If you tell your customer everything about how you work and they still buy you, that means that you're solving a hard problem. Thought leading should be the outcome of you trying to get at some truth or mission that you really believe in.

So DevRel is back.

How you do really great developer relations is literally. I'm

joined today by Swix. Swix coined the term AI engineer, founded the conference series which has 6,000 attendees. He co hosts Latent Space, which is the most popular AI engineer podcast, and he writes d x tips, a newsletter for people in DevTools. It's incredible. And he works at Cognition, which is the creator of Devon.

In this episode, we talk about why DevRel is so back, how you can do DevRel when your startup is not a rolling snowball of success and it's an uphill battle. We talk about a lot of other stuff, and I even ask what I should be doing with scaling DevTools to make it better. Enjoy the episode.

First of all, you know, we have a list of things we don't cover on it in space, and I think it's very interesting for podcasters to define their lanes and define what is not in their lanes. Because you it's usually podcasters and media people wanna cover everything. Right? And lack of focus really kills you. So for three years, I've had no bio on my on my we don't cover list.

And then Mark and Priscilla come along, and we're like, okay. Yeah. We do bio now. So they came inbounds. And but, like, I I think this is part of a general shift in AI.

You know, our job is to cover AI, explain it, and keep people informed and highlight good work. And a lot of people are moving to AI for science. It is not just Mark. Like, they're just Mark and Priscilla happen to be ten years ahead of us, but OpenAI's, like, head of products, moves into AI for science. Anthropic is spinning up AI for science.

The people have left OpenAI to start up material sciences labs. Terry Tao is working on, like, frontier math problems. I think the next stage, you know, once you've kinda solved MMLU and all this sort of general knowledge questions, is really starting to be PhD level stuff. And you probably need to actually focus in those really specialized domains. We cover coding very well because I have me and the last year have the career developer tools, but I have no business covering science.

So what we're actually doing is spinning up a second podcast for covering just AI for science all day all day long. And I think the challenge here is to make it less intimidating. And so this is why I titled my post for Mark Zuckerberg, Biohub for nonbio people. Because I anyway, I talk to anyone in SF, they are like, oh, yeah. I stay away from that because I you know, all I know is that mitochondria are the powerhouse of the cell, and I don't really know anything else about the the human body.

But I think, like, there's this is much more about scaling frontier AI than it is about specifically this this structure of biology. Like, it's actually very interesting how much nonspecialists empowered with AI tools can do better than specialists who only know the traditional thing. And, obviously, if they work together, which is the CZI approach, you put them in the same room, a specialist with an ML person, they could actually do more than they individually could alone. So I think that's that's a very heartwarming message. I think it means that we're solving harder and harder problems, and probably the next phase of AI is going to be a lot more in the hard sciences.

And because it's, like, kind of, I guess, AI has been most successful in coding so far, and science is, like, somewhat similar.

Yeah. There's a self self reinforcing loop, but let's let's not kid ourselves. Right? Like, we're just building b to b SaaS. Like, it's not gonna solve any dire problem that anyone has in the world.

Right? But, like, the you know, they are literally focused on solving diseases as rare as one. This is a term that they have. Right? Like, there are so many genetic mutations where there are no drugs, no treatments, no prevention, no no FDA approved things, and there's no resources for them.

So, like, one, CCI can can go and solve it. But the the interesting thing is that when you study rare diseases, you actually figure out what's wrong by looking at the disc between normal human and and the the sort of genetic mutations. And then you actually just understand the human the normal human body better anyway. So looking at the tails tells you the more about the the main body of of the cell. So, like, they're building up this catalog, the human cell atlas.

By the way, I I feel like I just turned this into, like, a Mark Zuckerberg sort of debrief pod. But Let's I think it's just so like, the main thing I wanted to understand like, tell people, software engineers particularly, is you have a toolkit for understanding systems. And when you understand how to solve problems and break them down, you could actually apply that to other things apart from just code or, like, just, like, I don't know, b to b SaaS. Like, can actually try to solve it to solve, like, biology and, like, just trying to really understand that. And that's that's my framing of what Mark and Priscilla are trying to do.

Right? Like, Priscilla's the doctor, Mark's the engineer, and they have $200,000,000,000 to go after this. And the human genome project I I I've tried to put a scale on my write up. Human genome project was something like 8,000,000 base pairs of DNA. 1,000 times larger than that is AlphaFold, where you map the structure of proteins.

Something like eight to 10,000,000,000 times larger than that is the virtual cell. Cell can contain billions of proteins. Okay. Once you have the cell, then you build up a few billion more times into organs. Once you have organs, you have a few billion more times into systems.

And what Mark is going after right now is the they they spent the last ten years imaging the cell. They're not done yet. They've only cataloged a thousand. We need 8,000,000,000 of them. And then we need to build up to the immune system because the immune system keeps us healthy.

It is nature's health care system. When it breaks, we are sick. When it when it's working, we're less sick. And and and so if we can figure out how it works and can completely model in computer systems how it works, then we can develop trials a lot quicker drugs a lot quicker.

So it's just about having being able to map these things out on a bigger scale than what they're doing and then

Than ever ever done. Like, how much, like, how much scaling do we have in LLMs? We have scaled. GPT one was, like, let's say, to the 20 in terms of amount of compute flops. GPT five, let's say, is about 10 to 26.

So that's six orders of magnitude of scaling. In biology, we probably have 20 to 30 orders of magnitude of scaling to do. Do they give a sense of how hard to that is? I I mean, I don't know. He's, like, super optimistic.

They wanna solve it before the end of the century, right, which means we have seventy five years to go. And he thinks that because of because AI, it will be faster. Right? That's the thesis. I don't know.

Okay. Right? Whatever. Like like looks like Super. All I know is, like, if they succeed, Priscilla Chan will probably have more impact on humanity than Mark Zuckerberg.

Scaling DevTools is sponsored by WorkOS. If things start going well, some of your customers are gonna start asking for enterprise features. Things like SSO, SCIM provisioning, role based access control. You could spend ages tearing your hair out, building these things yourself, or you could use WorkOS. Will, what do you guys do?

My name is Will Stewart, co founder and CEO of Northlink. We're a self-service developer platform, and we help teams deploy their most critical workloads into their VPC.

And you guys use WorkOS?

We use WorkOS for our SAML and OIDC integrations. It's a pretty exceptional product. It makes everything regarding authentication pretty seamless, and it's been instrumental for us to onboard our enterprise customers much faster. Building integrations with lots of different SAML providers is really challenging. We can do that for cloud providers, that's our job.

But we don't wanna do that for the other type of IDP. This is quite interesting because internal developer portal platform and then there's the identity provider. So IDP has three different meanings and we're internal developer platform and WorkOS is IDP for authentication. But yeah, it's a great product.

Thanks, WorkOS. Back to the episode. And so is that gonna be the I don't know if it's the third would it be the third career path for Swix to do? I'm gonna

I I I picked my lane. I joined Cognition. I'm I'm gonna do coding for a while. But we got two new hosts, and we're we're gonna start a little little second podcast. Yeah.

Just because, like, there's there's no answer to this. There's no, like I I think, you know, encoding is very competitive. There's, like, five, six different, like, frontier coding agents out there. They're all good. You know?

And, like, we're all competing on stupid minor differences. But, like, in in bio, if you solve cancer, like, go ahead. Like, you know, like, I'm rooting for you. Like, no no one's no one's against solving cancer. So, yeah, I I think, like, the this the the fun thing about hard problems is that they are just hard regardless, and, like, everyone is rooting for you.

The the only problem don't is the don't forget competing, just solve. Right? And it it is, like, the sheer hardest thing you can possibly do.

Yeah. I guess you don't have to worry about, like, DevRel for it. If you if you solve cancer, like,

Go ahead and say

People will come.

Hey, guys. We solve cancer. You don't need any DevRel.

Okay. Cool. We got the we got the sound bite.

That's it. DevRel is back. So

let's transition to selling b to b SaaS.

Oh, yeah. B to b SaaS and infrastructure rappers of AWS.

So DevRel is back. So back. Unbelievably back.

So most people in DevRel reply and say DevRel never left. That's because they never left DevRel. But I think in the broader world, if you just talk to founders and VCs and people whose job it is to have a broader mindshare than, like, just be in DevRel all day long, it did kinda die out for a while and and get uninteresting. And for a long time, you know, I I was thinking about this, even my temporal days. The sale that we made the the our first sale was a $4,000,000 sale, and it and it wasn't from developer relations normally.

It was just from a long period of open source plus top down sales. And I think DevRel obviously contributes to top down sales, but it is not a main driver. It didn't doesn't get any credit for it. Mostly DevRel is for driving bottom up sales. And for a long time, bottom up was kinda dead.

And the the main philosophy was, oh, you just you should just work on products. If you don't have a good enough products, then adding more DevRel to it doesn't matter anyway. So you just were you should just have all your employees with the product. And so I think, like, that's the that was the main sort of philosophy. And then you could see the numbers and and the the the the engagement on any DevRel topic going down.

And then suddenly, like, at the start of this year, I think it kinda started getting up a bit. And I think it really came to head for me with Anthropic, where I quoted this in the blog post that Anthropic, the the hottest developer AI company in the world with $1,000,000,000 of revenue in Cloud Code cannot fill their head of DevRel position. It's still open, by the way. They're still hiring. I just had the topic of hiring manager last week.

Yeah.

Yeah. Actually, just on a side note, why do you think they haven't been able to because surely I mean, they're obviously getting out

of this. I have info. I I I have info. I can share. It's it's some Yeah.

Some part of it's internal. Some some part of it is, like, they're really looking for the right fit. Right? And,

Yeah. Yeah.

They they can afford to be very picky, but having any job requirement open longer than four or five months is probably a bug, especially with the pace of the app. Right? So so and and but, like, Anthropics at the top end, but, you know, just and I get a lot of pings from founders, founders I invested in or just founders that I'm friends with. And, yeah, anecdotally, are pinging me a lot more for, hey. Do who do you know that's DevRel?

The standard answer is all the good ones are already taken. Who, you know, who can you poach? Right? And I think, like so the answer for me is I I don't know the answer. I just observed that DevRel is kinda back, and then I kinda leave the conclusions open to people.

Right? Like, all I know is, like, I think the last time I heard felt this way was early Netlify in twenty seventeen ish when people were like, oh, like, having really good DevRel drives adoption and actually lands us in big companies and and starts first to drive those use cases. That actually kinda went away during the COVID period, and now it's actually a lot more back than it used to be. So I can at least come from personal experience and say that that is happening, and I think a lot of people kinda resonated with that. I do think that the nature of DevRel has changed.

You know, I I came from an age where blogging and and and doing the conference circuit was a lot more traditional and and a part of the job. I think now being a lot better on short form video and streaming and, like, just online stuff is much more important. So the the job of the DevRel has changed, and, therefore, the people that used to be in developer relations back then may not be as relevant now, and they're kind of coping and feeling it. And the new kind of talent maybe doesn't wanna work for you because they have perfectly viable careers just working for themselves as influencers. Influencers.

So this is the the the link I posted in the in the chat here where you you can't hire these people that you wanna hire because by definition, what makes them good at their job is the reason they can't work for you.

Yeah. This was a really good post, And I think you, you know, I guess the my my takeaway from it is, like, you you were talking a lot about how the a lot of the best creators, like, wanna create content on lots of different topics because there's only so many interesting ways you can talk about most b to b SaaS. I definitely you said, like, if the company is like a CRM with a super base plus plus, like, a super base database wrapper. It's like, how how many amazing, like, hit technical deep dives could you guide to that?

Yeah. Yeah. I think this is a lot of venture pressure. Right? Like, I just went to the vet this morning with my my dog, and the vet the the vet was using a vet specific SaaS.

I looked at their booking system. It was, like, all tailored for vets. And, like, yeah, sure. It's a it's a wrapper on top of a super base or whatever, but, like, they are doing well. I I'm sure they're not, like, venture scale, but they solve a problem for vet clinics, and that's how normal businesses should be.

Not everything has to be number one in the world. You can just be, like, you know, we're we're we're your local one that you're familiar with, and you come here. It's it's just like a restaurant. Imagine if every restaurant tried to create, like, restaurant relations and tried to, like, be on Twitter and say, like, we're, like, the the best restaurant in the world. Like, the world would be very noisy, and, actually, that's that's that's how the world is for a lot of developer tooling.

It's very noisy because everyone's trying to fight for your attention. It's good in a sense that a fantastic time to be a a customer because they're they're they're all trying to solve your problems. But I think it is definitely fueled by VC and this desire to be number one when a lot of in a lot of other industries, there there is no desire. Just just make your customers happy and and make a good living, make a profitable business, pay your taxes, you know, sir support your customers well, make people happy.

Yeah. And I guess there's also the other point that, like, you were talking about how a lot of the top creators, like, they could just make way more money as solo people.

Yeah. It's interesting. So they there's the dream of way more money. Right? This is like the the way I put this is the the JK Rowling problem.

Right? JK Rowling destroyed a generation of writers because all of them thought they could just hide away in Scotland for, you know, six months and write a book and then immediately become a millionaire. Right? Because the the existence of superstars creates and funds this long tail of dreamers who will never make it. Right?

But they they'll try. And so even the dream, even if if they don't succeed, they're just they they just want that independence. They want the possibility of making them I think the the reality of creator economy is that it has gone higher in inequality than it was in COVID. Right? And I think that's a lot of the reason why creator economy is not hot anymore is because people found out that, actually, this is a there there's a paper called the economics of superstars.

It's a very, very old paper, but I highly recommend it because people try to aim for what they think the mean is, but the mean and the median are very, very different. And you're more likely to get the median outcome, which is much lower than the mean outcome because the world is not fair. Right? The world concentrates on pole positions and doesn't really give a shit about number four.

Yeah. I guess, like, it's almost like every single time someone talks about, like, dev creators, they talk about Theo and Primogen in, like, just

This is actually like And act honestly, they're not that big. Like, no no no hate. Like, they're they're they're way better at YouTube than I am. But, like, you know, like, look at real creators. Right?

Like, the the ones the the the actual, you know, outside of the tech bubble creators. Like, they they those guys are doing really, really well.

Yeah. But I would imagine they can charge a lot more if you're, like, very niche down there in DevTools. Right?

I think it's a mix. I think that so, like, Dwarvesh, you know, charges, like, 500 k per episode sponsorship. And he's a generalist. He talks about history. He talks about politics, and he talks about yeah.

I think just the general, like, go after if you're in the media business, go after the the most amount of audience. It's just this general rule. That's that's the source of your power. A subset of that is go after the thought leader audience, which obviously helps you get audience anyway, but he wants that respect. And then I think, like, that focus is good, but at the end of the day, even developer tooling startups and founders, they want to get on CNBC and Forbes and all that bullshit in order to reach the c suite who converge to the median audience anyway.

They they're not spending all day long in DevTools because I do care about health. I do care about politics. I care about history. Mhmm. So you you get my attention however way you can get.

And I think niching on DevTools, obviously, which is what you're doing, I think I think builds builds a library for great specialists, but, like, it's really hard to break out. Every everyone just, like in order to break out, you kinda have to generalize. Mean, I just told you, like, I had this no bio policy for a long time and now I'm breaking it. Like, you know, when you're lucky, the market pulls you. When you're not lucky, you have to, like, take a risk and generalize to that.

I do think that every creative actually feels that pull if their main goal is to primarily subsist on the audience attention that they derive.

Yeah. Yeah. Yeah. One one thing I wanted to go back to you on that you said earlier. You were saying about how there was this, like, consensus that around DevRel that for PLG, products was what mattered.

So you should just hire more people on product. And then your feeling is that it's different now. And so I kinda wondered, like, why can't I guess, can't you just do all in on product?

You can, and some people still do that. Right? I think right now, there's there's a couple of things. One, there is more of a shift towards platform plays rather than products. And whenever you have a platform, whenever your products is an API, then you gotta tell people how to use the API.

That's where DevRel comes back into the picture. It's obvious. I think two, money is a lot cheaper now than it was three years ago. So if people are funding crazy like, there's $200,000,000 seed routes. So if you have unlimited money, why not?

Right? And so this is the same reason the last DevRel bubble happened was because of ZERP. And, effectively, now there's localized ZERP. I mean, you know, we're we're not at zero interest rates, but there's localized ZERP in AI. Like, money costs negative amounts of interest because we want you to spend it.

So why not spend it on DevRel or try it out? Even if, like, it's hard to get any ROI metrics and attribute stuff, we just think it's generally a good idea to have good vibes and teach people how to use our products and create content on it. We do also happen to know for example, I think I posted in one of those blog posts that's in the show notes, that a lot more tech companies run on Twitter these days, including basically, the loop is CEO sees something on Twitter, copies and pastes it into the Slack, says, why don't we have this, or why is a competitor not gonna talk about this? So, basically, owning Twitter well is actually a job and actually a pretty good top of funnel, and much more so than it used to be five years ago. This is obviously an intentional outcome of Elon, which I don't love, but it's just reality.

And if you don't like accepting reality, then I think you're gonna do very well in developer relations anyway. But I think, like, maybe that will be my explanation of, like, why is it relevant even though PLG is obviously very important and honestly never went away in terms of its importance. I think, you know, we we just we just want to make sure that we we can spend the money wisely, but right now, there's not that much accountability in spending money anyway.

Yeah. What what do you actually mean by, like, negative interest rates?

Yeah. And so that's a that's a sort of analogy. Obviously, that's not a real thing. But so a negative interest rate in real life would mean that when you put money in the bank, you have less of it tomorrow. So you should not put money in the bank.

You should spend it as as fast as possible. The equivalent of that in VC is that I give you my $200,000,000. I expect you to use it well, and I'll sit on it. Right? Your the money should be burning a hole in your pocket, figuratively.

So you should try you should be trying everything you can with a positive ROI expectation to to return money because they are looking to for proof points, they are looking for people who can deploy money well, and you should have this feeling that the money is burning a hole in your pocket. So it when I say negative interest rates, it's kind of that that expression of feeling of, like, you should like, you you are rewarded for using money quickly and well. And so there's this incentive of, like, I can raise easily for anything in AI right now, so why not? And if part of that goes to developer to DevRel, then great. I heard this works from your buddy.

I think, obviously, me posting that DevRel was back further fuels that bubble, but I it was already happening without me. Right? So I'm just calling attention to something that's obvious.

Yeah. And I guess and then and then just, like, just to wrap that up is, like, the VCs wanna spend them want that money to be spent fast because it's seen as, like, a kind of a race to be the breakaway success or in all

these spaces? There's a there's a lot of bubble dynamics that do happen. So the faster a company ramps up in valuation, the more it's viewed as a success, and then the more you draw down your initial commitments so that you can raise fund two, fund three with those high numbers that are mark to market. So it feeds on itself until it doesn't, and then you get the crash. Right?

Everyone kinda knows this, but so far, it's been, like, a very good move. Like, you can't say you're not money good on investing in OpenAI at $30,000,000,000. Right? Like, now it's, you know, on paper 500, maybe the real valuation is up 100, but you're still up. So, like, you do whatever mental gymnastics you need, but at the end of the day, did you make the right call in investing?

And you probably did. And you probably want to do more of that in, like, in the cursors, in the cognition. Cursor went from 0 to 30,000,000,000 in two years. And if you miss that, you should you like, you you should be kicking yourself as a VC. Right?

Like and, obviously, it's super hard and, like, whatever, but I think it it it's I mean, that's the job. Yeah. You have to do this. You're like, if you're not in this business doing this, you shouldn't be in their business at all. So I I mean, I think that's great.

But, you know, I again, I don't wanna spend too much time thinking from the VC point of view. I do wanna think and spend the time thinking for DevRel and dev and and and engineer point of view because, obviously, those are the ones actually doing the work. And I I do think, like, those are the the the that's just the macro environment in which we exist, and we have to just kind of at least try to understand those dynamics and adjust for it in whatever decision we're trying to

Yeah. I will ask you questions on DevRel in this right now. But I do just wanna cover with you on that, because I think like, it's always interesting that you're like former life. I just think you you're very good at explaining a lot of the market dynamics, and it's it's always interesting from my perspective.

I'm in happy to indulge there too. I mean, I it's I'm a few years removed from the finance world, but, you know, I always think about it. Alright. So one thing I'm considering, I've been angel investing for a while, and the first investment was super based. That's obviously paid for everything else.

No problem. Not a bad one. The problem is I think I really like what Roloff Botha said on the the the podcast that he recently did before he he got fired from Sequoia. He was like, well, okay. So much more money has gone into venture capital, but the number of breakout companies has not changed.

It's still 20 per year. And that was, like, a really like, yeah, one one is objectively true. Two, it probably means, like, if you're the marginal player, if you're not Sequoia and you're playing this game, you're screwed. Like, unless you, like, just got super, super lucky, and I do know some of these lucky people. They're college classmates with the guys at Cursor.

They, like, obviously got some money in and, like, they made an early bet and got lucky. Great great for them. Right? But the the again, you're you're going for the mean expectation of what what what investing looks like in private markets, but the median expectation is you should you should understand that so much money is crowding out your returns right now, and economically, your expectations should be a lot lower and not higher. But people chase past returns rather than future expected returns, and that that's what's going on in in those dynamics.

So I am exploring starting a public fund to yeah. Because there's two reasons why you wanna do that. One, actually, there's a lot of public markets public companies that are navigating this AI trend really well. Like, Oracle, for the last three years, people my founders on the pod on my podcast have been talking about like, Oracle Cloud is doing super well. And, obviously, now they're now it's, like, very, very obvious.

But if we just bought Oracle, you kind of were just kinda long NVIDIA anyway. Bought NVIDIA. You're just kind of just long startups and long AI anyway, and you don't have any allocation limit. You don't have any limit on how much you can deploy. And yeah.

I mean, like, you know, I I think, like, there's think about it. Like, if you wanted to invest in ETF of the top tech stocks that everyone in Silicon Valley agrees is doing well, you probably do well. Like, there there's no there's no ETF like that apart from ARK Invest. And ARK Invest is a little bit too concentrated for for most people's taste. They have positions in, 30 stocks when you should probably have to a 200.

So the the volatility is super high and, like, you'll be up 50% and down 50%, and that's, like, it's

That would be so cool. I'm excited to hear about that. If people are thinking about starting a DevRel program where there is not much kind of momentum in the sense of, like, there's not many people coming in and it's it's kind of like from that standing start to what what do you say to people? Like, how

Yes. So I have a concept of uphill DevRel versus downhill DevRel. I also link this in the blog post. So that's definitely uphill. What you're describing is by default, you're not winning.

By default, you're dead. By default, nobody gives a shit. Right? Downhill is when you're already popular, and those are where, typically, when people regard someone as good at DevRel, they typically point to people who are downhill. Right?

And but it's very unclear how much is them and how much is just the value of the seat. Right? It's the same thing I had we had in trading, by the way. I like, sometimes traders are viewed as very good, but, actually, just they're just at a good company and a good good job. And, like, anyway, you put any monkey with a with a beating heart, you can do they they they would do well.

Obviously, that's not true. Okay? I do not believe actually, good companies can afford to hire great people. Right? So, obviously, there's a correlation there that, again, is super unfair.

But most people are uphill rather than downhill. Okay. So what do you do with with your uphill? Right? Like, you are a contender or challenger in some kind of market.

You have some belief on, like, what needs to exist in the world that doesn't believe, how well are you communicating that. Right? So you try to engage people on nothing else but the mission because probably you don't even have the products. Probably you don't have social proof. But what you do have is a mission that people should hopefully agree with you on.

Right? You like, and this is all a lot of this version is, like, sales is what people call it. It's not kind of sales because the yeah. You don't like, oftentimes, it's open source or it's, like, free trial. You're not even selling anything.

There's no money involved. You're not even, like, getting some enterprise buyer to, like, really champion it to their purchasing department or anything or use their budget. But it's still a form of sales of you selling the idea of something. So, like, can you get people motivated and excited about that? And I would definitely try to work on that because you need that for hiring and and fundraising, not not to mention actual sales.

Right? So, like, being that mission oriented vision like, articulating clear vision of the world that everyone wants to join you in is is the job. And if you don't have that job done, how can you expect anyone else to do theirs?

What makes, like, an interesting mission? Because I felt like a lot I felt like there's a lot of companies where it's like, their mission is, like, kind of more or less the product space that they're in and,

Yeah. And and I think that's a fair place to start, and I don't mean to say that I know any better than them. I just am articulating what I think you reliably works in the in the way that I do things. And so I I I definitely you know, there are multiple schools of thought, and I will defend to the deaf other people's way of doing things. But I I may not be cut out for it.

Right? I do have a rule for for missions, which is very, very simple and is a really good litmus test. The difference between a mission driven company and one that is just out for themselves is if someone reaches the mission before you, are you happy? So if you don't accomplish the mission, are you happy? Right?

The the the Silicon Valley HBO the cancer one. Did you cure cancer? Oh, someone else cured cancer before me. Alright. Fine.

You know? My my work is wasted, but humanity is saved, and I care more about that. Right? Do do we solve global warming or all that stuff. Right?

But, like, same thing, like, you know, do do we solve, like, serverless databases in Postgres? So, like like, any any mission that you particularly care about if you're, like, actually just really passionate about about about that mission more than your company winning, then you are actually truly mission driven. Very few companies are, and that's okay. Right? Like, you know, like, at the end of day, you're still trying to make money.

You're still trying to, like, play a competitive game, and a lot of people, all the best startup founders are just competitors. Right? They just want to be faster than the next horse. And because, like, they they grew up, like, competing in sports or whatever, and, like, they're just really good at it. And then it was great.

It's I mean, like, go go ahead and compete. You're you're not mission driven, but that's okay. It just it's harder for people to to get on board with you if you're primarily if they're primarily just supporting you. You know, they wanna believe in something bigger than themselves, and that's that's there's gotta be a universal truth about human human behavior.

So so is it, like, almost like you either they believe people would join you because they believe they're gonna win or they believe in the mission, but if you don't have either of them, then it's probably not.

Yeah. Yeah. Yeah. So so okay. I have a very fun clip from HBO Silicon Valley where the guy I I forget who the Hooley CEO is.

I want Oh, Richard. Was. No. Hooley Hooley. No.

Oh oh. It's Pied Piper. Gavin Belson. Gavin Belson goes, I don't wanna live in a world where someone else makes the world better place than we do, like, before us. Right?

Like okay. I I fucked up that quote. I don't wanna live in a world where someone else makes the world a better place better than we do. Right? So, like, he's clearly doesn't give a shit about the mission.

He just he just wants to win. He's a very competitive guy, and he's he's very successful. But, like Yeah. That like, a lot of people are effectively that. Like Yeah.

So so I think, again, coming coming back down to, like, the founders who, like, no traction, they're trying to fight uphill. I think get people on the mission, then get the social proof, get and build that into the the sort of community, build the products up there in as as far as you go. But, you know, the the product side is a given. I do think that for developer tools specifically, you want to nerd snipe people on some hard technical problem that you're solving. So the best version of this is you can tell your customer everything about how you work.

Right? You should not be like, oh, this is secret sauce. It's proprietary. If you tell your customer everything about how you work and they still buy you, that means that you're solving a hard problem. Because they're like, oh, yeah.

Okay. That there's no way I would do that by myself. Instead, let me hit this button to pay you $20 a month, and you can do that for me.

If they can copy paste the article into Claude Code, and they've got your product, then it's maybe less.

They they probably would not. Right? They like, in any in any code base of any sufficient capacity, it probably would not. Right? And and, like and that is where you earn your stripes.

But then by the way, that's also how you do really great developer relations is literally open up the box a little bit, demystify it, explain the system so they have a good working mental model of the tool that they're using. Because when people hire you as a startup, they're basically hiring you to be part of their team. Right? Like Mhmm. I have a team of 10.

I need this, like, widget that's, like, that's some b to b SaaS for thing. I expect you to provide me that software, and when I have bugs, I knee I report them to you, and I expect you to fix them in some timely fashion because you're a part of my team. If you're not a front working working, functioning member of my team, I'm firing you. So every, I guess, audition or every every public statement is you kind of putting your services out as a team member and saying, well, like, can we do this one very specific part of your job for you that's pain, but we'll make your life better and you just pay us, you know, a fair wage for it.

And you're, like, clearly know it better than anyone internally would have the time.

Prove that you know it better than anyone. Yes. Yes. So so demonstrating their expertise and experience is part of the the raw message. Most people, when they arrive at this stage of, like, they've started a company and they're trying to get get there, most people are very credentialed.

Right? They're just boring. They're just boring. They're they're like and I think part of that is arrogance. Part of that is, like, I worked fourteen years at LinkedIn.

Like, I did the Kafka thing, and, like, now pay me all the money for doing Kafka thing for you because I'm experts at Kafka. But, like, no. Like, you know, I don't really care about Kafka. I care about the the problem I'm trying to solve with Kafka. And, like, you have to tell me why your Kafka expertise matters at all, And that's a very different story than than what they're used to in inside of Big Corp.

And so I think a lot of the Big Corp former Big Corp founders when they transition to startups do struggle there because they're used to just dropping their employer name and winning. And, like, a lot of people don't give a shit.

Yeah. What makes it boring? What makes them boring?

Yeah. I mean, literally, they don't know how to sell because they never had to sell. So I think, like, there's a interesting bifurcation in former big core founders who's who become who spin out to startups. Some of them are so tied in and and sort of babied by the brilliant infrastructure that's all in house that that's at the big corp, and they don't know how to should the transition to the outside world. Others, because their work has been validated at big corp and now they're doing it for everyone else, that's sort of a class to the master story that's the same story that Temporal had that works very, very well.

I don't super know the difference there apart from you should probably try to use primarily open source tooling in that stack versus closed source stuff. So generic generally, I would observe that Googlers tend to struggle when they when they come out versus the other companies, because Google has a has a lot of internal stuff.

Because they're not working on intern they're working on open source projects internally

as much. I mean, Google a of projects. Right? Very famous ones. But all the open source projects are not owned and dogfooded by Google.

They they're just open source reproductions Oh. Of Borg.

Yeah. I see what I'm saying. So I guess, like, when you if you're the one building, I guess, this temporal story. Right? But, like, they built something that they used internally, and then so they were already the biggest.

They use Cassandra. Yeah. They they use Cassandra. They use the queuing systems. They use Elasticsearch, like, all these open source things, and here's the extra open source thing that we added on top at Uber.

Now let's spin out and and build a portal. Fantastic story. Easy to sell. Oh, by the way, one more one more thing for Uphill founders and Uphill DevRel, try to own a superlative in a important category. This is a blog post I haven't written yet, but I've been meaning to add it.

So, like, literally, like, being number one is something that matters. A lot of people fail to do that. They're like, we're more complete. They're like, I have fifteen years experience doing blah blah blah. I don't care.

Are you number one? Prove it. You're number one. Think that matters. Prove it.

It's really interesting. Like yeah. So and, like, the the this the the simplest easiest thing that matters actually matters is speed. Bunn is a huge success story now. And when they started out, it was basically just speed.

And then it was node compatibility, and then it was, like, here's, like, other innovative APIs, and then, like, just a broader developer experience. But, like, quote, unquote, better we have better developer experience. It's not measurable. It's not accountable. It's not number one.

It doesn't actually matter that much apart from dates they use. But number one in speed is good. You you should try something better than that. But speed's speed's a really good default because you can usually, as a startup, win on speed because you have less bloat than others because you are simpler.

Yeah. And and and and as a product factor that you push that's, like, the fastest, I don't know, database the fastest. You know?

Yeah. Speed, uptime, deep database obviously matters a a lot on the uptime side. I I do think, like, infra wise, things have shifted a lot for AI infra in particular, where you probably want to message, like, the well, Cerebras is is pressing a lot in speed, and, obviously, that works for Cognition. That's why Cognition and Cerebras are doing a lot of co marketing together. But I think just more broadly, like, infra, you want to have some kind of, like, vision on the standard that you're adopting and pushing because people don't have any idea what the standard infra should be.

Right? There is no docker of AI yet. Even m c MCP took a really good run, but I think it's kind of died out a little bit. I think it'll be around for a good long time just because it's it's, like, so prevalent everywhere, and it's it's really hard to get a new standard adopted. But it definitely didn't solve all the problems they were setting out to solve, and I think people are kinda looking for the next thing.

That's a hot take as well.

I don't think it's that hot. So, like No. The the simple measure is, like, roundabout. So and but in my way, like, I support NCP. Right?

Like and I but I did I never said I never said that NCP is, like, the be all end all. I just said, like, this I tried to explain why see NCP succeeded versus the other get attempts at sort of standardizing across AI companies. I think, like, the the measure that you have for this is run about in March, the last AI in New York, we did the MCP workshop. And, like, we were super excited about the MCP gateway that was the official one that was gonna be launched by Anthropic. And come around to August, Anthropic actually launched it, and no one actually cared.

I bet you didn't know that m Anthropic launched it because it just wasn't a big deal. Yeah.

Okay. Makes sense. Maybe

they could do a better DevRel.

Yeah. That that's very interesting. Okay. One thing I just wanted to ask you about about cognition and what you're doing.

Yeah. And I'm also figuring it out. I don't have a formal title yet. The the the the thing I'm sort of angling for is Cognition Building Cognition Labs, which is kind of the research publishing arm.

Interesting. I

think that Cognition does a lot of internal research, doesn't publish it because they're too busy working on products. And so I think it's, like, a nice mix of my talents as well as my interest in turn in terms of, like I'm mostly trying to serve the AI engineer industry, and I don't need to be too embedded within Cognition. You know? I still am running Lanespace independently. I'm still running AIE independently.

Independently. I don't need to be too too embedded. And so I I actually don't even know how Devin works. I don't even know how Cascade works or Windsurf. All I know is, like, what what what are the main tricks that we can share for the benefit of everyone?

What are the main evals that we can work on that, like, benefits OpenAI and Topic and Grok and all these other guys? Because they are demand they are coming into cognition for that. I think, like, also, for me, I've been neutral for the last three years, and I think that well, one, you know, you can see all your buddies getting rich based off of working taking aside an AI, and it's, like, really hard for me to stay neutral, really expensive for me to stay neutral. Cognition offered a really good deal where I get to keep running AI and Lanespace, and they in fact, they will actively support me featuring competitors of Cognition on Lanespace and AIE. And I just spent some time with them helping them with launches and with publishing research.

So I think that also contributes the overall mission of helping AIE anyway. So I I I think that's a relatively low lift. I I get to see things with the inside, which, like, informs my opinions. They get, you know, more than normal advising access to me for all things. And I've been advising many companies.

This is just a much big much, much bigger involvement than than any thing I've done previously.

That's super interesting. Do you think it's is it like DevRel in a sense, like, the publishing with it?

Yeah. So I I specifically asked to not be a developer advocate. They are hiring developer advocate, and I think I think the company does need one. So, again, they're also back, Cognition's hiring. And and and definitely feel free to out to me.

I think that the bar is high, but, like, you know, they they they do want good candidates. Yeah. I I I I think I just I I helped to phrase frame with launches, but I do think that my interests have kind of migrated towards wanting to make every time we do something, trying to advance the state of the art in that in in the coding, yeah, coding field. And I can if I can frame it and help that launch, great. But, you know, I think, like, it's more probably more important to just work on products and work on, like, publishing research.

So in some ways, this is just DevRel by another name, but I will try to do things that are not the normal scope of DevRel. Right? Like, publishing a paper is not usually part of the Dev Advocate job. But I think, like, that's how things will probably want to move for for AI DevRel, and I'm just exploring that, you know, because I I've done the other stuff, and I can hire and and mentor other people to do the other stuff. But this this this net new stuff is kinda interesting to me.

Yeah. Yeah. Doing things very differently. That's exciting. Actually, one question.

Putting you on the spot here is a pretty hard question to answer. I followed you for ages. You're like, I think the part that's hard to emulate is you're a very very smart guy. It's very obvious.

Oh, thank you.

Consume information at a rate that I think most people cannot, and then able to like synthesize it, seeing that from the outside. But I'm interested in the things that about your kind of success that people could learn from. What do you think your talents are that are like replicable for for other people?

First of all, you're you're very kind. I don't I don't feel that smart on many occasions, and I don't feel like I'm that successful yet. You know? And so, like, there's there's a lot of hunger left still. I think that in terms of, like, transferable lessons, I think writing is one.

I am holding a writing retreat again after three years off, and I think good writing is actually just the visible output of good thinking. And structured thought, the idea of how to thought lead, which is again, I still haven't published the blog post, but I've done one talk about it that that wasn't recorded. Leading thoughts for yourself and leading thoughts for other people is basically the same thing if you're just very authentic. So you should actually unironically just be very good at thought leading and study it as a field, study other thought leaders. And the problem with it with LinkedIn's version of thought leading is your your the your thought leading is the the end goal.

Thought leading should not be the end goal. Thought leading should be should be the outcome of you trying to get as some truth or mission that you really believe in. And you collect all the evidence. You collect all the social proof. You build all the mental models.

You do all the visuals. You you coin the catchphrases in order to arrive at the truth, and as a result, you will thought lead because you had the best thoughts. Therefore, you lead.

Is that like the Paul Graham quote where he says, to have good ideas, you need to become someone who has good ideas?

Oh, man. Think he says that. Yeah. It's still not so much like It could be misquoting. I I do so I I think the the quote is much more in startups.

Like, to to have a good startup idea, you must be the type of person that does have good startup ideas. And I I think that's, like Yeah. True but not useful. So I try to be a little bit more practical in, like, here are the seven things on how to thought lead, and, like, this is, the tricks of the trade. We'll study them and, like, then, you know, apply your own twist and or just do what's authentic to you.

And if you do it in this order, it's authentic. If you do it in the reverse order, it's not. So, like, I I have a whole system that I haven't really written up yet. But I think that, basically, clarity of thought and thought leading manifest in good writing, good writing manifest in good speaking, good writing manifest in good presentations and talks and what what have you. But I think, like, that focus is is basically learnable because I didn't have it when I started, and I just, like, looked at other people, and they sort of built my own playbook.

You can basically see it across my writing for for the blogs and and the the DX tips that I do. I think that maybe the other thing that I try to also tell people about is having a longer term view of what will really matter in five to ten years and keeping a mind keeping an observant eye on the stuff that you used to believe that you thought was the most important thing to do today, How much is it really gonna matter in a year? And if it doesn't matter in a year, why are you why does it matter at all today? And just really tuning in on the things that matter. This tends to be driven by economics because money, but also it tends to be driven by real technical problems and real hard challenges.

And so that always drives you towards being more real with yourself, whether it's, like, you know, technology holds you accountable or money. And so then then you then you hopefully end up making better decisions on how to allocate your time and and and resources. And I think, like, if if I can sort of impart an impart an economist mindset to people who work in tech and don't have the econ mindset, it is mostly that. There are a few drivers of your economic output. You have a production function, which is the Tyler Cowen version statement of this.

Stuff goes into your brain, in your eyes, in your ears, and what have you, and stuff comes out of you. How do you increase that quality of what you're doing? Just be observant about yourself. Sometimes that means journaling. I try to journal.

I don't do it very often. But some amount of reflection and honesty of yourself on, like, what the most important thing in the world is would be really good. Because I think even today, there's not enough of that. You still see YC put out a lot of stock. Everyone's hating on this, like, chat ID thing.

Everyone's hating on Cluevi. But even there's this, like, YC founder the other day who, like, came out with, like people they did their launch YC, and it was, like, React dev tools but better. And I'm like, bro, Facebook puts out React dev tools for free. And, like and I'll I I literally asked in the comments. I was like, okay.

So your plan is to be Cursor for React? And they're like, yeah. And, like, are you sure you used to work at YC? Because, like, Cursor will be Cursor for React. And if Cursor doesn't do it, Vercel is gonna it.

So you'll be number three at best. And it's like a really shitty position. And so, like, I I I don't think that these guys are being honest with themselves. They're just building. And and, again, no hate.

Right? Just you know, that's that's better than me. I I I I I don't think I I can build to a sort of what they call YC launch. But I I do think that it probably won't matter in a year, and they already know it, and they're just denying reality and just, you know, waiting for the right idea to hit.

Yeah. Yeah. And I guess that is if you are going through YC, you do need to be in a good position. Part

of this I'll maybe drop one more thing. Right? Do join communities. I think that a lot of people, when they listen to this podcast, they're by themselves. You know?

Usually, they don't even have a person to talk to about it. And that that and, you know, we are actually very social huge creatures. And so the way I put this is that, you know how Thiago Fortier has this idea of a second brain. Right? Your first brain is the one that's in your skull.

Second brain is the one that you externalize to your journal and your notes and all that. It is your external augmentation memory system. Your third brain should be your buddies that remind you of stuff that you missed because it's there are many studies that show that having two people focus on one thing helps because we're much better correcting other people than we are correcting ourselves. They they have other memories. They they know people who know people.

So your network is actually a really important part of your overall identity and value as a person, and a lot of people don't work on their network enough. And part of that is just being an active contributing member of communities, starting your own communities, and learning how to, like, basically have that relationship over long multiple years where you don't you sort of, like, give freely and don't expect anything back in return, but, you know, some days, some good karma will come back to you, and that's great. Like, you I think you just need to find a way a scalable way of giving good karma, and good things happen. So I I I do think, like, that does that that did factor into my decision to move to San Francisco for the AI wave, and that's benefited everything that I do.

Yeah. I I think that could be another sound bite of a scalable way to give good karma give out good karma.

This one's yeah. This is a harder one because, yeah, some people are very introverts. So, obviously, it's a it is a lot harder for for some people to to to sort of, like, participate in community. I will say that a lot of the introvert a lot of the speakers at conferences are introverts because the best way for them to speak is when no one else can reply. So Yeah.

So so, you know, you don't need to be the biggest party person because those are the extroverts. Right? They hang out there. They actually sometimes don't even do as well. They don't reach as many people as the introverts who find a highly leveraged spot to do community.

Yeah. Yeah. I definitely put myself in the introvert bucket. But as a

I also feel like fairly introvert, but here I am, like, running AIE next year is only 6,000 people. Algo ee. You know? And so, like, I my job is to curate the tracks of AIE, the track hosts of AIE, then curate the track talks. And this whole network it's a network of networks and networks, and Yeah.

You know, I I get to be associated with that. And and if I need any help, like, I can reach out to them, and and, likewise, they can reach out to me. So I think I I do think it's, like, a really nice form of human connection that I think is underrated when you just talk about personal accomplishment and achievements. Because, yeah, you you'll probably not be able to scale as quickly as just not like, you know, like, the way I for this is I don't know everything, but I know I know people who know everything. Right?

So That's very good. Good. I think it does make you nontechnical because then you traffic in relationships rather than code. But I think you can stay decently technical enough to effectively become a distributed engineering manager of those things. And and I think, like, the more senior you get, you you do tend towards that anyway.

Mhmm. Okay. This is a scary one for me and a very selfish question, but you're someone who's done a lot of things that I admire. Like, what if you're looking at scaling DevTools, what what would you advise

me today? You you can't I know. I can't.

You can't. You don't have to hold. You don't have to hold back. This is scary

for me. No. Right. Self selfishly, I kinda like it as it is because, right, like, the the what's what's brilliant is that you're pouring all your energy and effort into, like, this one very specific niche. And so you're you're kind of, like, you know, you're you're you're you're not getting paid by me, but you I'm paying with my attention and all that.

But, like, you're part of my team on, like, interviewing, you know, founders on dev dev tools and understanding, like, what their perspective is and asking engaging interesting questions. I think these are all, like, very valuable things, and, like, it's weird. As a creator, I want to generalize, but I don't want you to generalize because I I want you to be the DevTools guy, right, for for for for me. And, you know, clearly, I don't know if that's exactly what you want to to to do. I will say, like, I do think your recap thing of, like, a 100 episodes, I think, was good and actually could probably be developed further.

I think probably the path for you that makes the most sense is kinda like the Lenny path where Lenny is kinda owns product management. And so if you want to own developer tooling, then you build you sort of write resources for them. I do think writing does help to consolidate your knowledge better, and it builds you up as an as a domain expert. But, also, it just makes your contents easier to consume in other formats because you can't because podcast is is very limited, and there's only a certain amount of listening time every day. Yeah.

So, yeah, even I'm bumping up against that. Right? And so what I would say here is I have this concept of the stream particle stream duality of knowledge. I don't know if

It's the case that I got.

Particle wave duality of knowledge. Is this is is an idea of quantum physics. Like, light can either be a continuous wave or a single particle depending how you view it. Yeah. And the same way for knowledge.

What a podcast is is mostly a continuous stream of guests. They have most mostly overlapping things because you're one hour and, like, you're covering that bit of the hish they they are bundles of things. And so you can rebundle your podcast and your knowledge in other ways that are more consumable for people in that particular role. And I think part of that can be, like, here's the wiki entry on on doing, I don't know, like, marketing for ID terminals. And so if I'm an ID terminal guy, go straight there.

Right? So converting the continuous stream into a solid particle that is a consumable artifact is good, and then the other way around is also good. Right? So people just, like, listen in different formats. And I think every now and then, like, recapping is is a good way, or, like, doing, like, a limited series of, like, here's here's the ground truths that'll be evergreen.

So I think, like, the the other challenge or interesting challenge of apologetics is really very bound to time, whereas you probably want to build work on evergreen truths. You know, I kinda think of this as, like, a sawtooth of of content creation. Right? Like, a lot of content creation that it goes kinda goes like this, and you go back to zero every single time. What you instead wanna do is sort of build compounding assets that just keep going up, and so every new effort builds onto the asset.

And so I think what I would try to define for you is what the asset is. I literally I use this word asset with my Lean Space team. I'm like, well, know, the pricing chart for me was a was I did that for a year. I maintained the pricing chart. Every time a new model came out, I charted versus price.

And, like, I put out the chart again. And it's it's easy work once you've done the the initial work, but, like, it's useful resource for everybody. So, like, you can build a resource. That's great. If you can build, like, here's the ultimate guide to blah based on talking to a, b, and c, that's great.

You know? And and I think that's what a lot of what Lenny does is, like, at the end of the look at his end of year recaps. He will have, like, if you're trying to do x, well, I made this and, like, go back to it. It's probably still relevant. You know?

So that's how you sort of, like, have that sawtooth up in content because you're building and mapping out what the relevant areas are. To be clear, I don't do this for the in space. I don't do this for AIE. These are just my live thoughts on what the next step is obviously going to be, and probably Lenny is one of the best practitioners of that. Mhmm.

Yeah. This this is amazing. Thank you so much. I I'm gonna think a lot about this or try and act on it, I guess. Thanks for coming on three times.

I hope this is this is not the last one. I'm gonna be inviting you sequentially again.

I I will be I'll be a friend of the pod whenever you have me. I just I admire what you're doing, and I I happy to support it anyway. Yeah. And, you know, all the best. I I keep keep trucking.

Thank you. You wanna send anyone any anywhere?

Yeah. So probably by the time this is out, we'll have done the AI Engineer Code Summit. Watch the videos. Usually, we get, like like, last time we got, like, 90,000 viewers on the livestream, which is, like, pretty pretty crazy. I'm doing We collected all you know, you like the developer tools.

You like AI. Collect all the developer tool companies in AI and put them in one spot and asked them to recap, like, their year in 2025. So if you want, like, one place to recap it, that'll be that'll be it. You know? Which, again, you know, I I think coming back to the particle wave duality thing, it's really nice to just, like, look back on the year.

So I intentionally made this basically the last industry conference of the year, and so everyone has not like, they've shipped everything they're gonna ship for the year. Gemini three is coming out soon, and then we'll just announce a recap and have a year end perspective on everything.

Yeah. Amazing. And it is incredible. Best conference I've been to. Okay.

Thank you, Sweks. Thanks, overall for listening.