem illuminate why fresh food tends to be more expensive than processed food.
It's extremely difficult to dispense and properly meter and plate ingredients because of the almost infinite permutations. Right, Let's take kale for example. If you have kale, it will change its permutations or characteristics depending upon how it's cut. Whether that's sautaied, blanched of and roasted, it will change permutation.
Potentially if it's.
Sourced in Salinas you know, where we're at, you know, or nearby where we're at, or somewhere else, it will change depending upon the cut method, right, if you cut it a different way. So being able to be agnostic to the type of ingredient is extremely difficult.
So we basically have to you know, map.
All these materials and characteristics to basically better understand how to feed that ingredient, and then it will get better over time.
That's interesting, I mean, it makes me think about the difference between processed food and fresh food, right, And when you buy processed with anything that's like in the middle of the grocery store, you know, comes in a box, it's like become very homogenized, right, presumably so that it can be produced at scale by machines, whereas what you get at a Chipotle or sweet grain is very heterogeneous. Right, as you say, not all kale stocks are the same.
And not to mention, there's like fifty different ingredients there could be. So do you have to build or more obviously there's more than fifty different ingredients? Do you have to build a different dispenser for every ingredient? Like if somebody's like, we want to do cantilina beans, right, do you have to be like, oh shit, all we know how to do is chick beans. We got to go build a whole new machine for cantelini beans.
Yeah, So we have about six different feeder types or dispenser types. We'd have one that you know, has to deal with like guauca mole in the case of Chipotle, or pico de gayo, which is very different than guacamole in terms of this viscosity as well.
As it's kind of you know, the chunkier.
Ingredients, and then you have leafy greens, right, and that that's going to you know, perform wildly arugula, et cetera. So we yeah, we effectively have about six six dispensers handle you know, thousands of ingredients at this point that we've we've we've mapped and started to kind of understand their characteristics. Greens, especially a arugula early on was was
pretty difficult to dispense. It would just start wrapping around these these kind of paddle wheels or augurs, and you know, I think we designed about seven hundred and fifty prototypes to.
Get to the design that we have now.
We three D print when we test, so we can do it in about a day and then test the next morning.
Was there a moment when you sort of solve the leafy greens problem?
Yeah, I mean it was just through rapid iteration. There was no one thing. It was just you know, these micro improvements we made along the way. And then the case of like black beans, where you get like that pulled up juice, you know, sometimes it's pretty unpleasant.
You know, we effectively had to build.
An auger that would basically push the ingredients out while you know, effectively letting some of that black bean juice kind slide into almost like a gutter and a kind of a septic tank in the back there.
And so you do get some liquid, but not not that pool.
I like that, although not to give you on solicit advice, I'd suggest avoiding the septic tank metaphor with the make line, especially with the beans.
Great.
Yeah, and I'd say, like the trickiest thing that we haven't solved yet is fand avocado. So we can do guacamoley, we can do like you know, the goofs and mounds, but getting that beautiful fan avocado slices like quite difficult to do.
Yes, Slicing fish filets things.
Like that can be quite difficult when they're you know, very sensitive.
So do you give up, like it's fandavocado off the table?
Yeah, there are some ingredients like fand avocado that you know, the operator on the line, because they're just kind of waiting for balls on the line, they can just add It's called expo, but they basically would just add the the avocado or those last final finishing items at the end there.
That we can do today.
So you've been working on this for what a little more than three years now. I know you've got an investment from Chipotle last year and you're sort of building a version of your line for Chipotle. Now, where is your line your product in the world now? Like, is it in restaurants now?
Yeah, it's it's in a few and we have a lot actually coming in Q four of this year.
We're just kind of starting to ramp up production.
How many restaurants is your machine in now?
About a dozen? Okay?
And what's happening with Chipotle? Like, so Chipotle made an investment in your company last year, right, So like that seems significant? Like is that I don't know, that seems like a big deal. And what's happening with that?
Yeah, and they're actually investing in the second time too. We're just wrapping that up right now, so we'll have some stuff to announce also.
And Q four of this year with Chipotle, is.
It likely that your machine will be in Chipotle next year?
Yes?
The CEO of Chipotle, you know, I think Believe has publicly stated, you know, he wants to all this out fleet wide, So you know, I think they have like thirty five hundred stores today.
I mean you're saying there are thousands of Chipotlas in the world and they want your machine and all of them.
Yes, they to have all of their digital orders. Yes, front of house you know, will always be front of house.
Oh that's interesting. So the idea is your machine would be in the in the back of the restaurant, and when people order on their app, the robot your machine would make it, and when people walk in, a person would make it.
Is that is that the idea precisely?
And having that you know, that staff that's friendly and providing that hospitality is super important if you're ordering in person, right, But if you're ordering from your phone, you know, you just want your food to be fresh, fast and you know, accurate, ideally, and that's where we help. I mean, right now we're really focused on just getting these lines to produce and plate you know, perfect food every time.
How far are you from them? Like, I guess it's not binary when you say like every time. I mean it's never going to be every every time, right, But there's some percentage liability presumably where it like, if you get above X, then it makes operational and economic sense. Right, What is X in terms of the uptim or reliability or whatever the metric is.
Yeah, I mean, like you want to be at ninety nine percent up time, ideally ninety nine point nine percent up time. In the event that you don't, you know, it's just like when your Internet goes down or you know, your slack goes down, right.
And what are you at now?
Ninety five?
Okay?
Going from ninety five to ninety nine is really hard, and going from ninety nine to ninety nine point nine is even harder.
So in a sense, the first ninety five percent, like you've done the easy part, in the last four percent is hard, and then the last zero point nine percent is killer is in order it might be harder? Yeah, And that's where that like one weird kale leaf just screws you, right, Ah, that fand avocado, right, fand avocado. We gave up on fand avocado and you're still at ninety five to get So do you just have to keep iterating to get from ninety five to ninety nine?
So that's a really great question.
So right now we are going through we're basically doing in DFMA, so design for manufacturing assembly, and as we do that, we start to injectionable parts, we start to kind of get you know, production ready materials that will drastically improve.
Going from prototype to production basically.
Yeah, we're also doing a lot of accelerated life testing. So we actually, you know, we did this system.
We started in November on this new system, and we you know, we launched in April, so pretty quick, right, But you know, doing accelerated life testing, we have to run thousands of times. So I think we have like we've clocked like ten years on our subsystems. But through that type of testing and knowing where the failure points are and the hardware is how we get to that ninety nine percent. Yeah, it's just those types of iterations, right, and just micro improvements on the actual design itself.
So okay, So if we imagine this future where there is a much more highly automated, fast casual sort of restaurant universe, I mean one question is it cheaper? Right? Like, one of my favorite things about technology is it makes
things cheaper. I try and bring my lunch to work, but I do end up buying the fifteen dollars salad once or twice a week and feeling like a chump for spending that much on a salad, though it is a good salad, Like, will it get my fifteen dollars salad to a nine dollars salad.
Maybe even cheaper? Right?
Because again, you know, the two primary drivers of that price is the food, the quality of the foods, and then labor, right is this second. And so if you can make the marginal cost of producing food near zero, you know, you can drastically reduce that price, right, and you could get that you know, nine dollars five dollars salad.
Five.
I like that, You're going for five five dollars salad. Really, that's big for me.
You like that?
Now it's time for an ad that It'll be back in a minute. Okay, the ad is over. Back to Stephen. Stephen has this long term dream for Hyphen. He wants the company to open up its own make lines in cities around the country, and then he wants to let anyone who wants to to create their own basically virtual restaurant. So whatever I could go online, I could design Goldstein bowls, sell them for nine dollars a bowl, list them on Uber eats and seamless, so I would look like a restaurant,
but Hyphen would be doing the actual food part. They'd be buying the ingredients making the bowls, and they would just charge me a fee for each bowl that I sold. So that is a universe where you have a sort of automated ghost or virtual kitchen. Right, Like, there's no restaurant as we think of it in that universe. There's not like a sign outside or whatever, but it exists on your phone, and it can make any number of kinds of food, presumably cheaply and quickly.
I mean, that is what excites me. My sister in law, Sathara, she's in Brooklyn. You know, she's always wanted to start her own restaurant. I always just imagine her, you know, in her apartment, you know, developing a recipe on HERKMS on our Hyphen Kitchen management system, taste testing it on a nearby make line, and then launching her brand on door Dash or uber eats, right and getting it fulfilled that way. Why Hyphen you know, can scale when she's ready.
So I kind of actually see it now. I think about it kind of like you know what happened in the beer industry where you had you know, Budweiser and Anaheuser Bush and then you had all these craft beers right and from Portland, Oregon, so IPAs and things like that. I see a similar movement happen where you go from overly processed food like McDonald's to you know, more of the artisanal meals that are you know, local to you.
Huh.
And then if it's like the beer industry, all those artists and all craft of bowl makers will get acquired by Chipotle, right, just like all the craft beer maker it's got acquired by AB and BEV.
That is true.
That could happen, and so that happens in this new universe because the startup costs are lower, essentially, because the startup costs can be very low where you don't even have to start a restaurant. You just go to you and say I want to start selling balls, and you're like, okay, you're not selling very many, so we're going to charge you three dollars a bowl. And I'm like, whatever, that's fun.
I'm just sitting here at my computer. And if nobody it's like, you're like the Zazzle Shop, but for salad bowls exactly.
And if we're doing that for so many people where they can do three or four bowls and they you know, it's again are like maybe a better example is like a Spotify or square space, right, Like they're designing their menu on the website, but it's effectively that easy to run. They're not making the food, they're not driving the food, they're not you know, having to buy real estate, right, they can do it all from their own dormer.
Apartment in that universe. The sort of ghost kitchen restaurant, it's really a marketing question, right, Like, I mean, sure you want the food to taste good, but it's like a fixed set of ingreded against the Food's not going to be that different. What's going to be different is like whatever can you get everybody on TikTok to buy the brand salad with your name on it?
Right?
Yeah, it's like TikTok becomes like QVC meets the Food Network or something right where people are like you know, selling their meals on TikTok.
And then what do you see happening with traditional restaurants where you you know, sit down and a waiter comes to your table and you order food from that person and there's a chef in the back cooking the food. Like, what's going to happen with that.
Yeah, I think it's going to be bifurcated. It kind of already is like the food experience are eating is quite bifurcated, right, You have there are times when you want to have these types of conversations, got to a fancy dinner, drink wine, have a som kind of you know, tell you about that selection, and you know, really really experience the food and enjoy the conversation.
And then there's times when you.
Just did you get to your desk, or like your kids are at home and you know the house is busy and you just need food like fuel. Are you love food and you don't want it to be fuel, but like you just don't have the bandwidth at the time. Are the money? I think that's where you know automation can help.
And so do you think that like classic restaurant isn't going to be any different than it is today? You think it'll be the same.
I don't think it.
Will change much because it's just like coffee, right, It's kind of built around this kind of artisanal kind of craft experience and it's really conduit to culture.
Right, So you have a lot of folks that are going there for more than just the food.
They're going there for the hospitality, and they're going for that entire experience. I don't think you can automate that experience ever, So I don't see that changing much at all. But again, the rest, the other side, the other fifty percent, is going to change massively.
We'll be back in a minute with the lightning round. Back to the show. Okay, let's do the lightning round. Okay, do you cook? Oh yeah, what's a go to weeknight dinner that you cook.
It's gonna sound silly. We love salads.
My salad that I love it's kind of like a caesar salad, but it's it's it's kind of got a twist. We borrow the recipe from John and Vinnie's, which is a pizza pizza company out here in La where it's kind of a spicy caesar. So you got a lot of uh, you know, spice and zing to it and some some breadcrumbs on top. I'm a pretty simple guy when it comes to my salads, though.
What's something you think should not be automated.
Any guest interaction, like any anytime, even concierge at a hotel like which could be automated. You know with a iPad. I think it's like anytime that you're interacting with a guest, you can't take that magic away or you've lost your value proposition. So I'd say like that front of house or you know, front facing roles or types of tasks would be terrible to automate.
Yeah, long term, you ran a frozen yogurt chop in college, so I got a couple questions about that first one flavor or.
Swirl swirl Come on, it's more fun.
Swirls a little confusing to because you end up getting like it's not like you're eating the one and you're eating the other. You're kind of eating both the all the time.
I like a little spice, a little little interest in my life.
Worse stopping.
Chocolate sprinkles.
Yeah, I think I agree. My kids like them, but I think I don't even get what's going on. How will you know it's time to do something else for work?
You know, I'm pretty persistent, but just like if I bang my head and I'm not moving forward, and how I I eating for it? You know, like measuring everything and kind of seeing how it's progressing. You know, it's
it's pretty demotivating and that's usually a good time. You know, if you're not waking up in the shower and thinking about, you know, what you're working on for the day, that's usually a good signal too, because usually when you have something like Hipeenner, you know, an idea that really compels you, like it's occupies your brain, share your brain space, and you can't get it out of your mind.
When you lose that feeling or that thought, probably tend to leave.
Stephen Klein is the co founder and CEO of hyphen Today's show was edited by Sarah Nix, produced by Edith Russolo, and engineered by Amanda k Wong. You can email us at problem at pushkin dot fm. I try to read every email I'd love to hear from you. I'm Jacob Goldstein and we'll be back next week with another episode of What's Your Problem