Practical AI

Jerod: 00:04

Welcome to Practical AI, the podcast that makes artificial intelligence practical, productive, and accessible to all. If you like this show, you will love the changelog. It's news on Mondays, deep technical interviews on Wednesdays, and on Fridays, an awesome talk show for your weekend enjoyment. Find us by searching for the changelog wherever you get your podcasts. Thanks to our partners at fly.io.

Jerod: 00:28

Launch your AI apps in five minutes or less. Learn how at fly.io.

Daniel: 00:44

Welcome to another episode of the Practical AI Podcast. This is Daniel Witenack. I am CEO at Prediction Guard, and I'm joined, as always, by my cohost, Chris Benson, who is a principal AI research engineer at Lockheed Martin. How are doing, Chris?

Chris: 01:01

Doing very well. Looking forward to talking some fun stuff on this beautiful spring day.

Daniel: 01:08

Yes. Well, I've always hoped that AI could make me a superhuman. So really excited to hear about maybe something in that realm today from Loic Ussier, who is head of engineering at Superhuman. How are doing, Loic?

Loïc: 01:27

I'm doing great. I'm super excited to chat with you guys, I would say with a pretty humbling, I would say, set of guests in the past. So I'm super happy to have this opportunity and discuss in length.

Daniel: 01:40

Yeah, that's awesome. Well, know this is kind of interesting because I know Superhuman, I think, was one of maybe the sort of first really integrated AI first kind of engineering tools that I remember seeing. And of course, the AI space has advanced a lot in that time. Maybe could you give us a little bit of a state of AI in email or productivity more broadly, if you want to think about But really, obviously, we're going to talk a lot about email and messaging. So could you give us a little bit of a sense of what that landscape looks like right now and kind of how Superhuman fits into that?

Loïc: 02:21

Yeah, totally. And and it's like a it's incredible, like the time that we're living right now. Of course, like everyone has been shocked, like, when we had, like, the first version of those LLMs, like, doing some crazy, crazy stuff, and, like, analyzing text, summarizing, and doing, like, all sort of of magic. And and, course, email, I mean, is all text based for the most part. And so it was like a really nice test bed, like to try out, like all the cool stuff that you can do.

Loïc: 02:55

And interestingly, that's also like, helped like this category, like the email client to thrive, like for quite some time. Superhuman was almost the only one supercharging Gmail and Outlook. We were the only one on the space making like people faster, going through their emails, and all of that. And with the rise of LLM and agents and everything, now there's a bunch of people are like, oh, damn, this is like a great, great, great environment to play around and to make like things better. And right now, this is what we see.

Loïc: 03:29

We see like a bunch of, I would say, other like tools trying to do stuff with LLM, and to create like a better experience for email, and this is indeed like an interesting time for us, because this is the proof that that category needs to exist, was existing before, but we are the only ones there, and now that more and more people are getting there showing that there's deep interest in it. And, and it's challenging, and it's like super interesting, and would probably talk about it, but it will also like help us understand what makes like a good product, and is like just the LLM and AI sufficient, or like do you do you need some sort of like a secret sauce on top of it? And happy to to discuss about it.

Chris: 04:15

As you I'm curious, following up on on what Daniel was saying with, you know, with you guys being so early into the space. And and obviously, we're you know, the not only LLMs, but just AI in general has been going at light speed, you know, increasing steadily over that time period. How is that? How is the the space change for you guys from being kind of the early only player, you know, into into the space where there's others, you know, it's becoming, you know, somewhat congested across not just the space you're in, but just like everything. How has that changed the world for you guys in terms of staying differentiated and all that?

Loïc: 04:56

Yeah. So it's very interesting because, like, there's multidimensions that we can talk about. Like, the first one is this is, like, the raise of those AI features and capabilities are bringing, like, a new set of features that you can implement that you couldn't do in the past. In the past, AI was mostly classification, like adding labels and and stuff like this, and that was kind of like the limit of what AI could do really for everything that is text based, so typical classifiers, typical models like this. And more and more now, you can do, like, some intelligent stuff.

Loïc: 05:34

So we moved from a place where we were making things faster for our users compared to Outlook, compared to Gmail. But now there's like more that we can do, we can make things smarter, which is probably like a parting shift in terms of like the value that we're creating for our users. The other dimension is like, this is raising the expectations for, I would say, the different users. Like, for a long time, they were like, damn, this is so fast, and I'm, know, I'm gonna say, winning like four or four hours a week to go through my emails, but like, everyone is used to chat with GPT, everyone is used to the complexity, everyone is like crafting images, or like even movies with SoHa and all of that. So like the level of awareness and the level of understanding of what the technology can do is raised dramatically.

Loïc: 06:27

So for our users, the level of expectations like, hey, superhuman, I expect this now. I expect this now. The other dimension is like from an engineering standpoint and like a building standpoint, our tool set is totally different. Like, the the tool that changed, and engineers that were working like in in some ways three years ago, even two years ago, even six months ago. Like right now, the tool set and like your flow and like all your setup to work has dramatically changed.

Loïc: 06:58

And maybe like the the last dimension that I think is like really tricky to apprehend is the perceived quality. So Superhuman was seen and built on the kind of like the the one single dimension that was like, it's highly qualitative. We were in charge of the quality because we master everything. So you can be like, have like a zero bug policy. You can take the time to deliver the value, but it needs to be perfect.

Loïc: 07:27

And now, with LLMs, a bunch of the perceived quality depends on your prompt. So you have users that are prompting with different skills or different level of skills, and the outcome of that prompt may be perceived as low quality. But that's something that is really hard to control. And it's creating like something that is like sort of like mind blowing from an engineering standpoint. I mean, we've all been working in tech, and the craft, the bugs, and everything.

Loïc: 07:57

There are some processes to limit the number of bugs. But now, quality is not only bugs. Like, it's also like this perceived quality based on the user, and that's an interesting thing to tackle.

Chris: 08:11

And I'm curious, as you kind of mentioned the fact that with some of the prompts, and having different users' skill level and stuff like that, could you talk a little bit about how you tackle This is one of those interesting things, from my standpoint, to hear about where there's all these little gotchas in this world that a typical person isn't going to ever have thought about going ahead of time. And so as one of those things where prompting itself is fairly diverse in terms of the skill set. Can you talk a little bit about like, how do you deal with that, when you're trying to put together a product and focusing on the quality issues and stuff like that? Because I'll be honest with you, that would not have occurred to me to have to think about addressing that kind of issue. Can you talk a little bit about that?

Loïc: 08:59

No. No. It's it's I will tell you, like, about one specific feature that we released, like, in q one. So we have those auto labels. So automatic labels that will basically flag your emails, and based on the label, you can decide to skip your inbox altogether.

Loïc: 09:17

Typical stuff, like random pitches from a company that want to get in touch with you to sell their product. I receive like probably like 30 of them every day. Do I want to take look at those, like, 30 and answer, like, all of that? Probably not. Probably not.

Loïc: 09:35

So I'd love them to to basically, like, be skipped altogether. So for those, we built classifiers that do not rely on user prompts, so that we control the quality, precision recall, like the typical stuff. But we also allow our users to create and craft their own labels. Let's say, like, you want to have like, oh, I want all my podcast invitation to have the same label. But like, cannot just have like a deterministic rule to say it because I don't know all the podcast, like people and everything.

Loïc: 10:11

So you cannot just do like the filter, like Gmail would do, where you say like, if then, then this. So you have to point it, and you have to basically allow the user to craft a point that will surface all of those. But then, that point is tricky, because like if you have someone that is just like putting just a one liner, you you start having like some issues, because the precision and record based on the one line prompt is not great. And we know, like, as you, I would say, I guess your audience have been working with like, chat GPT, or like prompts in general, the more structured and extensive they are, the better the result. And there's a bunch of isolation hallucination that can happen if you are like just one liner, because lack of context and lack of all of that.

Loïc: 11:00

So of course, you do like some like system prompt to suit basically surround this user prompt to try to avoid like too much too much issues, but there's also like a part of education that you need to have, and we are working on this now, which is like, your prompt seems interesting, but like probably you want to structure it that way. So there's some stuff like this that we will be working on. Also sharing prompts, like libraries of prompts is something that we're thinking about more and more, because not everyone is able to craft a nice prompt, and maybe someone in your team will have done like a prompt that you would really use happily if you get access to it. So it's sort of like, mean, it's very product centric, so it's not AI centric, and you need to work around this new problem. And I wish we had a silver bullet, and like the answer to that problem, but I think we are like learning as we walk, and it, but it's super interesting.

Daniel: 12:00

I'm wondering I'm always intrigued by I read a book by Richard Hamming, and one of the things that he talks about is how if you rethink a process that was very human and manual before, often the way that you would make that an augmented or machine driven process is very different from what the original human process would look like. I think in the email client, we all sort of expect a certain process, a look and feel to the email client that's developed over time. What have you found in terms of presenting an email client to a user that is drastically different? What sort of needs to be preserved? What's kind of up for grabs in that experience?

Daniel: 12:45

What should stretch the user? What needs to be preserved? How do you think about that?

Loïc: 12:51

That's really interesting. That's a really interesting point because we are at that moment where the user interaction with the computer, with the system is, like, dramatically changing. Like, people don't expect to click in different windows anymore. Like, the expectation is different, like ChattyPT to to, like, or I would say the the other like clones, from like different providers, you basically have a chat box, and you ask everything there. Like even if you're working on a document, you ask on the chatbot, and like, modify my document and rewrite my exact summary.

Loïc: 13:26

Oh, make my tone a bit more like X and Y and Z. You don't expect to have like a button like World would have like Microsoft World back in the days. So, and we are only at the beginning of this shift. So I think that, it's kind of like coming back to like competition and all of that, the barrier to entry to like pretty much any SaaS application, or even the consumer application, is very low now, because it's very easy to, at least to build a POC, at least to build a POC. I wouldn't go like on like further than that.

Loïc: 14:01

And what will make the difference is the product test, and how you want to understand your users, and how you understand their user interaction. And this is where like I feel pretty, pretty proud to work at Superhuman, because our CEO is a freak in terms of like user interaction and vision, and he's already thinking about that, and like how the future of interaction will be, and it will change. It will be different. So like what will stay, what will, what will be slightly different? I'm pretty sure that the conversational aspect would be a strong paradigm.

Loïc: 14:41

Like right now, you don't talk, whether it is like through your keyboard or through a mic, you don't really talk to your system. You don't talk to the application. Maybe you start talking with Charge ept because they have this nice voice interaction. Maybe you use Whisker Flow, or like these type of tools to basically write your email, or like write in Slack and your messages. But you're not exactly commanding the device to do things as you talk just yet.

Loïc: 15:10

But more and more people are doing so. Like, I probably talk to my computer now more than I type, interestingly. So there's a change. And everything that we've done in the past was mostly click and click and click. Superhuman started with like the Common K and keyboard centric access to things for people that wanted really productivity, because like switching like with a mouse is like, is pretty slow.

Loïc: 15:39

And now more and more people are starting to engage with the voice. So all of that will change the way you think the way you face the data, the way you interact with the data, the way you bring the focus. So this is an interesting, I would say, area. One thing that I do believe will stay though, to your point, Daniel, and I would talk about email especially, The concept of inbox, like the concept of having like some sort of like a timeline of things that you need to go through, and get rid of the stuff that are top of mind, some sort of like a task list to some extent will stay. Now, how it will be surfaced, how you will go through it, will dramatically change over time, and we are already like seeing this.

Sponsor: 16:38

Okay, friends. Build the future of multi agent software with Agency, AGNTCY. The Agency is an open source collective building the Internet of agents. It is a collaboration layer where AI agents can discover, connect, and work across frameworks. For developers, this means standardized agent discovery tools, seamless protocols for inter agent communication, and modular components to compose and scale multi agent workflows.

Sponsor: 17:09

Join Crew AI, LangChain, Lambda Index, Browserbase, Cisco, and dozens more. The agency is dropping code, specs, and services. No strings attached. You can now build with other engineers who care about high quality multi agent software. Visit agency.org and add your support.

Sponsor: 17:32

That's agntcy.org.

Chris: 17:40

So as we were kind of going into the break, we were talking about kind of, you know, the the notion of rethinks on that. And I'm curious, as you're thinking about not only the rethinks, but you're also having to respond to the evolution of the technology itself that's available for your teams to implement stuff. And one of the things that we've seen over time is kind of, it's not this smooth increase. You may have evolutionary increase in the model capabilities for a bit, But you also have these jumps that'll occur along the way. And with your product teams, as you're looking at what the future of your products are going to be, and you hit these moments where it kind of goes from predictable improvement in the models, and you make these jumps.

Chris: 18:36

How does that affect the product development cycle that you have internally when you're saying, are those moments, you know, we're talking about rethinks, do you have moments where you kind of go, maybe it's time for kind of a deliberate rethink because something just happened in terms of the technology capability that we weren't expecting last week. And we're gonna do that. How do you guys handle this kind of an industry being in it

Loïc: 18:59

for that? No, it's it's very interesting. Okay. And so Danielle, you were mentioning a book, but one book that comes to mind as you're asking this question is Zone to Win by Geoffrey Moore. And he talks about, like, continuous innovation and, like, disruptive innovation, and this is probably what we're talking about.

Loïc: 19:16

Like, we continuously innovate, and we continuously add more features and new stuff into the product, and sometimes, you have this opportunity to provide something that is disruptive, whether it is like the underlying technology that is disruptive, or because you have like some sort of a wow moment, and you are like someone, I would say with a vision that is like, this is the duration to take, and we need to either pivot, or we need to do something like drastically different. What we've seen, especially with AI, is like the rate of those disruptive innovation is mind blowing. I would say before AI, to some extent, like the technical innovation where maybe once a year, once every two year, like you have something that is like brand new and like, holy shit, I need to use this, and pardon my French. But what is interesting with like LLMs, like every two weeks or three weeks, if you're not on Twitter, you're not on Hacker News, like, you can miss like the new big stuff. Like LLM, like multi models, reasoning, MCPs, like every that came in six months.

Loïc: 20:23

And all of that is coming with like a new set of capabilities that you can decide to implement in your product. So to come back on your early question, what is the impact on the product development? How do you handle this? One, you better be agile, meaning like the true agile. So you better be able to stop what you do and say, wow, focus, we need to sit down for a moment because this is coming.

Loïc: 20:48

What do we do about it? And you need to have like and that's why I love like a small companies to some extent, because it's very easy to have like everyone, listen, there's this new thing. We need to do something about it. Let's change the roadmap right now. When you're in a bigger company, it's way harder to do it because you have your early planning that is like coming into quarterly planning, and you have all those OKRs that you you need to report on and everything.

Loïc: 21:15

So like you need basically like almost a six months business plan to explain why you want to to pivot into something else, which is obviously not the case when you're a company that is of a small size. Superhuman engineering and product and design is probably like, don't know, I don't have the strict number, but like forty, forty, six, maybe 50, but that's about it. That's the the size where you can be like super agile. You can stop everyone doing something because something is coming up, and we need to to focus on it. Of course, we can do better.

Loïc: 21:49

If my engineers are listening to this podcast, they would say, like, maybe you're like your caricaturing a bit. So probably I'm caricaturing a bit.

Chris: 21:56

And of course they are, right? I mean, Of course, they're right. Of course they're listening to you.

Loïc: 22:01

And of course they are listening to it. No, so it's having this understanding that everything is changing right now. So you will you need to reassess your priorities like almost every two weeks, almost every two weeks. MCP is coming, people are standardizing on it right now. What do we do with it?

Loïc: 22:20

What do we do with it? Should we like invest like crazy? Should we stop everything that we are doing? Should we, I will say, we still believe in the vision, and it's providing more value, you need to make those decisions every two weeks. So, or like almost every week.

Loïc: 22:34

So being close to, I would say, the close knit team that is talking like basically on a daily basis to make sure that you're making the right decision is key.

Chris: 22:44

And by the way, just for listeners, you may have heard MCP in there if we did an episode explaining what MCP is. So anyone who's not familiar with it, you should jump back a few episodes and hear that out. It'll give you some context around that.

Loïc: 23:00

Yeah, thanks. Thanks, Chris. And I'm sorry, if I use, I would say some jargons, but

Chris: 23:04

Jargon jargon is fine. But we all we always try to jump in and point people to it.

Loïc: 23:09

So this is perfect. This is perfect.

Chris: 23:11

And I think, of looking forward, one of the things that that I'm really curious about is we've tackled some of the bigger issues of AI and email. But I'm curious, if we dive down into specific functionality at Superhuman, how do you see kind of the most, maybe the most useful AI email functions that you're currently either kind of releasing or kind of thinking about forward? You know, how do you, when you get kind of granular on the product, how are you starting to think about that now?

Loïc: 23:45

The I would just like like the the the feature that all our users are basically talking about because they just love it is a feature called auto draft. You receive an email as part of a thread, someone is asking you some questions, and or you send an email basically saying like, hey, can we meet next week or whatever, and after two days, you don't have an answer. You usually want to bump that into their inbox and everything. We build this feature where we create those drafts for you, ready to be sent. It's, it's not mind blowing in terms of like usage of LLMs, like you provide the context, you use the tone of like you're with that person and everything, and you craft a draft that could sound like a good way to reply to it, and the results are just mind blowing.

Loïc: 24:34

Like the users find it like so addictive, because it's relatively accurate, and they win a lot of time. Like, it's just about like winning time. Our users are mostly CEOs, CXOs, on the sales side, as well as some consultancy firm. They leave, like, basically day in and day out, like in their emails. So every ten seconds that you can make them win in their day, it's a huge win for them, given like the amount of emails that they have.

Loïc: 25:05

So this is like one of those feature that is super effective, even if it sounds simple.

Daniel: 25:12

So, Loic, even with what you're just describing there, creating an auto draft per email, maybe an LLM call, doing classifications, auto labeling, maybe other calls. I don't know how many calls or chains of LM calls are happening per email, but that could potentially be a lot. And if you do that for one email, that's fine. You do that for all my emails, that's more. If you do that for all emails of thousands or hundreds of thousands of people, that's a lot of GenAI workload.

Daniel: 25:48

How does Superhuman as more of a AI application company think about that in terms of optimizing infrastructure or AI, Gen AI use, consumption, hosting your own models, fine tuning your own models, using smaller models? How do you think through some of that?

Loïc: 26:08

So this is a great question, and this is a real challenge to some extent, if not a problem sometimes indeed. I guess, like, my engineers are, like, very much into, like, the finance. Like, they understand, like, the cost of inference, the cost of the input, the cost of the output. They understand the difference between the different models. So we we have to put some some sort of like a high level principles to keep moving fast so that they know, I would say, how to default, and only like escalate if they they have some questions.

Loïc: 26:41

I will give you some example, but if it's a new feature, we don't know if it would be working or whatever, so still testing. We want it to be great. So we take the most expensive model. It's working, and we have traction. Great, good problem to have.

Loïc: 26:59

And now this is the moment where you start thinking about like optimizing the cost, and maybe you will switch to like a cheaper model, maybe more fine tuned, maybe you would switch to like a different type of model altogether. So for example, like the classification that we discussed, LLMs are okay with classifications, but you can have way cheaper for the same quality with like a BERT type of models. And inference cost is like a fraction of it, a fraction. So long story short, this is the way we provide, I would say, value to our end users. We try with the best working, we do optimization after the fact.

Loïc: 27:37

Does that answer your question? But this is like, more generally, I think this is like always like the right approach is like, don't care about the cost right now if it's not becoming a problem, because you always want to provide like the best experience, and and if you don't have traction, too bad, because the risk, if you try to start small because you're afraid of the cost, you will use like a cheaper model, and the feedback from the users would be meh, and they won't use your feature. And then you don't know if it's because the feature is, I would say not well targeted, or if it's because of the model. So targeting the best, you have like better answers and better on better insights.

Chris: 28:19

That was a really interesting answer from my standpoint. You explicitly called out, as you're getting to the feature and going ahead and going with the best, the most expensive thing, and then pulling it back to what the efficiency will be. And once again, one of the things that we often call out on the show is kind of the fact of kind of software engineering being applied and kind of the analogies on that on the AI side. So I just, I really wanted to kind of call that out because I thought that was a great insight that you made there.

Loïc: 28:54

And it does impact. So I'm sorry to cut you off, Chris, but like it has like significant impact on the way you build your application, because you want to be able to switch models to switch, like, the heuristic associated with the output that you want to have. So you have to invest some time to have like a way to do this switch relatively easily, potentially do AB testing with different population to measure like the difference of perception, because again, there's not everything is like black or white, like there's like nuances of gray now in terms of perceived quality. So you need to have like more of statistical approach in terms of understanding the impact of like one model versus the others. And of course, we have internal evals and all of that to do our own testing in terms of with our golden dataset, but the reality is we have a diverse set of customers, and everyone is different, so we need to have broader perspective than just relying on our own dataset.

Daniel: 29:56

Yeah, Loic, I appreciate you getting into the technical side of things a bit and talking through some of those optimizations and how you think about them. Obviously, you're leading the technical efforts with Superhuman, and I'm wondering if you have any sort of hard lessons learned from doing AI engineering over time. We have a lot of practitioners that listen in. Any kind of general principles or lessons learned that you'd want to impart?

Loïc: 30:25

That's a good question. Maybe one thing that I've learned is to, like as a CTO, I need to discuss with the rest of my leadership team, and we talk about the the success of features and everything. And the typical way to talk about, like, quality is typically in terms of, like, number of bugs and everything. Now and and I I would say touch on it early on, but the perceived quality depends now. Like we are in a world with way more subtleties with LLMs.

Loïc: 30:53

So setting the right expectation, basically explaining that the way the feature can be built, and sometimes failing because the feedback is not great, might not be because it's not well implemented. But maybe there's, I would say more to it. Maybe there's a part of the perception, maybe there's too much latitude that is offered to the end user, maybe there's some work on the prompt side. So that's something that hit me in the beginning, where perception of the feature was like, like this is, this is terrible work. Like it's not working, people are complaining.

Loïc: 31:24

Guys, what have you done? And the work was done properly. It was like well implemented and everything, but the perceived quality of such, I would say, some of those feature can be completely different based on like the, those new aspects. So maybe like my lesson learned was to, is now to just like be very explicit when you basically launch a new feature about the risk of that perceived quality, and like the source of the mistakes being a bit less on the engineering side, and maybe a bit more on the user, and there's a lot of work to be done to control that in terms of like user education, in product education. So putting a bit more effort on like the product led growth, typical aspect of the business that will have like a tremendous impact on the success of the future.

Loïc: 32:17

So that's probably one. The second one is, and it's interesting because I see it every day, we are moving upmarket, right? We, like we have a lot of startups that are moving upmarket. So you start having like your companies that are like part of the Fortune 500, and they want to use your product, and I come from a world where moving to enterprise is pretty heavy. You need a lot of features, you need to have like a lot of compliance, you need to have like a, basically a lot of things that are not directly improving your product, but improving the confidence of those companies that you are the right partner to work on those for them.

Loïc: 32:57

There's a shift now. There's clearly a shift in those Fortune 500, and by extension, all the enterprise market, where, especially with AI, the risk associated with lesser compliance, or you're a small company, should we trust you, is completely counterbalanced by the risk of missing out, like the cost, the opportunity cost is too big, and now we see definitely push from CXOs on their security teams for the, like those AI tools and productivity tools, basically saying, hey guys, you need to make it work. You need to make it work, because it's improving so much the efficiency of the C level, and by extension of the rest of the company, you know what, we're probably ready to make the risk, to take the risk, even if it's like a series A, series B, series C company, and it's not like fully established maybe, or like maybe the, yes, they are processing our emails, which is like a core data set of our business, and we need to be like straight about it, like maybe they are more okay. Of course, we need to do the work. You need to be like, you need to prove that you're the right partner, and, but the first approach is changing, and the dynamic is changing.

Loïc: 34:13

So it's basically a bias toward let's make it work compared to two years before, where it was probably prove us that you are a reliable partner, and then we'll see if we do this POC. It's completely the reverse right now. So yeah, that's an interesting dynamic that is useful in the way to build a product right now.

Chris: 34:34

I'm curious, and, you know, we get to talk about all these really cool things happening in the AI space and how they're affecting products and services. And, you know, LLMs can do so much now. And, you know, we're kind of moving into the agentic age, you know, of AI, and that's increasing. But, you know, there's still a human being in the workflow. And, kind of what are the critical factors that the human's still bringing into the workflow as opposed to all this amazing technology that we're able to utilize on that?

Chris: 35:10

How do you see the human in the workflow going forward, given the fact that you have so much capability from technology playing all around them?

Loïc: 35:18

That's an interesting question. And I guess the answer is almost in the question. Like, it's like the human part that is hard to replicate. And so I mean, creativity, ability to define, like, to detect patterns and stuff, like so I I think that the the rise of LLMs is helping us get rid of everything that is mundane. Like, spent, like, was, I used to, I I will give you one example.

Loïc: 35:46

I I do a lot of interviews, because I hire, like, engineers, and as part of every interview process, you used to write up like a debrief for the team to consume, and so, and writing a debrief, like a thoughtful debrief, like takes time. It takes time. Like I would, I was like probably spending like between twenty and thirty minutes after each interview to basically put like the pause, the pause, and I can like question mark, like area to dive in. Now we are pretty much all using like meeting minutes that are like, you know, using the transcript, formatting that the way you want, and you just have to add your quick thoughts here and there, and on the lightning like that. So now, like from twenty to thirty minutes, this is taking me three minutes, and boom, this is uploaded in the whatever ATS, like tools like Form HR.

Loïc: 36:33

That's an example. Meeting minutes with my people, like I have one to ones. I do one to one, like meetings with my people. I want to keep track of everything that we said. I used to take notes.

Loïc: 36:45

I'm still to some extent taking some notes, but like the transcript itself is so good now that I don't have to take notes of everything. So I just like to put notes on like the the two key highlights that I want to keep somewhat private. The rest is already shared. And now it's building like a database for me, like of information on my desktop that I can query anytime to find information. So this is replacing all the mundane work that I was doing, and I can just focus on like brainpower to some extent.

Loïc: 37:17

And that's, that's definitely changing. So same for my engineers, my engineers, they've lived like, I would say padding shift to padding shift and like changing the way they build software over time. They keep increasing their velocity because of those like new tools. They have also to, to think differently. But like, it's still stupid to some extent, like all these toolings, like, it's basically an intern.

Loïc: 37:45

It's an intern. So you need to review, you need to spend the time, like you're reviewing the, what has been output, what's the output of like your new ID being cursor, being C Line, being what, whatever, like those those tools, you need to review everything, because sometimes it will make like some crazy mistakes that like a regular engineer won't do. But I think that it's saving a ton of time for our engineers that they can just focus on the core of their job, which is like understanding the user, understanding what needs to happen, and what is the smartest way to get it happen. LLMs are just a nice helper to go faster, but so far, it's about it. But it's changing every day.

Loïc: 38:27

It's changing every day.

Daniel: 38:29

Yeah, Loic, you mentioned kind of coding and vibe coding, you know, to mind. And I almost wonder, there's going to be a new reality for email with all of these AI features coming in. I know when I'm using Vibe coding tools, I have to learn a new way of working. There's different types of mental loads that I have to manage, like a lot of context switching, guiding the model in different ways. It's a different kind of mental load, a different kind of skill.

Daniel: 39:03

Do you see a similar thing developing in terms of my interaction with email and learning a different way of working through those things in good ways, but also in challenging ways to retool my mind or retrain my mind of how to work in this kind of vibe emailing way?

Loïc: 39:28

No, no, this is a good question. And then we are talking about like the user interaction and how this is evolving. And our work is to make that transition, if there's any transition, like the smoothest possible. We need to take the user where they are to bring them where they will be eventually with this vibe emailing, if that even mean a thing. I'm not sure what would be behind it, but but clearly, there's a there's a change that we are, what we are facing.

Loïc: 39:55

And interestingly, I was talking lately, but right now, startups, I would say, startup typically over index on seniority for engineers, because you need people to be able to manage the noise, manage the shit, like it's always changing. Like you need people with like a tough skin to be able to to manage that. That said, and when we see it, it's harder right now for like a new grads to get into this market. So, but they have like one asset that makes them probably different. It's the brain plasticity.

Loïc: 40:27

The new grads of this year, like for the last three years, they've seen so many different technologies coming like every six months, they had to readapt, they had to relearn. So their brain is used to this mental shift, like every six months, like, oh, damn, this is the new way to code. Oh, damn, this is the new, new way to code. Like in my days, like the biggest shift was moving from SVN to Git, that was about it, or like you have a new framework, or you have like a new language, but like it's same old, same old, like different flavor of the same thing. So I do think that the people that aren't like just born with it, can like, we were born with internet, they are born with LLMs, and with AI, and they have this brain plasticity.

Loïc: 41:14

And I think this will be like probably the challenge for like practitioners, like engineers globally, is how to adapt to that, because I'm 45, I'm not sure that my brain plasticity is still there, so I need to keep up, I need to still try new stuff and everything, and challenge myself every day, compared to even like five years ago where, like, I was just like tuning my own ways, and like making it like slightly better over time. Like this is a part of the shift. And if I don't take the wagon, I'm probably lost. And same applies for engineers. So it's definitely an interesting time.

Chris: 41:49

Definitely an interesting time. I gotta say, you hadn't, if you hadn't dated yourself intentionally, revealing your name, I was gonna say the the SVN to git switch would have done that for you. I don't think anyone out there under 30 is is going to know what SVN is anyway.

Loïc: 42:07

I'm sorry. I'm sorry if that it's kind of like my gray hair that we're talking.

Chris: 42:11

I'm just definitely a brain plasticity is on my mind as well. I'm older than you are even. I'm curious, as we wind up here, there's so much ground is getting covered right now. And you've talked about the evolution of the product and new technologies slamming into your current plans and having to adjust and stuff. If you take a step back or done for the day, and you're thinking about the future, and you're thinking on a little bit longer timeframe than kind of what we've been talking about, kind of where can email and messaging and stuff, where can it go with these technologies in a little bit longer timeframe, when you kind of get into kind of just letting your mind wonder and kind of kind of dreaming what could be.

Chris: 43:03

What are what are your thoughts around the future around that, you know, in the large? What should we be thinking about that's not necessarily going to be science fiction going forward, but, you know, day to day life given where things are are kind of generally headed?

Loïc: 43:17

No. This is a I I wish I knew. I wish I knew. But, like, if I have to to do, like, a bit of science fiction, like, clearly, I see the the communication globally, communication between people is so fragmented, so fragmented. Like with my family, I use WhatsApp.

Loïc: 43:34

At work, I use, and with my partners, and all of that, we use emails internally. We use Slack, but we also like discuss in like Google Docs, threads, like in comments, and all of that. So communication is so spread out, and so in different places that it's really hard to make sure that you have everything that belongs to the same topic into this, the sort of like unified inbox. So if I have to to guess where we would be like in, I don't know, I would say ten years, but like maybe with like AI, it would be like in six months. I would say that there's probably a need of a unified and central way to communicate for you, which is your preferred preferred interface, regardless of where this will land.

Loïc: 44:26

And doing so like in a way that brings focus. When I want to work on a specific partnership with like, in AI, with like all those like big providers and everything, I want to focus only on this, but I don't care if like the the information is in my email, is in Google Doc, is in Notion, is in in WhatsApp, or whatever. I want this to be consolidated, so that I know everything that is happening in one place. So I think there would be like a lot of work around this. The other aspect that is really interesting is the, where LLM sits, what is the entry point?

Loïc: 45:04

We see Charge pity being like one entry point, but like all the tools have like an embedded CHA GPT equivalent. So whether you use like Confluence, Notion, whether you use like Salesforce, whether you use like any kind of b to b application, have their like own specific chatbot. And then you have like actors like Glyn, for example, and and some others that try to, yeah, like unify everything. Where is this going? And so that's something that I'm really curious about.

Loïc: 45:39

Do we want to be where people work, or do you want to have like some sort of like a unified experience regardless of the the vertical people are working in? I'm curious. Like, I I have more questions than answer. What's for sure is it will evolve, and that I I do believe that Superhuman is doing that in a nice way, and people tend to love it. So building on that experience and that empathy with users, I believe we'll be like well placed for that race, basically.

Loïc: 46:10

But interesting race.

Chris: 46:12

I appreciate the insights. And thank you so much for coming on the show today, and kind of sharing, and not only where superhuman's at, but kind of, you know, how you're how you're tackling the challenges and thinking about the future. A lot of insight there. I really appreciate it.

Loïc: 46:29

Thanks, Chris. I appreciate the time with you and Daniel.

Jerod: 46:39

All right, that is our show for this week. If you haven't checked out our Changelog newsletter, head to changelog.com/news. There you'll find 29 reasons. Yes, 29 reasons why you should subscribe. I'll tell you reason number 17.

Jerod: 46:57

You might actually start looking forward to Mondays. Sounds like somebody's got a case of the Mondays. 28 more reasons are waiting for you at changelog.com/news. Thanks again to our partners at Fly.io, to Brakemaster Cylinder for the Beats, and to you for listening. That is all for now, but we'll talk to you again next time.