David Paffenholz. Juicebox.

David Paffenholz:

We got to $1.5 million in ARR with just me, my co-founder, and one engineer. The first thing we did is we posted a video on LinkedIn and we just walked through the, product workflow. So we got 8,000 LinkedIn likes on that initial post, and then we went from 0 to 30,000 free user signups in 3 days. So we went from kind of not having a product to suddenly a bunch of people trying it out. We suddenly had this demand and we knew there was message market fit. 90% of employees don't need to be AI-pilled. They need to be good consumers of AI workflows or happy to adopt them, but they're probably not going to be building the actual workflow. We looked at the most contacted person on Juicebox. Unsurprisingly, a software engineer in the Bay Area was reached out to 240 times over a 12-month period. That's like almost every day. And they responded to, I think, 4. What are those 4 companies doing differently than the other 236? We're 5 days a week in person. We have no exceptions. Ishan and I are in the office 7 days a week. That said, we are also explicitly not 996 or, expecting work on weekends. Like, people work intensely during the week. People should take the weekend to do what they like to do and then show up as their best selves on Mondays. I think I'd— I would tell the earlier version of myself to be even more comfortable in being wrong and telling others that I was wrong.

Nakul Mandan:

Ask any AI founder about their scarcest resource. It's not compute and it's definitely not capital anymore. It's people. The talent war is becoming the defining bottleneck of the AI era. My guest today has a view of that war almost nobody else has. David Paffenholtz is the co-founder and CEO of Juicebox. 5,000 companies from Cursor to Ramp to the Fortune 100 run their recruiting through his product. It has reached more than 3 million candidates. And when the best companies in tech fight over the same people, David sees the whole board. But also, David's 25. He started this company straight out of college, got to $10 million of ARR with just 16 people, and today he's backed by Sequoia and DST at an $850 million valuation. How is he building one of the most AI-native companies out there? How is he growing himself as a CEO along the way? And what does he see across 5,000 companies fighting for talent? That's the conversation we're having today. Knuckle Up! David, welcome to the show.

David Paffenholz:

Thanks for having me.

Nakul Mandan:

AI for recruiting was one of the most obvious opportunities right at the beginning of the AI wave, and a lot of smart teams saw it, not just you, right? But Juicebox broke out. So what did you understand about the market that others may have missed?

David Paffenholz:

I think this was true then, and it's still true today, which is if you look at what a recruiter is actually looking for, or like the biggest pain point that they have in their process, almost everything comes down to search. Are they finding the right talent and are they bringing net new talent into their funnel? And it sounds like a simple realization, but I think it's actually just by like an order of magnitude, the most important one. And, I'm glad we had that very early on because it impacts all downstream product development from there. And so the very first version of the Juicebox product was purely a search interface. you could type a query and start seeing profile results, and then you couldn't do anything beyond there. Like, you could click a profile's LinkedIn and then open it from there, or like download the name and then, do a different workflow somewhere else. And so in terms of like traditional SaaS, or like how one used to think about building a product, is like you'd map the workflow and then you'd try to recreate that workflow. In a way, this is the opposite, because the first thing we did was just focus on the search and be obsessive about making the search as good as possible for the first 2 years of the company. So I think that ended up being the most important thing we did early on and the right insight because it helped us then go to market with the most important piece of we're helping you find better candidates. And today the platform has grown a lot. There's a lot more workflows built into it. but it's still the number one thing we focus on, and over half of our engineering capacity is always dedicated to search.

Nakul Mandan:

Even like, let's double-click on that, right? Like LinkedIn has a massive recruiter product, right? I mean, that is— it's probably— I don't know if Sales Navigator is bigger than Recruiter, but top 2 product, right? And it already— it focuses exclusively on search. So what did, back in the day, people— GPT before Juicebox, which was your earlier name, got right about search?

David Paffenholz:

So I think the, thesis we bet the company on was basically that LLMs will change the way that you can do search. And so initial search, or like if you look at traditional search algorithms, be that BM25 using different types of keywords to rank results, there's a lot of companies that have gotten really good at those. And, frankly, I think LinkedIn is probably one of the best teams at, building a really good search given the technology that was available at the time. And so our bet was that LLMs are actually gonna change the way you can do search, through a few things. One, you can predict what the person is searching for, even if they aren't explicitly searching for that thing. And so to take like a really simple example, if you are searching for a software engineer, I can make reasonable assumptions about the type of software engineer you're searching for. Say the fact that they're based in San Francisco, you probably want them to have some kind of startup or venture-backed experience. You probably have a preferred tech stack. We can add all of those things into the search and then we can start reviewing profiles files, using an LLM against that search. And so in a way, all of the best practices that used to exist in search or the different machine learning techniques, the, way people went about optimizing it completely went out the window to a large extent because today you can do search in a fundamentally different way. That's a little bit more akin to how a human thinks about assessing a profile. And perhaps the whole dimension of search goes beyond search towards search and assessment. And that's what gets you the right results. and so I think that was like the, biggest thing that changed and why, from a pure like ML-based search perspective, a company like LinkedIn or other search platforms are probably way more advanced than what we do. We just think that the future is actually not in that. The future is in an LLM-based search.

Nakul Mandan:

Maybe just to double-click on this, right? So LLMs making search more human-like makes sense. I would imagine your other competitors who were starting up around the time that you were starting up would have also been on that insight. Was it more that you guys tripled down, 10x down on this one insight and build the entire product around this versus they probably overthought all other aspects of AI recruiter? Like, what was it that still differentiated you against all the new AI startups that were coming up in the LLM era for recruiting?

David Paffenholz:

I do think the majority of companies at the time took the approach of we will be your AI recruiter, which means— and a recruiter does a variety of things that go beyond the actual search, and we will try to solve all of those things for you. And I, I see the vision on that, but in practice, I think it's extremely hard to solve all of those things at once. and so I think focusing on just the one thing that we could actually do exceptionally well, which was search, was the right thing. And it also, like, search is the kind of thing where once you get into it, you can always go deeper. And so there's over 300 filterable fields on a profile level that we have in our search index. We don't expose all of those to the user, but that's what lets our LLM-based search be really good because we're so obsessed with making that really, powerful. And I I think that was like a really important foundation to set, before even starting to look at the other areas of the product. But it's hard because if you ask a user or you're in a conversation with a user, at least half the time they're gonna ask you about downstream workflow, or other things that they're doing or things that they're doing after other problems that they face. And, I think that was the really hard decision at the time is like, can we just focus on search and like, let's just make that really good even if our users also want other things.

Nakul Mandan:

Kindly, I mean, it's an interesting one because, you and I met when you were doing PeopleGPT. I, I actually recall even for me, the biggest question— of course, a big mistake at my end, did not invest in that seed round— was that, hey, is, is just a search product, search-focused product enough? And, you know, and even from the VC perspective, I think a lot of VCs felt like, no, the full AI recruiter is where the future is versus you just double down there.

David Paffenholz:

Yeah, and, I think long term that's probably right, like, right, like long term just search is not enough, but I think in the beginning search is such a hard problem to solve to In order to actually solve it, you have to be focused on it for a surprisingly long amount of time.

Nakul Mandan:

The product also then went from a search engine to a Copilot to now agents. Can you talk about each of those leaps as to what gave you comfort that we are ready— our users ready for the next step, not just the technology?

David Paffenholz:

I think that's actually still one of the hardest things today. And especially as we work with like larger and larger enterprises, cuz there's so many more stakeholders at play. And so today the product has kind of two main interfaces. I think what you described as like the search and Copilot, we've kind of fused into one. So every search is now automatically done with a Copilot effectively, where every profile will be assessed. we're able to serve those really quickly. So it's an experience that feels like you're just searching. and then we have our agent product, which is slower because it does a ton of reasoning, but it gets you to a better outcome in the end. And so we've like now started to use similar UI elements to what like the foundational model providers do, where if you're in like ChatGPT, you can toggle between chat and co-work. same thing on Claude, et cetera. I don't know if that's a form factor that will last, but So far it seems to be working because it does serve both user types of like the user that just wants results quickly and the user that wants to like, you know, kick off the passive agent or someone who wants to do both in parallel, which is maybe the most sophisticated use case.

Nakul Mandan:

How do you kind of decide as to, where does the human stay in the loop? Because that's a tricky thing. Users want autonomous agents, but it's also scary. Autonomous agents can do things or make bad decisions. And so do you let just the user decide as to how much is it completely a sliding scale or— in the co-work, it's a binary thing. Hey, more human in the loop in the search and co-work, sorry, Copilot product versus the agents is like, how do you decide this?

David Paffenholz:

Yeah, so with our agent product, we've decided to go very like, let's say, agent-pilled in the sense that, it will just continue making decisions for you, until any, what we call like a final decision is made. And the final decision is basically, do you want to reach out to the candidate or not? And by default, that still requires a human approval of like, accept, though you can also switch that to automatic acceptance. Now that's only really possible because we're at the top of the funnel of the recruiting journey. We're helping you find net new people that you're inviting into the pipeline. And so it's not like you're deciding on applicants in, the process. Rather, you're just bringing more people into your funnel. and so because of that, we're like in a fortunate space where I think being a bit more proactive doesn't have a huge downside. In fact, it's just creating more opportunities for people in the end.

Nakul Mandan:

Actually, can you double-click on what is the decision an agent— what is an example of a decision an agent will make even just within the top of the funnel?

David Paffenholz:

Yeah, so let's say you're looking for account executive in SF, and you just only say, hey, I'm looking for an account executive in SF. There's so much missing context that in theory would have to ask the user. So what the initial version of the agent would do, it would be like, okay, to help me succeed in this, please answer these 3 questions. Do you want a mid-market or an SMB account executive? Do you want someone with 2 years or 6 years experience? Now we just look at the recent hires you've made. We look at your job description, we look at your previous searches and we just assume those things for you. And so you can still go and correct those, but the agent will actually show you in its reasoning, the user didn't tell me this, but I'm gonna make these reasonable assumptions. Here's the outcomes I get from it. And if the user objects or has other feedback, they can still change it, but we'll just go and assume from there.

Nakul Mandan:

So let's talk about the elephant in the room, which is LinkedIn, right? LinkedIn is presumably one of the biggest platforms, maybe GitHub and others are also there, but around which— they own the graph on top of which you guys are built. So why, wouldn't they want to build this? Why can't they build this? Why does a separate company need to exist in this space?

David Paffenholz:

I think first off, LinkedIn is an incredibly successful business. they're like talent solutions based on like public data seems to be around $8 billion or so in revenue. And then there was a report that I think they released, a few months ago where they said that their AI products surpassed $400 or $500 million in ARR. And so it's actually a pretty massive AI business that they've built. It just takes a pretty different format from what we think today of like agent products. And so LinkedIn's primarily seat-based, the kind of KPIs that all of the recruiters sign into the platform, and that's one of the ways you measure engagement is like, how many LinkedIn searches did I do? How many profiles did we view on LinkedIn? I think that's where we have the biggest difference of opinion. So we think agents will actually be able to do a lot of those things, and we think naturally that will mean there will be less seat-based usage, and there might be less recruiters that have to go do sourcing because so much of the sourcing can be done autonomously by the agent. and so I think that's where like maybe LinkedIn faces a classic innovator's dilemma of, you know, they have such a successful business that is based on the seat-based model. It's probably going to continue to be successful for 5, 10 years, especially on the enterprise side, because enterprises are slow to change. And we can go all in on the agent side and build a long-term business that is not just tied to seat-based usage.

Nakul Mandan:

A couple of years ago, maybe actually just a year ago, things in AI move so fast, it feels too long ago, but, there was talk about OpenAI building their own AI recruiter. I don't know if you also heard this. We had heard this, but Like, talk about the labs. Like, why would the labs not build this these days? That's not just a question for you, it's for every application layer, horizontal application layer product.

David Paffenholz:

I think the short answer is the one— in theory, the labs can build anything if that's like their main focus, right? They have access to the best talent, they have infinite capital effectively. And so if they want to, they can go and build. I guess it's a question of where do they focus and how widely applicable is that to their customer base? And so. If you think of what you need to be able to build a good recruiting product, the first thing you have to build is your profile index. And so a lot of our engineering work is focused on building and maintaining and enriching our profile index. It's roughly 800 million profile records, all of which are enriched with full company data. So for your profile, for example, we'll do research on what Audacious does, what type of investments you do, what you said publicly, what you said on podcasts, and we'll append that all into your profile. and that's a pretty intensive, both in terms of engineering work, but also just focus, to, build out that data index. That's what allows us to run good search and an agent on top of that index. And so even if you have a really good harness that can, genetically go and, do things, unless you have an underlying data index that you're searching on, I think it gets really hard to deliver a strong sourcing product. And so then the other dimension that exists is like browser automation. So, okay, can you go build a brow— browser automation product? That would effectively mean you have to browser automate on LinkedIn, which LinkedIn is pretty good at preventing. And so I think in is it actually gets really hard to build a good recruiting product unless your main focus is building a recruiting product.

Nakul Mandan:

What about the competitors? So the direct competitors with AI, CodeGen has become fast, shipping velocity is insane. So, and again, this is a question probably true for every single category leader in application layer, especially how do you stay ahead?

David Paffenholz:

Yeah, so I think short term it's all about good execution. are we making the right product bets? Are we shipping them fast? how fast are we scaling go-to-market? Are we going upmarket faster than competitors? and so I think that's like the, in our control and execution question. I spend a lot of my time thinking about that because I think that's one of the highest leverage things I can do for the business is measure and benchmark how well we're executing and then pushing us to execute better in the places where we're not. however, that's only a, a short-term thing. Like that's like the sprint part of the race. And so, far we have an early lead in that, but as you said, it's competitive and there's gonna be more and more companies and there's gonna be more and more well-funded companies. And so then the second part is like, okay, what is durable? What do we, build in a self-reinforcing way? And there, there's kind of two interesting types of data that we learn and that we collect. the first is candidate intent and interest data. And so today in the platform, there's millions of emails being sent out to candidates, and increasing pretty rapidly. and that also means we track tens of thousands of interested candidate responses at any given point in time. And so that's a pretty massive data set that we can use to start predicting when certain people are most likely to start looking for— at a new job. And that's the type of data that you can only really do once you have a certain skill. I'd say right now, like, probably LinkedIn and Indeed are the only players that have, a sufficient amount of that data to make really good predictions. In some markets, I think we're getting pretty close. And so we can start actually making very good predictions of who might be in market at certain times. The second part is the company internal data. And so we call it a talent bar is, unique to a specific business. How does that business recruit differently from another business? What does the company think they do differently? And then what does the data actually show? And so if you look at the recent hires of a company, you can usually make a pretty good prediction about the type of persona that someone's recruiting for. We show that to the user in the product so we can say, hey, this is the type of persona that a user is looking for. And in some cases, the company might want to actually change that, right? Like they might say, hey, you know, we have this like hidden hiring pattern. Is that something that we want to be doing or not? Either way, we get pretty deeply ingrained in that hiring process. And so then we start becoming a part of the decision-making process of these are the personas I want to recruit in the company, but we can only do that if we're deeply embedded and have a lot of context on the business. And so I think that part is also kind of a lot more sticky in the sense that it becomes more like hiring someone internally than it does, buying another SaaS product.

Nakul Mandan:

Do you foresee a world where there will be a candidate-facing AI from Juicebox, for candidates to search better, or—

David Paffenholz:

It's a good question. One we've thought a lot about. I can't make any promises.

Nakul Mandan:

Okay, fair enough. so let's shift gears towards the GTM side of it. Maybe I'd love to talk about how did the early go-to-market look like? And then how is, it scaled? So maybe start with the early go-to-market engine. What worked? How did you break out above the noise of today's like launches every day, every week on Twitter and Product Hunt?

David Paffenholz:

To take back to the very beginning of the product and our first million dollars in revenue, to contextualize, we got to $1.5 million in ARR with just me, my co-founder, and one engineer. and so it was the three of us at the time. The roughly 70% of that ARR was PLG. And so we did a lot of focus on social launches, viral launches, how can we get on every recruiter's mind even though the product barely exists? and then two, for the enterprises or the teams that have the biggest pain point that we serve, and at the time that involved like the AI labs or different AI applications. There are companies that wanted to go really deep on search. can we monetize them on larger agreements? And so that was like roughly the thinking we had in the beginning, which worked out for us. And so the first thing we did is we posted a, a link, a video on LinkedIn. it was a 60-second video, pretty short. it was introducing PeopleGPT and we just walked through the, product workflow. it was recorded in like a Loom-style video, so it was pretty approachable. We actually, had like a lot of heavy editing going into it though. So it took like 2 days for us to put it together, even though it intentionally looks pretty basic. and that really worked. So we got 8,000 LinkedIn likes on that initial post, which in LinkedIn terms is pretty high engagement. and then we went from 0 to 30,000 free user signups in 3 days. And so we went from kind of not having a product to suddenly a bunch of people trying it out. Frankly, a lot of it breaking at the time, and there was a bunch of issues, but we suddenly had this demand and we knew there was message market fit. so we knew people wanted what we were offering and started paying, small amounts of revenue for it. And so that was kind of the very early days, our first $200,000-$300,000. And then to, get to that first $1.5 million, it was, it was primarily more self-serve, more people talking about it word of mouth, plus, an initial set of larger business plan deals, which I was personally doing at the time.

Nakul Mandan:

It's interesting you call it message market fit and not product market fit. Can you differentiate between the two?

David Paffenholz:

Yeah, so I think message market fit is when you have people signing up or like, starting to use the product, and then product market fit is when they retain. And so in the very beginning, we had only message market fit, not product market fit. It was just the search product. You could sign up, you could do a search, you could see an initial list of profiles. but it was very basic. We could— we only showed you 50 profiles a month and then it was done. Like you had to wait until the next month to get your next 50 profiles. And so it was a, it was a hard to use product in that sense. And I, I think users signed up to it cuz they could see the promise for it, or like they wanted something that did this for them. But it wasn't really working yet. And so that took us like 6 months to get it to work. and I think that's when we then started to see signs of product market fit, where we started to see churn decrease, we started to see expansions naturally in the customer base, and then we started to get those like larger customer contracts as well, which aren't a sign of PMF in and of themselves, but at least directionally better than, only self-serve users.

Nakul Mandan:

So many founders face this. They have a good launch, users come, nothing sticks. What was this, that 6 months like? What were you fixing? How were you getting the feedback on what to fix?

David Paffenholz:

In some ways it was actually a pretty brutal time because we launched, we were like, you know, it took us a while to get to launch. we were so happy. We had these like 3 moments of excitement where all these people were signing up and paying, and then almost immediately it kind of flipped of like, okay, now the people who paid are complaining. And so the upside of that though is that it made it abundantly clear what we had to do better. And we got an abundant amount of feedback. A lot of which was tied to search. And so we knew, okay, the search is already promising, but it has to get better. And then two, we obviously need to fix this 50 profile limit. You should be able to see as many profiles as you need to. And so that's where we then started investing in our data index and actually building that out in order to serve those, more than 50 profiles. And we knew we had to do it. We— it took us 6 months to successfully execute on it. And we had a lot of step-by-step improvements along the way. And then once we kind of did that and the search started getting pretty good, only then did we start looking at other parts of the workflow. And so that's then where we started doing things like managing your contacts, reaching out to them, and more.

Nakul Mandan:

Okay, so that got you to the first 1.5 to how much, in ARR, like that first phase?

David Paffenholz:

I'd say phase 1, 2024, was us going from like basically zero to $1.5 million, and it was just the 3 of us. At the end of 2024, we raised a seed round, and so we— it was our first like real round of funding. And then 2025, we knew we had to grow the team and scale the business. And so last year we went from 3 people on the team to 25 people. We grew ARR over 10x and we kind of started to see that scale happening, both in terms of customer base, but also in terms of team.

Nakul Mandan:

Did the same kind of PLG, product announcements, PLG that would then lead to upsell work through that phase? Or did, you introduce new, go-to-market motions?

David Paffenholz:

Yeah, I think we, did a good job of pretty quickly leaning into the sales-led side as well. So if I think of us at like, okay, how do we get to a billion in ARR, it's pretty inevitable to me that at the billion in ARR, maybe $200 million, maybe $300 million can come from a self-serve customer base that has like, you know, a $2,000 or $3,000 ACV. Because if you think of the number of people that need to sign up to do that, there's only so many recruiters in the US. And so if you, if you have 100,000 people paying you $2,000 a year, then you have $200 million in ARR. And it's hard to conceivably go orders of magnitude beyond that. Now that means that we have to be able to monetize on the business or the enterprise side to a greater extent. And so that's where that hypothesis came from, and I think that was so far true, that we have to invest in the sales-led side as well. I think we really started doing that with, over the course of the last year of like hiring AEs, and then, you know, we today have a 50-person go-to-market team, and so we've continued to scale that, quite aggressively as well.

Nakul Mandan:

Do you feel like the— even now, the early adopt— it's— you're still in the early adopter phase where it's a lot of AI-native companies, people who want to do AI for recruiting, or have you broken into some of the traditional industries, traditional companies yet?

David Paffenholz:

Yeah, so only 35% of our revenue is from venture-backed, and our top 10 largest customers are financial services, defense, a recruiting agency, and then a big tech company. Those are like our top 4.

Nakul Mandan:

Are these coming inbound? Because actually that's a fascinating, thing that all of us are grappling with is the non-AI, non-Silicon Valley company actually proactively embracing AI, or is that still a push? Like when, a big financial services company adopt you, it's a sales-driven motion, or there are two recruiters who sign up and then you go from there?

David Paffenholz:

Yeah, in almost all of those cases, it's a very sales-driven process. Now the— I'd say the, most interesting part of PLG there is that, because recruiting has like a fairly high turnover rate in the sense that people join new companies somewhat quickly, for many of our largest customers, by the time they started looking at Juicebox, we already had advocates in the business who had used Juicebox in a previous role or on like a self-serve freelance basis, though not in the formal context of the company. And so the actual company's evaluation is very much a sales-led process. We have an enterprise AE working with their VP sales. We try to get a bunch of other stakeholders involved. We do a trial and more. That said, I think the, enthusiasm and the, even the appetite to consider that comes from the fact that there's advocates in the business that already used it in the self-serve.

Nakul Mandan:

But are these buyers hungry for AI or you're educating them on what the power of AI is now?

David Paffenholz:

It depends. I, I think for, the people that have that are customers today, there is at least a natural inclination or interest in those businesses to implement more AI-forward workflows. and so I think that's kind of a given for those customers. That said, I'd say that's also probably a pretty large share of enterprises today that at least have an appetite or interest in exploring it. And so then I think the, thing that gets harder is like one level deeper. Okay, to what extent are they like interested in it from, you know, being able to say they have an AI workflow or they deploy something to a small team versus to what extent are they truly evaluating it for like a company-wide deployment? And so that's really then on us to assess and help them discover. And then see, if it makes sense for, us to pursue that together.

Nakul Mandan:

Let me ask you the other opposite question, which is if an— if a large company today which is hiring a lot is not adopting Juicebox or any of your AI competitors, what's the resistance at their end?

David Paffenholz:

Adopting Juicebox successfully, requires a few pre-existing beliefs in the company. First, the belief is that the talent bar at the business really matters to the business, so that having better talent in the company will lead to better business outcomes. The second is that to get that talent, they can go outbound and find that talent for their teams. Now, sounds obvious in like Silicon Valley, just given the, amount of recruiting that goes on here. But for a lot of large enterprises, the, far majority of roles are purely filled through inbound. And in many enterprise cases, inbound can actually also be really high quality. And so the, necessity of going outbound is not always, clear to every stakeholder. And that's on us to have to, prove that of why you're able to get a better, a better talent into your business. And so if those two things are, clear, like the company knows they want to increase, the talent bar in the business and they know they can go and proactively do so rather than just be reactive. Those are good preconditions to then adopt an AI solution. And then from there it's just like a change management.

Nakul Mandan:

So if somebody's like, hey, we have 40,000 resumes in our database that are already inbound, we don't really need this. Like we just need to sort through it. They, might not care for you.

David Paffenholz:

Yeah. And, there's also different business contexts. You know, there's, businesses that just require massive volumes of new people in the company, that don't have prior training where they have extensive training programs in-house and they know they can build that talent pipeline within the company. And so maybe in those cases it's that's important for business to go outbound and find that talent, because they've built the in-house resources to do so. And so I think a lot of that depends on the context of the business.

Nakul Mandan:

Got it. Now, if Juicebox is massively successful, would internal recruiting teams exist, or will business users directly— hiring managers directly use Juicebox as their AI recruiter?

David Paffenholz:

Today, the companies that have the most hiring managers using Juicebox are also some of the companies that I think outside in many people would consider the best at recruiting. So, companies that a lot of people would be very interested in, they're also companies that actually have pretty large recruiting teams. And so I think that speaks to the fact that in businesses where everyone is excited about recruiting and they're really passionate about getting in the best talent, that shows an investment in recruitment overall. That also means they have pretty large recruiting teams because those recruiters provide an exceptional experience to the candidate. They build a real relationship with them. They check in with them after inter— every interview. By the time they're closing, they, might have built a real relationship between the recruiter and the candidate. That requires a pretty high recruiter headcount.

Nakul Mandan:

Account.

David Paffenholz:

And so I think for those arc type of businesses, which I think will be more and more where hiring is a priority and they know they want to attract the best talent, they will need to continue to have relatively large recruiting teams. But the work of those recruiting teams is going to shift much more towards the candidate relationship than the administrative parts of the job.

Nakul Mandan:

I, I understand that these, are the companies that are most recruiting savvy, hence they're making it as a real business function. Recruiting is a business function, not a supporting function for these companies. So it makes sense. But as you go more in an agent-first world, more away from search and autonomous agents, like, what does the recruiting org look like in the future? Is it just one head of recruiting with several juicebox agents? Yeah. What, does a human recruiter actually bring to the table?

David Paffenholz:

I think the same things hold true in that the recruiter is the one who has to build that relationship with the candidate. But then too, and, I think this will probably increase in importance over time, is The recruiter should also be able to be consultative to the hiring manager of what type of person they can actually help fill in the role. Usually what the hiring manager has a good understanding of is like, this is the business problem or process problem we're facing, and they have headcount allocation to go solve that problem. And so that's like kind of the, preexisting set. Now, in some cases, they've already hired for the same role multiple times. They know exactly what they're looking for. Maybe in that case, they can run it somewhat independently, but in many cases, they don't know the exact role that they're hiring for. They don't know what the process for it should be. And so the recruiter in that way has two responsibilities. One, building that relationship and educating the hiring manager, and then two, building that relationship with the candidate and bringing them through the process and then ultimately closing them.

Nakul Mandan:

Do you think there's a future where the best-in-class candidates talk to an AI interviewer for the first screen?

David Paffenholz:

I think it's unlikely, unless the intent is to purely provide information to the candidate and the candidate just prefers to consume that information in a live call format. I often think about the AI interviewing because I think it's a really interesting application of AI in recruitment. Like, it seems clear that there should be something in AI interviewing, but it's also in a way like an antithesis to what Juicebox does of helping you go outbound and hunting for the right talent and doing everything to be the most attractive employer for that talent. And if that's the goal, do we immediately want to bring them into an AI interview? Now, there's probably a happy middle ground where, you know, we invested in creating like a candidate-facing website that just explains a bunch of things about the company, videos, et cetera. Maybe that information could be consumed in a different way. But should that AI interviewer actually be assessing the candidate? I'm not sure about that.

Nakul Mandan:

As you said, this is like a multiple billion dollars of category, right? Like LinkedIn alone does $8 billion. So, and Juicebox is one of the leading— like, for Juicebox to go from where it is today to the multiple billions of dollars of revenue, what does the product surface area look like in 5 years? Like, fast forward 5 years, what is the full scope of Juicebox?

David Paffenholz:

So I think that's one of the things that makes Juicebox such a fun business to build is that our product scope can— will continue to be extremely narrow. And the best examples of that are actually LinkedIn Talent Solutions and Indeed, where in both cases, you know, they would be the best position to go build an applicant tracking system, to go build scheduling software, all of the things that happen downstream in the recruiting funnel, but they haven't because those categories exist. They're perhaps in some ways, somewhat commoditized of like one ATS is not that different from another ATS, and that's then reflected in the value that they capture where like the ATS scheduling, note-taking, assessment, all of those categories, total revenue is, not that large when compared to like a LinkedIn or an Indeed. And so in, Juicebox's case, we want to keep that same focus. We want to stay on the top of funnel. Are we helping deliver better net new candidates to the company? and then we want to basically hand off that process at that point and, end our involvement there. Now there's some things where we can actually make better sourcing decisions later on, like are we aware why certain candidates make it through? What feedback do we learn from them and how can we use that information? That's important. but it doesn't mean that we'll be building an ATS. And so we'll stay narrow on sourcing. We want to help you find the best candidates and, do just that.

Nakul Mandan:

So let's shift towards, how Juicebox is run and built and how, you've been, building it, as a company. If you were to describe the culture at Juicebox, how would you describe it?

David Paffenholz:

Yeah, we have two values, that, Ishan and I shaped in the first two years of the business when it wasn't really working. Those are decision-making speed and intellectual honesty. And those two values go hand in hand. decision-making speed, how quickly do we make a decision on something? And that explicitly means that we're not like overthinking decisions that, don't have a purpose, to be overthought. I think oftentimes the reason for people having a meeting is because they want, permission from others to make a decision. We try to avoid that. And so we wanna have maximum decision-making speed. Everyone is empowered to make those decisions and it's totally it's totally okay if someone gets them wrong. Now, that model only works if you then also have a form of correction afterwards, which we call intellectual honesty. And so that means we're open to talking about cases where we made the wrong decision, we took the wrong bet, but then we quickly change course and take the right bet afterwards. And so even if our, decisions are, let's say, majority wrong, because we're making them quickly, we then get data and we can course correct quickly. I think the end outcome is still better than if we take a b— a bunch of time to decide beforehand. Now that results in a pretty intense culture, where there's a lot of change and a lot of, things happening quickly. I think to counterbalance that, we try to put a big emphasis on winning and celebrating wins. And so we try to do celebrations every time we have a big new product launch. We get like a matcha cart in the office and, we get to celebrate that. We have a physically nice office space. We try to be generous with, all things around to help kind of make it a positive culture that counterbalances some of the intensity as well.

Nakul Mandan:

Is it all in person?

David Paffenholz:

Yes, fully in person.

Nakul Mandan:

And do you guys have some mandate around all 5 days in, the office or like, yeah, how do you manage some of the workplace culture, cultural elements? 996 is a big phrase thrown around, like where are you on all of that?

David Paffenholz:

Yeah, so we're 5 days a week in person. We have no exceptions, for remote work or any other reasons. That was hard because it means we have to say no to candidates that we'd be very excited to have on board. That said, we're also explicitly not 996 or expecting work on weekends. Ishan and I are in the office 7 days a week. Ishan's my co-founder. That's primarily because we enjoy the time together and enjoy the work that we're doing. sometimes others come in as well, but it's like fairly unusual and never an expectation. and I think that's been a good balance. Like, people work intensely during the week, people should take the weekend to do what they like to do, and then show up as their best selves on Mondays.

Nakul Mandan:

Anything that has changed in the culture as you've grown from those early days where those first two values were kind of formed to now you're 90 people?

David Paffenholz:

One of the biggest learnings for me over the past, like, couple months is that, I used to have an approach of we just go ship product and people will figure it out and the sales team will figure it out and that's how we'll talk about it to our customers. that started to break. We started to have a bunch of, I mean, problems is a strong word, but like cases where, people didn't know what was happening in the product. We had like AEs who were seeing new things on demos that they weren't familiar with. and all of those things then start like, also not being what we want to happen. And so I think that's been one of the things we've changed as we've tried to put a bigger focus on internal enablement, actually providing resources, doing like a internal training session before the feature goes live. And I was initially quite against all of that. Like that to me seemed very big company coded. It all felt very slow, like, oh, why are we doing this all hands meeting to go through a feature? Now that we do it, I think it's actually immensely valuable. People feel ready, they know the value of the feature, they can talk about it, and they're also like more of a product expert than if they were just self-discovering everything. so I think that's changed and probably for the better.

Nakul Mandan:

What's the hardest thing about working at Juicebox?

David Paffenholz:

It's still very focused on your own agency. And what that means is there is— like, we've now started to build a management layer, but like until recently that didn't exist. There is still zero engineering managers in the business. So everyone in the EPD org reports directly to Ishaan, my co-founder. On the sales org, we've now started to add more structure, but it's still very much go and figure out your own thing. And that means you can have an exponential upside. You can be closing a ton of deals, you can be, shipping really high high-impact product, but it also requires the right type of person who kind of wants that, and that's not everyone. And, now as the business will scale, we'll slowly have more structure, but hopefully not too much.

Nakul Mandan:

And you mentioned, you know, I think when we were talking before this, that you're at 90 people, you've never lost an employee, never, right?

David Paffenholz:

In the last 2 years.

Nakul Mandan:

So like, what has created that? Like, is there— is, it some kind of celebration of winning? Like, what's the— this What's the ingredient that's leading to that kind of retention, which is kind of crazy in today's world where people move around, there's lots of seed funding thrown, there's money being thrown from the big labs. How are you retaining these people?

David Paffenholz:

There's this concept of being like the place where someone feels that they can do their life's work. And I think it's actually kind of a hard question to answer, like what are the preconditions that need to exist to be able to do your life's work somewhere? And I, I think one of those preconditions is you need to have independence. So independence usually comes with the ability to make an impact. You can actually make decisions yourself. I think we try to maximize the amount of independence that everyone has in the sense that they're able to go and do things and try new things. Second, it has to come with some form of reward. You have to feel that the, impact that you're making is seen, be that impact on the customer, be that the commission payment that you can earn, be that, the career development you can have and you feeling like you're getting one step closer to where you want to be. and so I think maximizing that impact and then also making it visible to everyone. And then the third thing is, you know, the, company is growing fast. we try to make sure everyone has a, a sizable equity stake in the business so that they also have strong financial upside to be at the business. And I think that's one of the coolest things about Silicon Valley. That's also a very different— like, I grew up in Germany, like, the idea of equity ownership is like very different. And so, I think making sure everyone is able to lean into that and get that upside too.

Nakul Mandan:

Let's talk about the weekly operating cadence, like what meetings matter in the week, how are you spending your time, and how is Ishan spending, like, at a macro level How does a week at Juicebox look like from an operating and leadership cadence perspective?

David Paffenholz:

Yeah, so team-wide, we have, extremely minimized meetings. We have a biweekly all-hands every Tuesday, 1 hour, and then we have a monthly product meeting. Those are the only mandatory meetings in the business, for the whole company. And everything else is essentially optional. From a leadership perspective, we've recently introduced a biweekly leadership team meeting. again, something that I actually resisted for a long time. I said no meetings, nothing recurring. We, should be executing on work, but I also had to kind of go through that learning journey of I can't just be executing 24/7 because it means that we don't have like a process in place to get something done. And so now we've started to introduce that leadership team meeting. transparently, I don't think I'm doing a great job of running that yet, and it's something I have to learn. And so in the beginning, we kind of— I, I put like what I thought were the discussion items or what I thought we were getting wrong as a leadership team. And so it became like a pretty intense meeting in the sense that we were just talking about the things that I thought were going wrong, which wasn't the right setup. And so then now I'm like trying to reflect of like, I'm creating a pre-read or like an agenda for it, or like also looking at positive things. And so I haven't quite found it out yet, but, that's the part that I'm still working on. I, I guess when, we used to have problems, the business was just so small that like the problem didn't really matter. And I think now as the business has scaled, there's like more problems as well. And so it's like, okay, how do you talk about those problems in a productive or constructive way?

Nakul Mandan:

So it's interesting. You, said you resisted the leadership meet— a standing leadership meeting, but you know, in a fast-growing business, especially in, AI application layer category, the market is changing dynamically, the tech capabilities are changing dynamically, your product-market fit might have a lot of moving parts. Like, why resist that? Because you also have to dissipate what is going on in your mind through the company and especially the leadership team on product and go-to-market. Like, how was that getting communicated before this leadership team meeting was in place?

David Paffenholz:

So I think I initially had, which I've now learned was probably not the best view of, I was like, okay, you know, we hire different leaders. The reason we hire different leaders is because they can do a phenomenal job of executing and they're part of the business. And so our sales leader can help us close a lot of new business. Our CS leader can make sure we're succeeding with our existing customers, et cetera. And I almost viewed it as like somewhat isolated, you know, like we go solve this thing, we go solve this thing. What is the remaining alignment we need as a company? Like everyone should just be doing their own thing well. And like, yes, we have to collaborate and know what's going on, but like, I didn't feel like there was any kind of missing leadership or discussions. I then realized that like, one, I think people can execute in their job better if they have a feeling of what's going on overall. And then two, they are more bought into it because they understand the bigger picture behind it too. And so then I think I overcorrected where like, I would bring like basically everything that was going through my mind or like even things that were like, I was basically just thinking about for a day or something and talk about those pretty openly in the leadership team meetings, which then I think like led to the, opposite outcome of like suddenly we were spending so much time discussing so many random things that didn't really matter either. And so now I've like, again, course corrected to what I think is a good medium of like, you know, here's certain brainstorm items that we want to discuss as a team. Here's something that's on my mind and here's a win that we can talk through from the last week. or at least that's the, rough format that I'm testing now.

Nakul Mandan:

Let's also— didn't talk about how you guys are dogfooding AI across the company. you, probably are one of the most AI by natively run companies. And everybody talks about CodeGen, so I want to talk less about that. Maybe talk— start with recruiting. How does your recruiting function work, using JuiceBox with other AI tools? Yeah, let's start with recruiting and then we'll go to sales and marketing.

David Paffenholz:

We have a 5-person recruiting team. We— which I initially, I wasn't sure, like, oh, how big should our recruiting team be? Like, how many recruiters will we need given our product, etc.?

Nakul Mandan:

So 90 people, 90 total employees, 5 recruiters.

David Paffenholz:

That's right. and the recruiting team has actually been, I think one of the, highest impact investments we've made in the business, both in terms of using the product but also in terms of, and giving product feedback, but also in terms of the hires that we're making. and so our recruiters are almost exclusively now really focused on that model of the candidate relationship. We do a call with every candidate after every single interview. We update our notes, we figure out where, are we in the process. And then we have an exceptionally high close rate, for candidates that get to the final stages. And I think that's because of the way we've, set up our recruiting team, to be able to do that while also hiring very quickly. In terms of AI use, of course, we're heavy users of Juicebox. the majority of our hires are outbound sourced through the Juicebox platform. Beyond that, I think we've done like a, a decent job of building other Slack automations of looping in hiring managers, making sure there's like automated channels existing for every role we're working on, projecting what number of hires we'll be able to make. So tr— treating it a little bit like a sales motion, the forecasting that goes with it.

Nakul Mandan:

Is AI a thought partner for your recruiting decisions? Like, is every interview captured on Granola or something? And, you know, AI, is also saying this is actually a good candidate, or you didn't ask this question, which we should.

David Paffenholz:

We record every meeting. I use that data to frequently find out additional information about the candidate, but we explicitly tell our interviewers to not use those notes to make the decision on their recommendation. And so they have to submit a score of 1 to 4. That should be their score based on their own reasoning, explicitly non-AI influenced. And then what we can use the AI for is to contextualize. That of like, okay, what information have we gathered? And then because those scores are consistent from the recruiter, we can normalize scores, amongst each interviewer. So, if we know a certain interviewer is specifically harsh or too generous, we can normalize within that interviewer, and then we can use some of the notes to back up that information. But we never try to use like the actual transcript or content from an interview to make a decision.

Nakul Mandan:

You guys are obviously the most super users of Juicebox. What is a super user at the company using Juicebox, for, or in a way that most of the users of Juicebox might not be fully leveraging it?

David Paffenholz:

So I think 3 things that in my mind make a high impact are actually fairly easy to do, but most companies don't. The first is network sourcing. So, we have every employee upload all of their LinkedIn connections, and then Juicebox has a feature that predict— predicts additional connections that might not be a formal connection, but we think you know them. We then go for every role, we start with our network. So we look through all of the people in network for a given role, and then we ask intros to, them. it's like a Juicebox workflow, and we do that pretty aggressively. And so we have a good amount of referral hires, because of those network introductions. Second, we re-engage candidates that showed interest or responses previously, especially if it's more than 3 months ago. And so oftentimes when we reach out to an AE, we might hear, you know, right now is not the right time, I'm like wrapping up my quarter. Most recruiting teams then kind of like let that go because maybe the role is done and like what's the cause to like revisit that 3 months later. We actively track and I personally revisit all past interested responses. When have we re-engaged sense. And then the third thing is we use our hiring managers pretty aggressively. So every hiring manager is expected to be a part of the outreach sequence. And so we usually do first 3 emails are from the recruiter, 4th email is from the hiring manager directly. And then depending on the response rate, we might move the hiring manager further up in that sequence too. And that gives like so much better response rate and it's actually not that much work for everyone involved.

Nakul Mandan:

What's the function beyond engineering and recruiting that's the most AI native at Juicebox? Sales, marketing, product? Out customers?

David Paffenholz:

I think actually customer success, we've done some really good work on, because of our business model. We, so we have over 5,000 customers. We have around 1,000 sales-led customers. So those are eligible for our customer success, team. The ACV is like fairly distributed. So we have like a good amount of customers that are minimum ACV of like $6K, but then we also have customers that are close to a 7-figure ACV. And so the, way we handle those varies a lot. We have some customers with 500 users and others CIOs with too. And so what we did is we built two different systems, both of which I think are high impact. One is basically content creation for CSMs before every call. we have an automation that runs, it looks at all the past usage data, it puts it into a deck, and then it looks at benchmarks for, what do we expect the usage to be and where's the biggest opportunity for the team to improve. the CSM doesn't have to do any work in prepping for that. They can just hop on the call. They know they have the deck ready and they can start that conversation. and then second, it's important to us that especially for enterprise customers, we don't have AI responses. It's always a human responding, but we have an AI draft the response in every case. It has access to our code base, it has access to past tickets, et cetera. So it's kind of our own version of like an internal support agent, but it's optimized to be able to then be sent by the CSM themselves rather than an automation. And so I think that's actually made for a very efficient CS function where we're also continuing to invest more and more on what we can build. Built there.

Nakul Mandan:

Do you feel like everybody at Juicebox, because of the nature of the company, anybody who joins is already very AI-pilled, or you're finding it a spectrum that there are 5% of employees who are just supercharging themselves with AI and 95%— even with the com— within a company like Juicebox are still lagging behind?

David Paffenholz:

Yeah, so, okay, I have a slightly contrarian take on this, which is that, 90% of employees don't need to be AI-pilled. They need to be good consumers of AI workflows or happy to adopt them, but they're probably not going to be building the actual workflow because that would be a bad use of their time and resources. And so we have like dedicated builders, be that our go-to-market engineer, we have like a, a product engineer that just build internal workflows. They're very good at understanding what the internal pain point is and building it. And then yes, it's an expectation that the IC adopts that workflow. but they don't really need to be AI-pilled to do so because it should just make their work easier. And so we have some people who like to go above and beyond and like build their own extensions, et cetera. But like, you know, usually those aren't things we can roll out to the whole company. And so we'd rather have one or two people who build at a really high quality for internal workflow, and then everyone else is AI-pilled in the sense that their work got easier, not necessarily that they're building something with AI.

Nakul Mandan:

Do you feel, your sales and marketing org has supercharged themselves? Because you talked about CS, but, you know, building has become faster. If distributing becomes, 10x more leverage, then everything changes in the world of building, right? Company building. So, ha— has your sales and marketing team seen 10x more productivity because of AI? It?

David Paffenholz:

I'd say no. I think on the marketing side, there's actually like a, a lot of different AI tooling that exists. we haven't found huge success with any of it. Our content is explicitly human written, especially high impact content pieces. And so I, I'd say actually a lot of the marketing workflows are like intentionally human led. There are some things that work well in an automated basis. So we have like automated, you know, competitor ad analysis, like what are the, key points that competitors talk about? Out, but you know, it's kind of an informational input. It's not like a workflow that gets solved. And so I'd say, especially on marketing and sales, it's less clear to me how much has actually changed.

Nakul Mandan:

What do you think is the new constraint on company building? If building has become so fast in today's world, what's the new constraint in company building?

David Paffenholz:

So I think it's easier to ship features or product. In a way, that also means it's harder to ship them reliably or at like an enterprise grade. And so I think a constraint is deciding who you're building for and then actually being able to do that effectively. and we've definitely suffered from that in the past where like historically our culture was very much just go and ship and we ship new features all the time. Now, okay, if there's a company with 500 recruiters using this feature, how can we make sure we're actually educating those 500 recruiters and making sure they succeed in having like the relevant permissions and, layers in there? Those actually get really complex pretty quickly. And so I'd say deciding who you build for is a, a newer constraint, or like one that you now have to think about that maybe you didn't have to think about as much previously. And then I'd, say a second, maybe it's not so much a constraint as like an opportunity, is if everyone is building so quickly, what is the message that you communicate to your user, or what do they perceive your company as? And that's almost like a branding exercise, but perhaps it's now disproportionately more important than it was before.

Nakul Mandan:

You know, we all, wanted to see AI employees by now and supercharging with agents on every function. Like, we just let's discuss sales and marketing may have seen less of it, but there was a dream of the AI SDR and all of that, right? And that hasn't fully panned out. Do you feel like we are still like very far out from that world where there'll be a few managers in the company and then it'll be all like the IC roles will in various functions will all just be agents, or we actually very close to that? Where, are we in that journey, or do you believe in a different future?

David Paffenholz:

Yeah, I, I, I think we're still fairly far from that. I think we try hard to like get there and adopt, or, build what could be like a real autonomous workflow. But I think the best that we've gotten so far is that like, we don't have AI employees, we have maybe AI tasks and like the tasks can have gotten more complex and they can do more things. But in the end, it's still a lot of human decision-making driving those things. And so who knows where we'll get to? I think software engineering is perhaps the one where it seems clearest of like, okay, you know, there's so much output being generated. Even so, I'd, you I'd never give up our software engineers. And so what does that mean for the future of what a business looks like is less clear to me.

Nakul Mandan:

I guess in engineering, let's just touch on that on two questions. So one is, earlier in the, discussion, you said Ishan is directly managing all the engineers. That's a big team to manage for him. Is that more an AI capability that has gotten added that a single— the co-founder can manage? There's no engineering management layer needed because there's a lot of visibility he has across people? Like, how is it— like, is this some new thing, or you guys are trying like at the frontier here?

David Paffenholz:

I think maybe a bit of both. so I think it's like Ishan's passion. He wants to be able to be directly involved in all engineering and like, that's what he, wants to be doing. And so I think it's like a lot of that is definitely driven by him and he's very good at that. and I'd say the second thing is because there is now so many new workflows, we actually can have scenarios where an engineer owns a larger feature than they usually would. we can do an automated code review on it and then we just need a peer review in the end and we can go into production. And so I think a lot of those things have gotten simpler. I think in theory, an org could have also existed like that before all of AI.

Nakul Mandan:

How are you personally using AI as a CEO?

David Paffenholz:

I think my job, my job has changed a lot. I now have actually very little things that I do that are like executing on things. And I think a lot of it is like consuming and then making a decision. AI is actually very good at helping you consume information. And so I, I have everything kind of hooked up to my personal Claude outside of our enterprise Claude. and many times it's just me asking it to aggregate information that then I just, I individually consume. And so for example, if I wanna do a deep dive on like how's our sales going in a certain segment, I can have it look at every single Gong call, curate a list of snippets for me, and then I'll actually go watch the Gong snippets so I know what the customers are saying. But I don't have to do that work of finding and aggregating that. And so I think the efficiency at which I can consume information has increased by an order of magnitude, but then ultimately my job of like making decision based on that information has perhaps remained the same.

Nakul Mandan:

You, have a unique vantage point where you're seeing all these companies recruit talent through Juicebox, right? And the talent war is everything. Whoever wins that wins, the bigger war. Maybe let's talk about from your vantage point, you see Cursor, Ramp, all these as customers. Like, what are the best companies doing better? What's been your learning on what are the best companies doing better to aggregate the best talent in their companies?

David Paffenholz:

I think first off, for even those companies that you mentioned, and like, I think a lot of the, companies that people view as aspirational brands or places that they want work, all of those really treat hiring as an obsessive priority and they do irrational things to reflect that. And so that means that people spend way more time on recruiting than they would in another business. They probably have leaders in the business who are obsessed about recruiting almost to an irrational extent.

Nakul Mandan:

What is an irrational extent though? when you say irrational, like, are some of those CEOs spending 30, 40% of their time even at scale?

David Paffenholz:

I think in some cases much more than that without naming any like specific names, but I, I'm, whether it's the CEO or other like similar CEO figures, in some cases that is like their primary job of like, I want to go and recruit the best talent for this business, and that's what they believe they need to do to win. And so I think a lot of those companies either explicitly have that attitude or have someone in the business who has that attitude and goes and drives that for the company.

Nakul Mandan:

Are, the signals that they are searching for that are different? If in your observation, do they move faster? Do they actually move more deliberately and slower? Like, are there any common

David Paffenholz:

Yeah, I think almost across the board, they move faster. I think they're more willing to try different things. so they'll try to find whatever recruiting advantage they have, be that, you know, they're going to go and search in unusual places. They're going to try leverage the network even more aggressively. They are going to reach out to people purely with the intent of learning about the role, not even recruiting them, but then using them as a second referral to get to someone else. And so I think just the, intensity that drives or like the, amount of shots on target that they have, are, way more. Now, it also is a question of like, does that— is that the right strategy for every business? And like, if you're an AI lab or if you're like building something that's research related, perhaps it is. If you're a different type of business, perhaps it's not because you don't need the same level of research or you don't have the same talent constraint that other companies do. And so I think it's also a, a forcing function for those businesses that truly need to compete to that extent on talent to go and execute that way. and for most companies, it's perhaps, you know, an order of magnitude recruit less.

Nakul Mandan:

Do they focus on different signals? Have you been able to observe some of that?

David Paffenholz:

Yeah, I mean, we've seen a lot of things. We've seen, we've seen, this is like a fairly early stage founder. He was purely looking for people who had the maximum number of GitHub commits on the weekend. And the, signal that he was looking for is like, okay, if they're committing on the weekend, that means they're like writing code outside of their work and they're doing so in like open, source repos. And that's a very niche signal. It's a hypothesis of like, can I find good talent with that? And it's like a not that hard one to execute on. You know, you can go find that talent set, reach out to those people, and then see if you have success with hiring them. But then you have to keep going. You have to come up with new hypotheses and ideally ones that not everyone else is searching for as well. and so I think there's that kind of approach of doing that more aggressively. I don't know though, necessarily, if that correlates with they're going to have the most hiring success. I think there's kind of a different approach, which is I want to be known as a great place to work. I want to build that brand. And as I grow that brand, it's going to be easier and easier for me to recruit talent. When we, did a large billboard campaign in San Francisco San Francisco, in the fall of last year, and we've kind of kept some of that going. Initially it was with the intent of targeting customers. One of the biggest positive impacts of it has been on the recruiting side because people have heard the word Juicebox. If you've heard the company name before, you're much more likely to open the email and respond to it. And so what like company investments can you make to increase your recruiting outcomes? Because you have to be a place that people know about or are at least interested in to get them to work there.

Nakul Mandan:

Any common mistakes you've observed that you guide when, you as a CEO go in, sales conversations, you're talking to these founders or senior execs, like any common mistakes you've started seeing where you guide them that, hey, don't do this.

David Paffenholz:

Yeah, I think— so I think a common founder thing is like founders want to be Founder Mode. They want to like, you know, control the full process, and especially for recruiting, they want to be like the person involved in all of it. That often like misses the leverage that the recruiter has, and it's a little bit like in a sales negotiation. If you're trying to close a really important deal, if you as the end decision maker are immediately involved in the process, you are sending the wrong signals to be able to close that deal. And so using your recruiter in a smart way to do the negotiation, figure out what are their actual incentives. In many cases, the recruiter can literally find out the list of other companies that they're currently interviewing with, what their comp expectations are, and what we need to do to close them, because the recruiter is their ally in that process. And so I think tying that back to the question, most common mistake founders make is that they just want to do it all themselves and they don't leverage their recruiter to get a better outcome from that hiring process. And I think especially early stage founders doing like their first, I don't know, 20 hires, maybe they have a recruiter on the team, but maybe they're clashing with that recruiter, or it's like not going how they thought it would. I think often the underlying cause is that they're not letting the recruiter do their part of the job.

Nakul Mandan:

You also talked about like how outbound is the thing that you really propagate through Juicebox, but are the best candidates available through outbound, or they are available through referrals? And because they are locked in on their jobs, their bosses will make a counteroffer, they, they're happy in good jobs. Like, can the best candidates be accessed through cold outbound?

David Paffenholz:

Yes. but not by every company. And so the— we recently did, an analysis on this, or for a talk I did, we looked at the most contacted person on Juicebox. Unsurprisingly, a software engineer in the Bay Area was reached out to 240 times over a 12-month period. That's like almost every day. And that's just through Juicebox, you know, it's like not even considering all the, like LinkedIn, all the other platforms. And they responded to, I think, 4. So now there's two ways to look at that, right? Like one, okay, that's a, that's a means you're going to get a pretty low response rate, but two, who are the four people that they responded to and what are those four companies doing differently, than the other 236? and so I think there's like one approach of like, okay, you want to get that person and you know, you want the person that is like on paper, the most obvious hire, then you're going to be competing against the 200 other 240 companies. You have to come up with a way to do that. And the other approach is, okay, I'm, I know I'm not going to go for that one obvious fit. I'm going to try to find other ways to, identify talent that's relevant for me. And then maybe I'm competing against 10 other companies rather than 240.

Nakul Mandan:

That's a crazy stat. That probably is somebody who's at a branded company already. So you're, talking more about finding hidden gems.

David Paffenholz:

Yeah. And, but I mean, like hidden gems implies that it's like hard to find them. Like it's like a real gem. Like you have to go digging for them, but in many cases it just means like, okay, if your recruiting strategy is like, I'm going to go and reach out to everyone who works at OpenAI, that's probably not going to be a very fruitful recruiting strategy. And so thinking a bit more about like, okay, what is the type of like background or skill set that I think will be successful? Or even going one step further and looking, okay, if I want to recruit OpenAI caliber people, where did OpenAI caliber people work beforehand? What are their backgrounds? What did their journey look like 5 years before they worked at OpenAI? And can you find them at that point in their journey? And so it requires a little bit more work and insight, but then it's much more likely to get to the right outcome.

Nakul Mandan:

You also described the big AI labs now as star-studded sports teams. how literal is that analogy? Like for the researchers, are there literally AGI agents, I mean, there's always been recruiters, but like specialized agents who are going after researcher talent and OpenAI talent and Anthropic talent, and they get a different level of commission, like the bidding wars, take us into that world a little bit.

David Paffenholz:

Yeah, I'd say almost like the, kind of researcher game has gone to like almost a different order of magnitude. It's like almost outside of the recruiting landscape in and of itself. I'd say within the actual like recruiting landscape of like non, you know, researchers that are going from OpenAI to Anthropic, on the candidate side, because a lot of candidates do work with a recruiting agency, they might be getting a lot more opportunities at once than they used to before, and they get a lot more leverage to be able to demand their own compensation than they did before. And so I do think, and I think that's actually somewhat similar to what was the case in 2021 already. And so it's like a pattern that's somewhat repeating itself, but with a narrower focus of like, okay, we're looking for people who are AI native or have built something in AI or similar.

Nakul Mandan:

And what's happening with compensation? So maybe put it in two buckets, the researcher, bucket and the engineering bucket, but top engineering, like what, are you seeing?

David Paffenholz:

Yeah, exceptional candidates demand exceptional compensation. And so I think the, whether that's in the form of equity, whether that's in the form of cash, I think often it is in the form of equity, especially for a fast-growing company. It's like, okay, if you're a Series B company, are you okay parting with 1% of equity for someone that you really, want on your team? Traditionally, the answer to that is No, but perhaps for the right person, the answer for that is yet. Yes.

Nakul Mandan:

Would you say the 10th engineer joining, exceptional engineer, should get the 1% in your recommendation now?

David Paffenholz:

It depends on the stage of the company. I'd say in— if it's like a, an early stage company still, by the time it's the 10th engineer, probably yes. For most businesses though, especially in the AI era, like when we hired our 10th engineer, we were already at our Series B. And so that percent at that point, probably not. so depends a lot.

Nakul Mandan:

But a founder with 20 people, they've gotten a Series A done. What is your guidance for them on going from 20 to 100 exceptional people?

David Paffenholz:

I, I think it depends on the, situation of the company and the type of talent that company needs. If talent is a determining factor for that company to succeed, especially if it has a research style angle, then it's no longer a question of like, how much should you give? It's like, what is the market saying that you should give? And then you're, probably gonna be giving that. And you'll see pretty quickly based on whether you're losing or winning a candidate, what you need to close them. If you're a business, that's probably the majority of AI businesses being built now, you want great talent, but you also wanna build it in a somewhat sustainable way. And so in our case, we budgeted a given option pool. We set like rough equity ranges that we want to issue per role. And then we do have a rule, between Ishan and myself that in exceptional cases we'll have exceptional comp. And so that means we are willing to go 3, 4x above that range if we truly think it's necessary for the right person. Actually, exceptionally frequently. But I think it's really important for us to have like pre-aligned on that. Because it means that we can actually then go do it when we want to, and we don't have to like make it a, a big questioning decision upfront.

Nakul Mandan:

But do you— are you seeing also location-wise people taking a different tack that— because in the Bay Area it's extremely competitive, and Anthropic and OpenAI and XAI and all these guys are taking up a lot of the talent. Would you recommend, or would you guys open a different hub outside of the Bay Area?

David Paffenholz:

So we have a London office. We opened it like, 2 months ago, and the recruitment process there was very different from what we've seen in SF, to the extent that the SF market is an order of magnitude more competitive. and so I think we, haven't hired ourselves in New York from, but from what I've seen, I think, New York is probably somewhere in between. but SF is definitely unique in how competitive it is. Also given that there's this wave of AI companies that are all fully in person. And so if suddenly there's like everyone's back in person, but there's not that much talent that's like moved back yet. And two, everyone's looking for the same elite talent that is naturally going to result in a really competitive market and what we call the talent war.

Nakul Mandan:

Are you finding, for Juicebox but also for your clients that, on engineering they are able to take junior engineers, maybe even like people in college, but they're just super AI native. Kids are— seem like they're much more equipped for the real world much earlier in their age. I mean, you, you're an example of that, but are you finding more success for AI native engineering teams going younger or actually that the technical architects are needed because agents can do the CodeGen, like where are you guiding JuiceBox's own recruiting as well as for others?

David Paffenholz:

LLMs are like a new technological paradigm, which means we can build more. It's perhaps a little bit more like when mobile came out and you could build more, and there was like more younger founders as well that were building social companies or other types of businesses. Perhaps LLMs are similar in that they enable a new wave of people with, kind of a new set of eyes, which then naturally enables younger founders to succeed as well. So I do think that's generally true that there is now an easier case to be made for, a young engineer or a young founder to have a disproportionate impact. And I think we actually see that in the big labs as well that do, hire 20-year-olds, 22-year-olds, into their software engineering roles. That's actually like a very different narrative than what was the case 2 years ago when everyone was like, oh, because of AI coding, we're only going to hire really senior engineers because that's like where the impact is going to be made. That was like the consensus. Everyone agreed that was the case. Today it's like the opposite. People are hiring the, really junior engineers on paper because they're the ones that can use the tools the most effectively and then have the biggest output. and they've always been building with this mindset of, you know, I'm building an, a finished product or a finished feature rather than I'm just contributing a small piece to something bigger.

Nakul Mandan:

What's your actual stance on the thing you touched on? All the AI companies are here, or a disproportionate company, set of AI leaders are here. At least everybody wants in person, so there's more hiring here. People are flocking here accordingly. Do you think AI has actually increased the distance between the Bay Area and the rest of the world, or it will actually propagate— this is a local maxima and it'll go back to a little bit more propagation across the world where engineers from everywhere will be able to thrive using CodeGen and— because otherwise, how will this sustain? Maybe, just the broader question is how will the talent war sustain itself? Like how will those, I don't know, the next Juicebox hire against the current Juicebox and Anthropic?

David Paffenholz:

Yeah, so I do think that naturally, as a lot of, like, a lot of these companies were only founded in the last, like, 5 years, let's say, and those companies are now, like, scaling quickly, and some were founded, you know, more recently, others, like, closer to the 5-year mark, but that means they're all starting to build, like, global teams, global offices, and they're kind of repeating that same cycle that a previous wave of companies did as well. And so I think naturally, as all the, companies that are, like, AI leaders continue to scale, they will also scale globally because they serve end customers that are globally distributed. And so I think it's inevitable that the talent market to some extent will have the same ripple effects internationally. Now, will SF still be an order of magnitude more competitive? For the short term, that seems pretty likely, or at least I don't see any reason why that wouldn't be the case. That's just from what I see today. I, I, I could be totally wrong. In 2 years, we will be sitting here and seeing something different.

Nakul Mandan:

Okay, so let's shift towards the inner game. You know, you went straight from college to being a a CEO. You'd never managed a person before. Now you're running a team of 90, which is growing super fast and will probably be, you know, several hundreds in the years,

David Paffenholz:

I think that's like, especially the last 6 months have been probably the most intense time for that because the, this journey of having gone from like 20 to 90, I feel like has felt like the biggest leap. Like it's even just walking into the office and there's so many people where I don't know anymore day to day exactly what everyone is working on because there's too many people. I shouldn't know that. And so I think that's been the biggest change for me where I used to obsess about the fact that I knew everything that was happening at every given point in time. And I've had to, start trusting others that they know what is happening in their given teams. And so I think that's been like a big mental shift for me. And I'm still somewhat slowly coming to terms with it. I've been able to learn a lot from other founders who have gone through that journey, especially founders that are like a step or two ahead of me where they've gone through those same learnings and same transitions. I don't think there's many careers apart from being a founder where you so quickly start managing so many more people. So because of that, I guess there's also a pretty targeted set of people that I can go to for advice on it. and almost all of them are former founders themselves. and so that's been really important to me. And then I think the third thing is I've had to be comfortable with being wrong and changing. It's like embracing our own company values of making a decision quickly, but then being honest about when it was wrong. And I've made a— I think a number of wrong leadership decisions that I've had to course correct and figure out and like have grown myself by being able to reflect on that and get something wrong.

Nakul Mandan:

Is there a community of people around you where you go to them for different decisions? So for product decisions, you like this other founder, but for management scalability of your own, like you go for this, like, is that how it is? Or is it a much more tight to 2 or 3 founders talking to each other all the time and scaling together?

David Paffenholz:

Yeah, so for everything that's product related, it's my co-founder and me that debate it out. And so we'll lock ourselves in a room and we'll discuss it for 2 hours until we come to a, a decision. It's an intense process, but it works really well, and we trust ourselves because we feel like we know why we made the decision. And so we rarely go to anyone external for product or eng decisions. I'd say for like company running decisions, that's where I often lean on other founders. And there's a couple different founder groups that I'm in where people are pretty open and have— share their own learnings from going through this. And so fortunately, I guess in SF there's more than a handful of founders that have scaled teams pretty quickly to 100+ people. and many of them are, generous with their time and, share what they learned.

Nakul Mandan:

When you have decisions where you have zero pattern recognition, how do you go about that? Is it talking to others or like do you go to your investors, board? Like how, are you thinking through where you don't have any pattern recognition?

David Paffenholz:

Yeah, I guess it depends on the type of decision. If it's a case where I think I can get it wrong and it doesn't really matter and I can course correct in 2 weeks, then I don't go to anyone for advice. I just make the decision and we see what what happens. That said, in many cases, I have a feeling that it will be a, you know, a one-way door, like where we will actually regret getting this wrong. And so then there I try to go to the person that I feel I trust the most for that advice. And sometimes that might be an investor. I think more often that might be another founder that I've, built trust with and that has maybe also been vulnerable to me where they've shared something that's gone wrong for them. And then that's why I trust that they'll be honest in what they think. I think I view those as data points, and then still try to feel very comfortable in the decision that I ultimately make.

Nakul Mandan:

You're also now probably— correct me if I'm wrong— managing executives who— execs who are much older than you, much more experienced than you with the company growing. Does that create mental friction for you that they have more experience but on this specific thing I'm disagreeing and we are going in this different way?

David Paffenholz:

To be honest, it hasn't been a barrier so far. one, I think we've been fortunate that we've hired a great group of leaders. And then two, I just try to be, very open when I, have a decision that maybe goes against, what the, what the leader recommends. We recently had a case where like we got a hire all the way through the interview process, but I felt we shouldn't be hiring that person. And I kind of, wasn't sure if we should be hiring the role either. And we kind of came to a good decision there, but I also like tried to lay out and, talk verbally about here's all the, like, here's the reasons I think, maybe I'm wrong, but like, I think we should make this decision. And I think we haven't had a case where it like really was something that someone felt super strongly that shouldn't be the case. And so I think in those cases it's like easy for everyone to accept. that said, you know, as the business continues to grow, I'm sure there'll be cases where it'll be harder, and I'm sure there'll be cases where I have like a really major disagreement, and, that'll be a learning journey too.

Nakul Mandan:

Has your own leadership style evolved, like, to your own observation, where you've had to make— tweaks in how you communicate, how you carry yourself in the office every day, every week, as you've gone from 10 people to 90 people? And is there something that you're currently working on?

David Paffenholz:

Yeah, so, okay, I used to, or I, I still have a bit of this tendency of like, whenever I see something that I think is wrong, like be that in a Slack channel or like an opinion I disagree with, I will write that I think it's wrong, or like I will share my opinion on it. And in the beginning, I think that was good because I trusted my judgment. It meant we were moving fast. Now it actually creates like real disruption for others because in some cases I might not have the full context and their decision is actually correct. But even in cases where I have full context and I think my decision is, the right one, it's not always productive for me to share that, because sometimes the team is actually more effective by having that iteration cycle, getting it wrong, and then correcting it afterwards, especially if they are focused on that domain. And that's been hard for me because I still always feel the temptation of like, oh, I disagree with this, I'm going to share why I disagree with it, and I'm going to advocate for, what I think is the right thing. and I've had to stop doing that, at least in, a good amount of cases.

Nakul Mandan:

Is there a CEO mistake that you can share with us that you still think about?

David Paffenholz:

We just launched our MCP server. I thought an MCP server for a recruiting platform is kind of silly because you should be going into the recruiting platform to build. and we are kind of your AI native workspace. You shouldn't— you need to use an AI— an MCP to connect externally. And I felt pretty strongly about that. It was like product decision. and we, you know, did a bunch of customer interviews and we're generally, like, seemed like that made sense. I was wrong on that for two reasons. One, there are a certain set of MCP use cases where you're able to get Juicebox data from the platform, be that usage data, be that, say, types of candidates you reach out to, their open rates, where you're then able to combine that data with data from other platforms and you get a better net total output. And it's impossible for us to integrate with every other platform. And so if you have a layer in between, say you use Claude for that, is actually a net value add for the user. And then two, from a marketing perspective, it made it look like we were behind because we also didn't go out and say, this is why we're not building an MCP server, or this is why we're building an MCP client instead. And so that was like a mistake that I made that we corrected. We've since launched it, but I probably set the team back by like a month by doing that. And it was also a good learning journey because it's like, you know, I have to be comfortable getting those wrong, and then I also have to be comfortable telling the team, hey, I got this wrong, and here's how we're going to change it.

Nakul Mandan:

And how has your relationship with your co-founder evolved as— or had to evolve as you've scaled?

David Paffenholz:

We still talk about all important decisions all the time. The level of what is an important decision has changed a lot. And so it used to be we would align on every candidate's comp, together. That's no longer the case now because we've hired for enough roles that we have a gut feeling of where it should be and we know the other person will be fine with it. And so I think the level of trust was already high, but it's increased because we know how we, operate and we know what the other person might disagree with. I think the thing that's gotten hard is if we feel like there's something that's like going wrong in the other person's org, how do we handle that, right? Like, how do we figure out what is the case? How do we still respect the things that are happening at the moment? how do we do that without like, you know, disturbing the ICs or the team members that are actually working on whatever the given thing is? I think that's something that we're still like figuring out is like, how do we collaborate like, not just as co-founders, but also as like managers of different parts of the business? And, I think that's a, a new skill we've both been trying to figure out.

Nakul Mandan:

What about psyche management for both of you and individually? Like, you guys are both in office 7 days a week. It's an intense thing, 4 years straight and decades ahead, hopefully. So how, are you managing your personal life? Do you guys get time to, recharge at some point during the weeks and the months? How, are you guys doing handling all of that?

David Paffenholz:

Yeah, so one of the things we do that I love is we do founder date nights, which is just Ishan and me, going for dinner like once a month or once every 2 months. And we always end up talking about work. So it's always like work in there, but we're also like really close friends. And so it's like always a mix of the personal things we have going on, the work things we have going on. We usually have more work things going on than personal ones, but I think it makes for a good dynamic and also just time that we like— time for us to trust each other and spend time with one another that's not just in, the work context. I think that's been a important. We've also just gotten to know each other's personal lives and families really well. like my girlfriend, Ishan my girlfriend Alex knows Ishan super well. They spent a bunch of time together. Like I think it has become a deep, like a cross between a deep friendship and being co-founders, which initially I wasn't sure, like, would that be the case and how would that work? And like, I'm really happy that it is the case, because it deepens the trust.

Nakul Mandan:

Did you guys know each other from before actually?

David Paffenholz:

No. So we, the, we got to know each other in the context of like building projects together. we built two consumer apps that failed, but we always kind of had that like, oh, we want to build something together. It wasn't like, oh, we met socially and like we were friends and then we decided to build. It always started with like the fact that we were going to build. And I think in many ways that was productive because early on we knew what we were focused on and we didn't have any personal things mixing up. But over time it also evolved into becoming personal.

Nakul Mandan:

And do you guys use an exec coach for each of you individually or together to scale as managers and leaders?

David Paffenholz:

no, neither of us uses an executive coach. We have had advisors for different functions, so I had like a sales coach for a while. I have like someone I go to for marketing help, etc., but all functional specific and never personally. I haven't felt the need for it so far, but I'm also not opposed to it.

Nakul Mandan:

You know, you and I chatted briefly in the, you know, conversation before this, before the recording, but one thing on your mind is you have a girlfriend, you have a dog, and you're a founder at 25. And this is everything right now. Like, how does life look ahead in terms of balancing all of that? What can you share on what's on your mind on that front?

David Paffenholz:

It's hard. Like, I, I think, my partner and I know we want to have kids at some point soon. I also feel like I want to be able to be there for my kids, at least to, some extent. and so I guess that's like a weirdly big life question that, I then suddenly have to think about that's very different from like the, work things I think about. And, I don't really know what the answer to that is our board member is also a parent and he, he's also pretty young, and so he's been able to, give me some advice on that too. But, I think it'll also be something we have to take a leap of faith and, figure it out. From what I've heard from other parents, who are— from parents who are running companies, as they say, it actually helps give them a lot of clarity of like where they spend their time and what actually matters to the business. And so maybe there's some advantages to that as well.

Nakul Mandan:

What I've heard is that parenting makes you very focused. Like you don't riffraff, you don't waste time, you get the job done because every, minute is taking away from the time with the kids, I guess. What, keeps you up at night?

David Paffenholz:

The team has grown very quickly. I think I'm, very happy about all the people we've hired, but you know, the majority of them also just joined within the last 3 months. And so are we keeping the talent bar where we want it to? I think yes, but I don't know. We won't know until we've spent a few more months with, the new team? and then two, are we growing the team in the right directions? Like, are we hiring the right leaders when we should be? And like, those are all questions where there's like no clear right or wrong. Like, you'd have to go figure it out and we'll know in retrospect whether it was right or not. and I think maybe that's part of why it's like keeping me up more is because I don't— like, there is no clear right or wrong decision. And then the, second piece is being kind of more unique to the fact that we opened this office in London. the team, there's a amazing, but it's also the first time we've had team members who are not physically in San Francisco. and so they're with each other, which I think already adds a lot of value. But you know that now there's in a way like two teams to manage, and two geographies to manage. And so that's also, I think, just something we have to get used to and like something that, I think we've done a good job of so far, but we'll see what, that looks like.

Nakul Mandan:

Okay, next up is our quickfire round. So red flag you never ignore when recruiting?

David Paffenholz:

Asking for a job title just for the sake of it.

Nakul Mandan:

Most overrated startup advice?

David Paffenholz:

Launching too early.

Nakul Mandan:

Something you believe about company building that most people would disagree with you?

David Paffenholz:

Oh, that's a tricky one. At times you have to be okay disagreeing with the user.

Nakul Mandan:

Most overrated signal on the resume?

David Paffenholz:

Promotion trajectory.

Nakul Mandan:

Most underrated signal?

David Paffenholz:

Being an immigrant or moving.

Nakul Mandan:

One interview question that tells you the most?

David Paffenholz:

What's one thing that your current company does really well and what's one thing it does poorly?

Nakul Mandan:

One company whose hiring machine you admire the most?

David Paffenholz:

Cursor.

Nakul Mandan:

Cold email or warm intro?

David Paffenholz:

Love a good cold email.

Nakul Mandan:

this is set up for you. LinkedIn in 5 years, bigger or smaller?

David Paffenholz:

As a company, bigger. Town Solutions, smaller.

Nakul Mandan:

If you weren't building in recruiting, what, were you— what would you build?

David Paffenholz:

Something in the German market. I'm from Germany and, there could be more startups there.

Nakul Mandan:

So final question, you're 25, you— this is still just the beginning, but you've packed a lot in these 4 years. So if you were sitting down with the David who was starting Juicebox box, what would be the singular thing you'd tell them— tell him to understand better about company building?

David Paffenholz:

I think I'd tell the earlier version of myself to be even more comfortable in being wrong and telling others that I was wrong. and I think that's something that I still struggle with is like, admitting when I was wrong and, talking about that. And, I think in the cases where I do, it's like it shows vulnerability and it builds trust with the other people. and I think I was reluctant to do that for a really long time, even with my co-founder, et cetera. And I think I've slowly gotten better at that, but I had like a real reluctance to do so. And so I tell myself to do that better.

Nakul Mandan:

What changed it actually?

David Paffenholz:

I think it was seeing the reaction of people when I admitted that I was wrong. and usually it's actually a positive reaction in the sense that it makes it easier to get to the right solution. And so, being able to, do that more proactively, then that kind of opened it up for me of like, okay, like you know, this is what happens and that's a good thing that happens, so let me, let me do more of that.

Nakul Mandan:

David, this was amazing. Thanks for coming on.

David Paffenholz:

Thanks for having me.