Grammarly Acquires Superhuman: AI Email, UX Moats & Building AI-Native Products (Loïc Houssier)

Now with AI coming, people are kind of like wanting to have what is the organic version of like the SaaS world. We just want humans. Okay? So superhumans make so much sense even more now with AI, because you want humans, you want interaction, you want like real stuff. The main difference between like successful companies versus not successful companies is execution, intensity, execution and like the way to. To brought this idea to life as part of it, you want to move fast, you build on top of OpenAI. You can be fast or anthropic or.

Whoever the latest best model is.

Yeah, exactly.

Welcome to the venture capital podcast VC fm. Today we're joined by Luick Hujier, the CTO of Superhuman, one of the most iconic productivity tools in the last decade. In fact, as I was preparing for this podcast, most of my friends were jealous that I was talking with you. It's redefining what it means to be truly AI native. Loic has always been leading the charge with integrating AI into Superhuman DNA, not as a bolt on feature, but as a core part of how people think, write and communicate.

And.

And you guys were recently just acquired by Grammarly and that is huge news. And if I'm correct, Grammarly will be taking the Superhuman name.

It's done. It's done technically, like Grammarly renamed themselves October 29, just like three months after an acquisition. So I can tell you my people were on deck full time some of the weekends to make sure we land that. So. So it's fun. So basically like, yeah, Superhuman acquired by Superhuman, something like this.

So we're spreading the news because a lot of my friends didn't realize that.

How much was the acquisition just driven by the fact you guys have the coolest name ever?

Oh, this is a really good question. This is a really good question. Honestly, the acquisition, like the price might have like been also depending on the brand. Indeed, because the brand was so cool. Yeah, it's versatile. There's a lot you can do with it. Compared to Grammarly, that was pretty much too narrow in terms of like semantics. So it just makes sense. It just makes sense. But Grammarly, now Superhuman is investing quite a lot into like communication and mail especially. So it's not just a branding, I would say brand acquisition. It's clearly like product market DNA, like engineering culture as well. Product culture acquisition. So it's full fledged. Full fledged. And cherry on the top, the name is cool.

Yeah, I think not. Not that we want to spend a lot of time on this because it is, you know, the name and not not the core of what you Guys do. But I do think in a lot of ways it's probably like one of the most interesting names in the a. In the age of AI to have. Right. This whole concept of, like, we're going to use AI to make you a superhero, essentially. Right. And that's what our brand really kind of communicates.

And one thing is what is like, really interesting is of course it was like, started before AI, before all of that, and it was like creating like bringing you, like superpowers at a time where AI was not there. So that was kind of like the DNA of the brand initially. But now with AI coming, people are kind of like wanting to have what is the organic version of the SaaS world. And like, yeah, we just want humans. Okay. So superhumans make so much sense even more now with AI because you want humans, you want interaction, you want like real stuff. So, yeah, it's. Yeah, that'll be cool.

I feel now as we were preparing for this podcast and we had a pre conversation with you, you, and we're talking about like the AI wars and like what's happening and how the world's changing, you shared an interesting thought where you said, you know, where. Where all these other tools are external to your process, like ChatGPT. Like, I go to ChatGPT, I go, I go to Google Gemini. Your main advantage is you're there where I'm at. Like, you're already in my Gmail, you're in my email. Like you're handling email for a lot of people. Talk with us about that strategy.

No. And that was one of the. So if you talk about superhuman male. So the original Superhuman. If we talk about superhuman male, one of the principles from, I would say design standpoint from Rahul the founder, was to not get into your flow. The users of the product were like really intensive email users, CEO funders, like, you talk with your VCs, you talk with your partners, you're trying to raise funds and everything. You need to be like, probably like seven hours a day in your email on top of your work. And one of those principle was don't get in the way, understand the flow, understand how people are, I would say are working and make that flow as smooth as possible. And the best way to do it is to basically like remove the friction. So when you think about it, Grammarly is really that as well. From a philosophy standpoint, like, Grammarly was where you work. They didn't ask you to do copy paste somewhere else and get away from your flow. Think about like, I don't know if you're on your email or if you're writing an essay like a bunch of students are using Grammarly, you want to stay in the context of where you are. You want to do like the correction. You want to do like the improvement of your writing style where you are. You don't want to get like somewhere else. Because every hurdle, every step outside of your flow is a friction and is reducing the pleasure you have to work I would say within a tool. So both Grammarly and superhuman. So basically superhuman globally now had this very, very intentional perspective that don't break the flow and improve the flow if you can and make the flow better. Because being in the flow is what makes us productive as knowledge workers. So I think this is where we have an edge as a company. We all are trying to improve the productivity of everyone and I think that these stance would make a difference.

Yeah, I love that approach. When you think about where does the product go from here, what are some of the use cases that you're most excited about that you're actually able to talk about?

Yeah, there's what I am excited about and what I am able or like entitled to talk about publicly. But there's a lot of things. There's a lot of things. One thing that is really exciting is the change of modality. Like the change of the way people are using and interacting with devices. Like if you ask my kids, they have their phone close to their ear and basically are talking to the phone and listening to voice messages more and more like texting for them is limiting, limited to environment where they cannot talk. So maybe they're on their bed and they cannot talk. Maybe they're like on a very noisy environment. But like their go to I would say communication channel is voice. So I am very interesting to understand how this cohort of like our generation of users will come and interact. I would say with laptops, phones and whatever like productivity suits they would be using. So most of us knowledge workers now are. I would say I'm talking every day to my device. Like I'm talking. I wouldn't name all the product that I use. Maybe I can like a whisper flow for example. But like many others, like I'm using that in Slack, I'm using that in like pretty much every surface because it's so much faster. Yeah, like I don't have to type, I don't make typos and my accent is not a problem for those tools. So which is cool. So I think that understanding how the user experience will shift and change is something that is really exciting, but it's a very macro. Macro triumph.

Are you coding by voice right now?

I'm. I'm so one. I'm not coding because I'm very rusty. I use, I would say a lot of, like. I wouldn't say vibe coding. Like, I think it's a very dangerous way to think about things. So you can. No, I love the flow. I love the flow, but not the vibe to some extent.

There's a technical founder here in Utah, and I won't mention his name, but he's trying to build his entire mvp. Like, a robust MVP without touching a keyboard. Working with AI.

I mean, well, it depends what we call mvp. If you're trying to raise funds, if you're trying to. Yeah, like, that's great. Like, you can get up with something. Like, the way we are using it is like all those side projects you don't have time to do because typically it would take time and everything. Sure. Like just cursor code, cloud code and cursor or whatever and boom, like, shift something on the side. You don't care about the quality, you don't care about the scalability. You don't care about, like, the security. But you. When you have like, a real work, a real application, and you start managing people's email or their contract or whatever, I mean, you don't want to fuck with the money. Part of my French. So vibe coding in those, I would say consideration, I would be very, very, very cautious about it, but clearly, I mean, I don't want to talk. I would say too deeply, but 95% of our, like, engineers are using AI. Probably 70% of our code is AI augmented. I wouldn't say fully AI generated, but like AI augmented. So, yes, it's a tremendous help. Vibe coding and everything to do like an MVP or to sell an IDE or to prototype. Sure. Please, please do it. My pm, I would say our PM at Superhuman, they do like, a lot of, like, demonstration of ideas with live, like, interactive product that are vibe coded. But when it comes to, like, yep, let's do this. This feels good. Like, now we. We do it right, if I may say so.

Okay, that makes sense. It's interesting that you talk about, like, how typing is slow and like, that's how I feel all the time is like, like, ah. But sometimes I can't talk to it because, you know, I don't want to be like the crazy person walking down the street talking to myself or there are other people or whatever. But yeah, it is, it's.

It's interesting.

It feels like there should be a way to move what I want in my head into the AI faster than what's currently available.

Yeah, so, so it's interesting like for one, like I'm a, I'm a productivity nerd and I'm like a typing nerd. So I have, I have like two different keyboards. Like I don't use my Mac keyboard. It's so like it's not good. So. And it's basically slowing me down when I type. So I have a mechanical keyboard. Like the weight like is like super heavy but like super clicky. Like I love the sensation. So there's also like a sensational, sensation oriented feeling about it and brings my like number of words per minute to be like way better.

Yeah.

But now that I the, that I talk to and like the transcription of what I want to say, it's like two, three times faster. And I talk fast. I talk fast. And the good thing as well is like I'm a rumbler. I rumble a lot when I talk about my IDs, I think aloud and everything. I don't care anymore. Yeah, I put that in the system and then rewrite that. Get rid of like the, all the feeling words and everything and you have like my content. So I think this is a really, really good acceleration. But what's next? We are already fair at meta. They already have some crazy devices. I wouldn't call them, I would say brain implants, but there's already something we can do. So in 10 years or my kids at my age, definitely there will be a new way to share your thoughts with an AI or whatever that is that will transform that into some sort of like a deliverable. That would be way faster than talking to it, I believe.

You know, the other thing too that I think a lot about is in the ad space, they are so good at predicting like what I'm interested in based on like the smallest amount of data.

Right.

Like if I am just thinking about like buying a kayak, it feels like all of a sudden I start to see kayak ads, right. And, and everybody jokes around forever about like, oh yeah, your phone's listening to and all those things. But I think a lot of it is like they're just picking up these things that don't seem to be related, but they actually are correlated and related. And then they're, they're showing you this stuff. And I've always wondered why, why we haven't brought that to more areas of our, of our lives with this where it would actually Be like more helpful and less addictive, I guess, in a lot of ways. Right. And just like starting to predict some of these things for us. So like in email, for example. Right. Like automatically writing. And I know that like companies like Fixer and Superhuman and others are starting to move in this direction, but it just feels like in the ad space, they figured out a lot of this with machine learning a long time ago, and it seems, seems like it's been kind of siloed there, but it feels like there could be a lot of broader applications. I'm just curious what your thoughts are.

Yeah, maybe. Maybe the first reaction to it is to some extent. To some extent. And the people that will. That work in that tech will basically scream at what I will just say, but to some extent it was easy. So it's pattern recognition and at the end you want like one world. Is it a bike? Is it a hammock? Is it a canoe? It's. It's easy to some extent. Now it's your context. So for example, when you draft an email, like it's not a world. It's like, what is the context of the project you're working on? What is the criticity of that customer you're talking to? What is the voice and tone you're using with that customer? Special? Because maybe you have some previous interaction with that person. And it's been like 10 years we've been working with him. Like, there's a lot of subtleties that are like way more complex that only like typical even NLP technologies. So back in the days, which was like the first approach to LLMs, to some extent that was not enough. That was not sufficient. Like the technology was not there. So yes, we had technologies to correlate some of the stuff you are saying to one word and say, yep, canoe, but like the context of your project, the criticity of your customer, your voice and tone and all of that. Different game. Yeah. So it's coming. It's clearly coming.

But now we actually have the tools to start parsing out some of that stuff that hasn't been possible before.

Right. And to close the loop, like we are talking about the fact that Grammarly is everywhere.

Yeah.

Compared to ChatGPT being nowhere or like in their own, I would say place, and they build connection to get access to that data. Grammarly is to some extent omniscient. Everywhere. It's basically with you all the time. So data. And pretty much everyone says it. Like in AI, there's two modes. Like there's the data and then not the models. But like the compute power that you have and the electricity behind it is right now like the fight for the big hyperscaler. And the data is there, the data is there. And there is, I would say balance between how much you want to look at that data, compute that data to be helpful for the end users. But like there's also something that is a bit scary associated to it. So. So we need to pay attention. But when it comes to the positioning, Grammarly is everywhere. Grammarly understands that you're moving from Google Dark where you were writing a draft about maybe like I don't know, a contract with a specific customer. Oh. And you're going to Salesforce and you're basically updating the context of the deal associated with that contract. And now you're writing in your emails, okay, we have this information. So the way to help you draft or craft your communication with your customer would be so much better. So much context aware because of the data we have access to. If you have ChatGPT, it's way harder. You need to. Maybe there would be some integration with Salesforce. There's some integration with I would say Google Docs, some integration. But then like it's so generic and so out of context that you need to bring that context there. So it makes things complex. More complex, I believe.

How do you.

Interesting positioning.

How do you think about kind of the debate around building on top of OpenAI or anthropic versus using open source, building your own models and your own LLM?

I think that's. I have a very naive answer to this. Uh, which is what I would say there's, there's like in startup world there's one thing that is making a difference between others that's not your idea. They're like, there's so many good ideas. Uh, the main difference between like successful companies versus not successful companies beside luck is execution intensity execution and like the way to, to brought this idea to, to life. Yep. As part of it, you want to move fast. Yeah, you want to move fast. Moving fast. You build on top of OpenAI. You can be fast or anthropic or.

Whatever the latest best model is.

Yeah, exactly. And does it make you like an AI wrapper or like an OpenAI wrapper? If you're really solving a problem, if you want to have your own model, your own infrastructure and everything. Sure. It might be, I would say cost interesting or like you will be slower to get to the market and someone else basically will get there before you. And like being the first is always like A significant advantage. Would you today start something on your data center with your own servers or would you go to gcp, AWS or you name it. Of course you will go with that because you don't want to build and take the time to create your data centers. Maybe at some point, and Basecamp is a good example, you want to move back because it's cost efficient. Eventually, but only when you have the success. So I would always start on top of those providers. Again, I think every founder, I would say think this way anyway, on the technical side, make or buy, I mean buy if you can just buy, move fast, prove your market fit, increase like your, your time. And when you're solid enough, okay, now you can think about like replacing those potential costly bricks in your architecture.

Yeah, I think that makes a lot of sense. How do you think about like building defensibility though, if, if you are built on one of these. Right. I mean the joke is every time OpenAI does a demo day, they wipe out thousands of startups.

I mean, my perspective is probably biased, but Superhuman? What is doing Superhuman that Gmail is not doing? I mean sure, like we have not. I make him caricature because we do more, but the reason people were able to pay and willing to pay for Superhuman was because we nailed a way better experience. Yeah, everything was faster and everything at the end of the day it's still emails, receiving emails, sending emails, but helping you to do it and doing it better. So if a company is built on top of OpenAI but nailed the use case, nailed the experience, people won't go back because you care about about like everyone is able to do anything. So the new moat to some extent is the user experience. Yeah, it's not the feature. If you build like a crappy experience on top of OpenAI that will be part of the roadmap at some point, sure, you will die. Yeah, I wouldn't invest, I wouldn't invest personally. But if you really care about like that's very specific niche on top of OpenAI and you craft the right user experience and it feels good, you have a shot. Yeah, you have a shot. So I'm not that afraid. I'm not that afraid.

It's interesting you say that because we've been saying that for a long time on the podcast that the AI wrapper, the thing that's challenging is that at the very early stages it's hard to know if a UX is really game changing or not. But the example I always think of is Canva and I've mentioned it before and that frankly there's nothing that special about Canva from a feature perspective. What is super compelling is the user experience. And that's how you have a company that competes head to head with Adobe and Microsoft Suite and wins because their user experience is just so much better.

And it's freaking hard to copy. Like getting the product sense, the product taste is something that is really hard to copy. And this is why I wanted to join Superhuman because I wanted to understand how they craft it, how they think about it. And you can't imagine the level of the. I would say scrutiny that is needed to. I would say check all the details, check like all, I would say part of the user experience so that it feels great. So interestingly like Raul, our founder is coming from, I would say the gaming industry and he was like a level designer. So he has like those game design approach to things. How to again think about the flow. You don't want the user to get away from the story. You want the user to stay in the story of the game. You want the user that has like really rapid feedback when he's doing something. Like if you're a console player, there's nothing worse than a lag between your command and the action on the screen. So everything associated with those concepts can be felt. In Superhuman, everything should feel fast. Like we have a rule internally. Like every single interaction should be I would say done under 100 milliseconds. Yeah, this is coming from the game industry, kind of like Playbook to some extent. And having this attention to details, like Raoul is always reviewing every single feature. Kind of like Branchevsky at Airbnb type of approach, like attention to detail and everything which is bringing everyone PM designers but also engineers to become very product centric, really understand how the user is feeling it. So that when they present something to Rahul, they already have. That is their answer to how is it improving the flow, how is it creating some friction? You want to provide some discovery. But this pop up is coming at the moment where that person was sending an email. This is painful, don't do that. So everyone is kind of like, I was about to say like brainwashed but like trained to have this muscle memory. And I mean good luck if you want to replicate that. It's not about like having like some nice designs that you steal on the web or whatever. You have to be like really intentional about it. Linear is a really good example of that approach as well. Very design centric. And like everyone in the company knows that the user experience is more important than the Feature, make it work. Not good enough. Make it great. Should be like the, the. I would say the mvp.

I feel like that's been like the Steve Jobs effect to a certain extent. Because I, I don't know about you, but I felt like there was a huge shift when Steve Jobs took over Apple the second time and just pushed it. Like, okay was no longer good enough. You had to be exceptional in terms of design and UX and quality of product and all of those things across the board. Or you just couldn't, you couldn't compete. Right.

Yeah. And it, I do believe that it needs a very strong funder.

Yeah.

And like to, to basically drive this. Like, it's interesting because Apple has been doing that for years, but job's gone.

Yeah.

It's like. And you would assume that everyone at Apple has like this mindset and everything to a certain level, but like the ability to keep and push like the limit is interesting. I will give you like just an anecdote, but I was talking with Rahul about like the quality of our AI inside. Superhuman. Superhuman male is a relative small team. Like we are like 120 people globally. Like in this, I was in this team. My engineering team is probably like 40, 50, something like this. But the quality of my AI is comparing me with OpenAI is comparing me with Gemini. This is the level of bar. I don't have this army of people to basically make it work. So. And this is the, I would say the level of bond that Steve Jobs had at Apple. Scale, of course, but like always willing to get like above and beyond. And I think when those founders are not in the picture anymore, the people that are behind you kind of like keep the last baseline of expectation and you can settle there. And it's good. It's good, but good is not enough. And that's, I think, what makes a significant difference between the founders like Raoul and of course like Jobs.

So one of the things that I think is kind of interesting, then you made a really interesting point, which is like, taste is so important. And I've been wondering a lot that, you know, for the last call it decade, we've been saying to people the very best thing you can do in college is go get a computer science degree. And I'm wondering if in the world of AI, the very best thing you can do is actually go back to humanities. Right? You know, the kind of the Steve Jobs perspective and like learn about people and how they think.

Calligraphy?

No.

Well, I mean, maybe to an extent. Right. Because it's, it's like how do you develop taste? And it's, it comes from like critical thinking and understanding art and people and. I don't know. What do you think?

No, no, no, this is a, this is a very, very interesting take. So for one we are pushing people to, to go into CS because it was like a good way to make money. Safe money. Yeah, let's put it that way. Does that mean that you will be like the best founder or whatever, but like safe money. So as a dad, like you want to do something, you don't know exactly what. Okay. CS can be fine. So it's kind of like the safe bet. Is it the same safe bet? I would say yes. I mean it's still fine. It's still fine. I don't believe you need to build those stuff. There will be always a need for new CS grad and working on all of that. That being said, clearly the people that will make a difference are the people that I call them product engineers. So even for engineers, and that's what I'm trying to hire like very product oriented engineers. Engineers that are like only core CS people. Like very like hardcore almost scientists like interested in like the wiring. They will be great for the hyperscalers, big companies because there's a need for that specific capability. But for, for startups, you don't want to build a new product. I mean you need those people with this appetite for design, for understanding the users and all of that. So to your point, I would probably push. So my son is doing marine biology, but if he had a doubt, I would probably something like major in cs minor in humanities and probably like human psychology, understanding the way people interact with things because this would be like the next inflection point or the next change. Like the interaction. As we were saying, keyboard and the screen. You've seen the meta glasses and the wrist brand and everything. The human interactions are changing. So clearly understanding the motivation behind the humans will be key.

Yeah, it's interesting at Superhuman.

How do you guys make sure that you stay competitive in just such a changing world?

I think the way we think about AI is mostly as like a new tech. It's just a new tech. Like at the end of the day it's a new technology. Do you care as a founder that there's a new database technology? You don't care. I would say engineers, they will make like the most of it. And so I see this kind of like when like the crypto, the blockchain came in. It's a new technology. Like don't be afraid like, and to some extent, just think about the use case. Is it enabling new use cases that you cannot be sold before or not? Maybe, maybe not. But like, focus on the use case, don't focus on the technology. So when you talk about the AI race, there's a new technology, cool. But do you need to push the technology up front and say, hey, this is AI, this is AI. End users, at the end of the day, they love good products that are solving their problems. If AI is helping along the way, that's good. So the race in AI, so it's interesting right now we don't talk that much about the capabilities of the models. It used to be like, early on we're like, oh, damn, GPT4 GPT 4.1, oh, Cloudsonnet and everything. And all those benchmarks on a daily basis, weekly basis, like to tell like, oh, damn, damn, damn, this is so good. This is so good. We are at a point where globally those models are relatively mature. Even like the open source ones, they can do some pretty cool things. So I think that the perception now is even like, I'm talking like LLMs and for language, but like multimodal models, like SOHA and like Grok, like managing images and videos, we are at a stage of maturity where we don't talk much more about, I would say the capabilities of the models. But we talk, we start like shifting back to like, what are the use cases? Kind of like at first there was Internet and like, oh, I have this new modem, it was like a 56k, now I have the new one. Or my Internet used to be like 1 mega 1 meg, now to 2 megs. We don't talk about that anymore. It's utilities, it's there now the question is like, what do you surface as value? So when you talk about the AI race, I think now we reach a point where okay, AI is there, engineers are using it. It's part of the, I would say the toolbox. Product managers, designers are also using that to expedit their process to work faster and use like rail plate V0 to get faster to like something that you can demonstrate. So it's kind of like shrinking time, which is like the one good thing. But then from the end user standpoint, it's all about like the use case. What are you solving? And I think we are pretty laser focused on solving like the communication space with the superhuman male. And we are doing a great job there. So I don't feel the race is AI specifically. It's still the typical business race. And we want to be the best at solving the problems. I'm sorry, it's a very canned answer, but that's the way I think about it.

So I'm curious though, to push on this a little bit more. Do you think we've kind of achieved what we're going to on AI capabilities? I mean there will always be increases, but we're hitting this point of diminishing marginal return.

No, that's not what I'm saying. What I'm saying is that the shift in terms of how we look at the race and what makes a difference was very almost infrastructure oriented. Is this model able to solve whatever, like medical test? Is this model able to solve X, Y and Z? So talking about almost like the, the mathematical aspect of the power and raw power of those models versus looking at what they were able to solve in the real like B2B SAS world and all of that. So that's the inflection point that I've seen. Like right now we're talking much more about what we solve rather than, hey, we have the latest model. Cool. What about me? So even you can see it like when you look at the, the different billboard in San Francisco or like on the 101 heading south or back up. It used to be like a very like, oh, models, GPUs and all of that. More and more it's coming about like, what is the end goal? What are we solving? Sure, we use AI. Like everyone, everyone is using it. So I think this is the shift that I'm seeing now. The capabilities of the model, yes, they are like always improving, but the same way Internet capacity was always improving, the same way database kept improving. So I think it's coming into a world where we don't care that much from a business standpoint about like those model capability. Sure, there's infrastructure players that will be like building those models and everything and it's important. But for the rest of us that are not, I would say core model players doesn't matter as much, I feel.

Yeah, it's interesting. I read something that Vinod Khosla made a statement recently that he, he was like, basically there is a company building an AI tool for every job out there that humans do. And to your point, like, it's kind of like the AI models have hit this point where they're good enough to start doing most things, but maybe the difference is that they're not. They don't have the data sets and they don't have the direction to handle very specific tasks. And that's what everybody's kind of working.

On, if that makes sense about the fact that so one do I agree that now for every single job someone is thinking about how AI can improve it or replace it 100%. Of course, of course on the second piece that those products will still like direction on the percent. So the way I foresee the future is everyone will have their own Jarvis helper doing whatever job they're doing. But I strongly believe that there will be like always a need for human in the, in the space making those calls, making those decisions. I see this a bit as a. So there's a really good book that I usually mention as part of the management practice which is Turn the Ship Around. Long story short, USS submarine commander talking about his leadership and how he build a model of leader of leaders rather than a leader of followers. I think that we can make almost the same analogy with the world of AI agents. Like as a leader you might have some agents but you still want to be a leader of leaders even if they know their task. You want to be the one making sure that everything is contextually right. You want to be there to orient, adapt, even if they have the capacity to do things instead of you. So I think that there will always be a need for like a USS commander to do the thing even if you have like the right team of agents I would say working with and for you. So yeah, that's the world I hope for. We are augmented but we are not replaced. So maybe I'm biased because I'm French and I have a social I was a democracy background. But yeah, that's kind of like the way I think about it.

Has AI affected your hiring strategy or your headcount?

That's a good question. For one, I think we do more with less. That's very true. So probably I would have higher, faster or sooner without AI. Now we're hiring, but I think it's allowing us to do way more than we used to do before. Like I see it like I would give you like some silly example of how AI changed my world, but I do, I do a lot of hiring. I used to do 45 minutes hiring 40, 30 minutes. Kind of like doing the wrap up, writing my notes and sharing that I was in the system so that others can take a look at it. Now it's taking me two minutes because I use like a meeting recorder. It's structuring my note the way I love it. So it's. I have already my notes that I'm taking live on top of this Transcript. It's formatting that in the way I want to push it in the system. I just review things. No, no, no, you. Five minutes done. So and I do like three interviews a day. So you can imagine it's like one hour and a half that I save each day in my, I would say day to day work. And I'm only talking about the hiring process. It's solving so much time. Also, like in terms of reporting, like as a leader you need to provide some KPIs, some reporting to different teams and everything just to take a shit ton of time. Like basically my Sunday afternoon, like one hour in the afternoon. On the Sunday, I was basically writing my Monday morning email. Now it's taking me five minutes because I have the context from all my teams and they're doing the same. And boom, boom, boom, boom, boom. You consolidate. You had your human touch on it and maybe like you put, I would say the highlight on what you think matters. Ten minutes, not one hour. So to answer your question, yes, it's changing. I would say the way we think about hiring, but not because we are replacing jobs, but because my team is able to do way more. So like the hiring needs come out slightly later. Now when it comes to hiring, it's interesting. I'm doing like rounds of interview for PMs. If a product manager is not using v0 is not using like those by coding tool to basically share their thoughts, share their thinking with designers, with engineers, I mean, that's kind of weird. So clearly this is something that you want to understand and for engineers, like you want to understand their flow, you want to understand how they're working. I would say how have they used AI? You don't care about like how they use it that much. But like are they using it? Are they thinking about it? Are they constantly challenging their own ways of working so their flow to augment it? So that's part of the, I would say the interview process. Now if the question is, do we allow AI to use something? I think it depends on teams at superhuman Grammarly, some people refuse and some teams are basically saying, yep, as long as you can explain the code that has been done and you explained the way you've used it to get to that point. And when we drill, you have some solid answer. Sure. Now we don't force the usage of AI just yet, but the same way we don't force the usage of a specific language or a specific database. We care about your understanding as an engineer of the thing that needs to be solved and your experience should tell us about your ways of working. I don't know if it's answering compared to your question.

Do you anticipate that people that don't use it won't be able to keep up, or is that not a concern and that ultimately you don't enforce it explicitly, but maybe implicitly because lower performers, you know, get replaced?

Yeah, I, I think that AI is really good at exposing or like making the divide between average engineers, or let's say bad engineers and good engineers. A good engineer with AI would be so much faster. Yeah, a bad engineer with AI would be even worse. So because it's kind of sharing crap that that person won't be able to really like, clean up and everything because the way they prompt is not, I would say, organized. I would say the right way and everything. So I think this is a. This would make a. I would say a very strong difference.

That's an interesting perspective because I've heard from other people that have said, like, you know, your top engineers just get better and your bad ones also get better.

But.

But you think it's actually the opposite.

I think they get faster, but they don't get better. So they get faster or whatever. They do more crap faster.

An example I saw yesterday is I. I sprained my knee and I'm gonna, I'm gonna use a non coding example, but I asked it like, hey, what could I do? I think I have a patella inflammation. And it told me to do the RICE method, rest, ice, compression, elevation. And I went back to it and I said, isn't this an outdated method? And it said, yeah, this is an outdated method. And I'm like, when was it outdated? It was like back in 2018. And to your point, I think an experienced person who's very studious of their craft, whether it's coding or engineering, will be able to pick up on things like that. But someone who isn't well studied would potentially continue going down the wrong path. That ChatGPT could have sent me down.

Yeah, and there's small stuff that of course LLM will help pretty much anyone. There's a small bug, like a small fix, and you say, hey, fix that bug. If it's like very local, like a typo, like a typo in your product and you want to fix it very easy and a crappy, like, engineer would do it and it would be faster doing it, great. But like everything that needs some level of complexity. It's kind of like garbage in, garbage out. Like an LLM. If you ask the right way, you have the right, I would say output. If you ask poorly, the output will be highly dependent on your query. So that's why I think that best engineers have like the best way to. And coming back to your humanities Peter aspect the best engineers, and this is what we're hiring for because we're a remote company, are the best at explaining what they do. And to use an LLM or like those tools, the more explicit articulate you are in providing the context of what you want to change and the way you want it to be to be changed will help the LLM to be more specific and successful in getting the tasks done. Compared to someone that has ideas pretty much everywhere, not able to articulate the need really, the LLM will try to infer what you're trying to achieve, but the chances of success are I would say, less likely to happen. So coming back to those humanities, it's very interesting. I do believe remote workers would be remote engineers have an edge because they're already used to articulate their thoughts. They're already used to work asynchronously, write down their thoughts for other engineers to consume. So those people have an edge using AI in the, I would say engineering capacity, but I think we can expand that to the. I would say the rest of the population. If you're a good communicator, you would be a good user of AI.

One of my thoughts was some of my, my junior engineers is I feel like a couple days a week they shouldn't be allowed to use AI. They should have to struggle through the problem. They should have to to exercise critical thinking skills. What are your thoughts on something like that?

That's an interesting point. I think for one they should be, I would say supervised with someone that has the scars, someone that has the review so that they can ask like the painful question, why are you doing that? And not that? What was the trade off? How did you met the trade off? If the answer is like, oh, the AI told me, I'm like, nah, wrong answer. Think twice. I think you need someone that is able to basically challenge them and obviously review the code. But really asking about the thought process, when you have a junior person that is doing something and it's not exactly right, you want to dig in. What was your thought process? Tell me about your process there. Whether it is like an engineer or not or going through the mental mindset. I was at the process that they were going through the frameworks they were using so that you can fix their ways of thinking other than just fix the outcome. So your Point about like hey, don't use AI on a Monday, for example. It's one way to do it. It's pretty harsh. I'm. I don't know. My, my, my father was a blue collar and he was always telling me like always let like the is a steel worker. Steel worker choose their own tools. Like you don't care what the brand is. Like they have the habit, they have their flow. They, they know like the how to use the tools like the right way. Don't force them to do like something different. They know better. So I think that I'm always trying to apply that to my teams. Like I don't care what your tool is. You want cursor, you want VS code, you want like OpenAI codecs. I don't mind. I don't mind. Yet own hone at your flow.

But choosing your own tools is a different argument than strengthen your brain. So in my case I'm fairly consistent with no AI Fridays. And writing an email now after like I use AI for everything, I'll say here's the bullet points. Sometimes I'll, I'll screenshot the prior messages and I'll say help me craft this message for this response. And then, and then I spend most of my time reviewing. But writing an email from scratch after two, three, four years of extensively using ChatGPT and Claude and Anthropic and all those things, I noticed it's significantly more difficult to put. Put it all together from scratch.

That's an interesting point. Never really thought about it like the, the way like what's coming to my mind like think about like a CEO in the 1980s, like 60 year old, I would say CEO never had the time to write any like mail using an executive admin to do it basically like talking I would say through like the context and everything. The admin would basically write, I would say the email and he will or she will just like review the copy. Like nah, rewrite this sentence. Like I don't like it doesn't sound like me and everything. This is what we're doing with an LLM today. So would you say that this CEO from the 1980s was kind of like brain slow or losing his ability to craft a good communication because he was only reviewing content that was proposed by the administration?

Or was he getting leverage so he could use his brain for other things that were higher value?

Yeah, or like making sure the structure is right rather than just like you're finding the right word at the right time? That's a good question.

One of the Things that was a big unlock for me with around Vibe coding was I tried to do a bunch of things and I realized what I was doing is I was just like outsourcing the task but not really understanding it and then like getting frustrated because I get stuck in some spot and I shifted it from hey, I'm gonna outsource this task to like, I just want to learn and understand it and get options. And it was really interesting. Like it, it shifted and I started to make a lot more progress, a lot faster. And, and it wasn't so much I wasn't just using AI. It was like, okay, use the AI to teach me. Let me make the critical decisions on which paths we're going in the app development and help me understand pros and cons of both. And I don't know, I found it really helpful, but also a way for me to continue to think critically throughout the entire process. And really like, as somebody who's not a computer programmer by profession or training, like was really cool to see this whole new unlock of opportunities and abilities.

And two anecdotes on this. But this is the way I teach my kids to use LLMs, ask questions, make it a tutor that is asking you questions and challenging your answers and explaining you why rather than just regurgitating whatever knowledge. But also in our executive team at Superhuman Mail, Kristen, she's our chief business officer and I basically replicated one habit that she has after some one on ones that she had with the team and her team. Everything is recorded. Of course we all have like granola or like you name it, like there's so many of them now. But you can ask after the meeting. Please challenge me. As I was enjoying my one on one. What was great during it? What can I improve? And it's really interesting. Like it's. You're facing like, oh, you shared that feedback with that person, but it was not timely because you're mentioning something that happened like three weeks ago. Next time doing right after stuff that you know, like you went to some like management course or whatever. Of course I know. But when you have like this constant feedback about these specific meetings. Oh, you talk too much. Like, that's, that's me. I'm super verbose. I talk a lot and I enjoy talking. But like having the granola or like whatever recording that is telling you maybe you are a bit too long and not concise enough. And it's coming, it's coming, it's coming. Basically eating my brain. And I'm trying to do better now. Not obvious during this discussion with you guys, but I hope it's more obvious in my one on ones with my team.

No? Well, this has been a super fun conversation with you, and we've appreciated all of your robust and verbose perspectives and opinions. So thank you for being on the show and sharing all that you're working on and your perspectives on what the future looks like.

Thanks a lot guys. I really appreciated that time. And yeah, that was cool. Thanks a lot.

And thank you so much. And what's the best way for them to follow up with you if they want to learn more about what you're up to?

LinkedIn is my go to. I'm an avid X Twitter user, but I usually don't post, so LinkedIn is the way.

Okay, sounds good. Well, thank you so much.