428. The Future of Large Language Models, Open vs. Closed Sourced, and Why 2024 is the Year of Inference (Ashu Garg)

428. The Future of Large Language Models, Open vs. Closed Sourced, and Why 2024 is the Year of Inference (Ashu Garg)


Ashu Garg of Foundation Capital joins Nate to discuss The Future of Large Language Models, Open vs. Closed Sourced, and Why 2024 is the Year of Inference. In this episode we cover:

  • Venture Capital Investing Strategies and Ideal Founder Characteristics
  • AI Model Development and Its Future
  • AI Models for Legal Tech and Their Potential Applications
  • Tech Giants’ Positioning in the AI Shift
  • Google’s Innovation Dilemma and AI Adoption

Guest Links:

The hosts of The Full Ratchet are Nick Moran and Nate Pierotti of New Stack Ventures, a venture capital firm committed to investing in founders outside of the Bay Area.

Want to keep up to date with The Full Ratchet? Follow us on social.

You can learn more about New Stack Ventures by visiting our LinkedIn and Twitter.

Are you a founder looking for your next investor? Visit our free tool VC-Rank and weโ€™ll send a list of potential investors right to your inbox!

Transcribed with AI:

0:18
On today’s episode, Ashu Garg joins us from San Francisco. Ashu is a General Partner at Foundation Capital, an early stage venture firm headquartered in the Bay Area. At Foundation, Ashu has invested in 1 decacorn in Databricks and 6 unicorns. Before joining Foundation in 2008, Ashu was the general manager for Micrsoftโ€™s online-adverstising business and was a consultant at McKinsey. Ashu, welcome to the show.
0:43
Thank you for having me, Nate, of course. So
0:46
prior to jumping in today, can you give us your path to joining foundation and ultimately becoming a general partner there?
0:53
obstinately Nate, you know, I’ve been at foundation since 2008. So it’s my 16th Theatre at the firm, which is roughly half my professional career, I’d spent 15 or so years prior to foundation at a variety of roles, mostly in software, and McKinsey Cadence Design systems, a couple of startups, and then most recently to different students at Microsoft, the most the last operating student I had was running the online advertising business for Microsoft. And that was really, you know, in 2006, one of the situations where Microsoft and others like Microsoft, Yahoo, Google, and others, were really applying machine learning to optimise online advertising, online advertising was the killer app for machine learning. And so I got very, I got exposed to machine learning very early in its cycle in 2006. And that’s really influenced a lot of what I’ve done ever since.
1:45
Got, if I read correctly, I forget where I read this, but you were thinking about starting a company at one point, I believe you’re a trinity. And then ultimately, from a conversations with a number of founders and ultimately prompted you to join the venture side, is that accurate?
2:00
You have a great memory. So I left Microsoft, very clearly, with the idea of starting a company I had never thought of venture as a career. It was it was not on the list of things I had seriously considered. And having spent a few years applying machine learning to video to display advertising. My Insight was, there was an opportunity to optimise video advertising, which was very, very nascent. This is circa 2008, a very different world. Most online video was short form, think YouTube. And the notion of having premium video on on the internet was was a new idea. And I wanted to facilitate that and create the monetization ecosystem for that. I spent, you know, six months as an EIR trying to figure out sort of the right point of entry, the wedge. And as I was doing that, I met a team, the founders of freewheel jog in, and John and others I was like these guys are so far ahead of we have the same problem. And so I consider joining them. But I was I was very convinced that I wanted to do my own thing. And so I kept looking at other ideas. But nothing really inspired me enough to want to spend the next decade. And at the same time along the way, I started pitching various ideas to the folks at foundation capital. And one day, they said, just come in and join us and spend a year and we’ll go from there. And so that’s what I did in the fall of Oh, eight weeks before Lehman happened, and weeks before sort of global financial crisis. And then my first investment was freewheel. Because I turned around and said, look, the best idea I have is is this company called free wheel. And the rest is history. That was my first investment a few few months later. And that really became the basis my career because free wheel took off it became the first real ad server for video was a very early acquisition by Comcast going back, you know, in those days for give or take $400 billion. Once upon a time, that used to be a big number. Yeah.
4:00
When I when I read about your story, I was reminded of something I think investors are a bit split on and that’s whether or not founders actually need help from their VCs. Where do you where do you stand on that? Do you think the best founders actually need help from investors? How do you feel? Or what are your thoughts on how investors can actually add value to the best founders? Or there’s any
4:26
I think there’s a couple of different questions embedded in what you say, look, the best companies succeed with or without their investors. You know, I think it’s it’s it’s very naive for any investor to believe that hey, this company wouldn’t have happened without me. So if I don’t invest someone else, well, if I don’t help with the right time, the founder will find some other way to make things happen. Because the best founders don’t take there’s no concept of taking no for an answer they take they find a path so I think that is that is absolutely, like we are at the same time, the best founders still need at critical moments help in getting over the finishing line on something. And that could be a critical decision. It could just be counselling and advice. It could just be a sounding board. Very often it’s closing a key candidate. I think one of the things where I feel like investors that we really spend the time can really help is closing key candidates, those key candidates want another perspective. You know, the CEO doesn’t have a lot of senior executives and senior people around him or her. And so being one of those being one of those people to help Polsky candidates is very useful. I think introducing them to key executives early in the journey of the company. You know, I think companies are always one step one person away from greatness. One hire away from greatness. And so you find the right VP of sales, you’ll find the right CRO, magic begins to happen. And then I think, if you look specifically at the segment, I focused I invest almost exclusively in pre revenue, pre product companies, started by very technical founders, with limited go to market experience. And often you know, it’s their first startup. So that’s zero to one, that idea to crossing the chasm or sort of crossing the valley of death, as we call it. I think investors can be very helpful. We can help on helping them think through how to position package the product, what is the product, you have a technical idea, you have an insight around a problem. But what is the product that solves that problem? I think we play a role. How do you take that product to market? How do you find your first 10? Customers? Who do you sell to? What do you pitch? Once you have a dozen customers? How do you scale that? How do you find the first salesperson, this first marketing leader? How do you how do you sort of kickstart dementia? And then, of course, you need to go fundraise. You know, the seed round only goes that far. And helping founders go through especially first time founders go through the process of raising a Series A, I think it’s something we do it for a living, we’ve seen hundreds of series A’s, whereas most founders have seen their first series A Yeah, even the most successful founders are gonna raise early on. But again, it’s a little bit like cooking, there’s no formula, you know, every recipe is different, every company is different. Yeah.
7:26
Yeah. I mean, someone said this not that long ago, but even the most successful founders are going to raise five, six rounds of venture capital, maybe, you know, maybe a couple more. But it does put in perspective, just the level of experience that you know, an investor like yourself brings to the table who’ve seen hundreds of rounds, versus the entrepreneur who even has success is gonna see very few of those. So as I heard you describe the typical founder that you work with it, it almost sounds like a company describing their ICP. For who, who they who they landed, and you know, who they were focused on? How important do you think it is for investors to have a similar thought process for like the ideal characteristics of what makes a founder that would be a good partnership for for you or for them? Because as, like, VC has been fairly generalists over the past couple of decades. And now we’re seeing hyper specialisation or we’re seeing more specialist funds, or we’re seeing investors that may focus on a particular type of founder. Do you think specialisation will continue? And how important do you think it is for investors that are looking to make a name for themselves? To focus on an ICP a founder similar to the way that you articulated to me who makes a good fit for you?
8:40
You know, I think there are infinite strategies possible in venture. And I think what’s important both at the firm level and at the partner level, is to have clarity around your strategy. I think a generalist strategy is is is very much on the table. It depends on the stage, it depends on the sector, set of sectors, it depends on sort of your ownership targets. I do think that for investors who are early in their career, and who are looking to go or looking for higher ownership, specialisation matters a lot. But it’s not necessary. I think there are alternative strategies. But if you’re going to be the lead investor in a seed stage round with that opportunity comes obligation. The lead investor in a seed stage round has taken effectively the obligation love of the founder of helping the company graduate to a Series A and your ability to do that is is enhanced by the fact that you have you know, some pattern recognition. If I’m if I’m investing in a cybersecurity company today, a social media network tomorrow in a gaming app day out After tomorrow, it’s just harder not impossible. I think there are people who there are people who probably are way more talented than I am and can do that. But for me, that kind of spread would make it really hard for me to sit in a board meeting, or have coffee with the founder and help introduce them to candidates, like the VP of Marketing for one looks so different from the VP of marketing. The other. Yeah, and even within sort of the narrow, I live in the world of AI enabled, or AI, first apps and infrastructure. And that can be anything from business apps, to, you know, the enabling data infrastructure, all the way to sort of things like cybersecurity, and dev tools, even that’s a pretty wide spectrum. I only do a small slice that I and I still keep thinking about, okay, is this to biggest like, should I should I narrow even further, but again, that’s a function of my strategy, where I am taking a lot of early stage risks. Almost every project I invested has very little code. There’s there’s the you know, there’s the a fragment of an idea of technical insight, a team that has some unique technical capabilities, and a massive market.
11:10
Going back 1516 years, how has foundation strategy changed, since when you initially joined the firm? Whether it’s how you guys think about price, whether it’s how you think about ownership? Are there any major changes in strategy that you’d highlight?
11:28
There’s three pillars to our strategy, high conviction, high ownership, high cost, in two out of the three have stayed consistent. So you know, we’ve always been a high ownership firm 15 to 20% is our target point of ownership at the point of entry, we’ve always been high cut, our goal is to be the first call a founder makes with good and bad news. We we we almost always take board seats, we are highly engaged investors, I think high conviction has been an evolution for us. And let me unpack that because it means a lot of different things. But it really starts with the fact that as a firm, we’ve chosen to focus on a few things. So focus on things where you have conviction and choose not to do everything else. So we are, you know, from a technology standpoint, we’ve made a really big bet for the last 15 years in AI, and over the last decade in blockchain and crypto. And those are sort of the two technologies or other, you know, things that we could have chosen or we could still choose to do, we’ve chosen not to do others. From a sector standpoint, when you overlay sector from those technologies, more than half of what we do is enterprise software. We also have a very distinctive in a lot of history in FinTech. And that’s a big part of what we do. And then crypto, which is both a technology and a sector. And that’s that’s sort of the world and that’s, you know, it’s a big chunk. But it’s you know, there’s lots of things we’ve chosen not to do. That high conviction applies to how we, the people, we hire the culture, we pray their decision making. And so our model is about, you know, when when someone has conviction, the organisational design, is to support that conviction. And then for that person to take accountability for that conviction. And so in our decision making model votes are advisory. And if a general partner really has conviction on a project, they allocate budget, they they have for the large part, the freedom to make that decision. That means the process can move quickly. But at the same time, it generates intense debate. Because if you don’t like a project someone else’s do your by just quietly vote against it. You have to create a ruckus and say this is why I don’t like it. That’s great. We have intense conversations,
13:48
would you say from just reflecting on the past 16 years that you’ve been investing at the firm, that the best investments have been non consensus and have been a bit more contrarian, where someone was going out on a limb,
14:02
you know, at foundation for sure that to be true? I don’t know if I could generalise for the industry. But we have a very diverse team in a good way. We have very diverse investing styles. And so our greatest investments have always had a lot of debate. Yeah, I
14:16
wanted to you mentioned AI, and that you’ve been spending a lot of time on the infrastructure layer, also the application layer, but I read an article that you wrote, which was titled Why 2024 will be the year of inference. So I want to dig in and talk about some of the sub topics but prior to going too deep, can you summarise at a high level what you mean by saying that this will be the Europe and France in some of the key takeaways. The article conveys
14:42
that if you if you look back at the history of AI, the more recent history everything has been around for 2530 years in some shape or form. But over the last decade, we’ve seen an evolution from narrow AI to a variety of deep learning, you know models, there were deep learning deep learning was applied to computer vision. Then more recently, we saw Bert, which was the first application in the language for smartphones to be applied to language and text generation. And then, of course, for the last six, seven years, we’ve seen open AI and different flavours of models of open AI. And you know, the host of other players like open AI. As we’ve seen that journey, the number of people developing their own model is now a relatively small subset of the number of people that are using an existing model and building an application on top of it. So there’s still a fair number of companies building models. But you know, if you go back in time, 10 years ago, for every 100 companies, applying machine learning 99, were building a model of some kind, you had to, there was no alternative today, for every 100 companies building a machine learning application. 98 or 99, are just using someone else’s model because you can when it’s good enough, and it does the job. And so when you think about that shift, the the up while the opportunity for model training will continue in it’s a big market, there will be more, you know, there will be companies like open AI, there will be companies like data bricks that provide a lot of infrastructure, I was very fortunate to be a seed investor in data bricks a decade ago. And they’ve definitely built a very large business and model training. And in enabling companies to build their own models. I think what players like Databricks any skin and others are recognising is that the vast majority of customers will start with an existing model and create compound systems. But what they will do in those compound systems, in addition to traditional software engineering, is have lots of inference. And so inference has become really a larger part of sort of the the AI ecosystem. And the challenge is that inference, you know, is, is becoming complex, the future will be multimodal, multi model, multi cloud, and multi chip. So the future that’s multimodal multi model, multi cloud and multi chip, the inference track becomes very complex, because as a and then when you when you think about, you know, an entrepreneur, and it’s building a compound system, for some features, performance, and latency will really matter. And the trade offs that you will make relative to cost will be very different from features where performance and latency definitely matter. In some cases, you’re going to use third party reasoning. And you’re going to take an off the shelf model for reasoning, because, you know, why do you you know, why would you believe that you can do better reasoning, on the other hand, for something else in that compound model system, you’re going to need rag, and therefore, then, you know, the models change, just because of that, it’s somewhere along that chain, you’re going to choose to have a fine tuned model. Some of these models will exist on the edge. And in some cases, you know, you’re going to use a model that’s being served from the cloud. So I think that complexity is definitely a both a challenge for entrepreneurs building AI applications, but also an opportunity for infrastructure companies. Yeah,
18:34
I want to talk about these model layer companies for a moment, because there’s a lot of debate about source versus open source, these closed source model companies like open AI, cohere, etc. Like, can they create durable revenue over time, they’ve been incredibly capital consumptive thus far. And obviously, open source poses a real threat to them. How do you think about the future of these model companies closed source in particular, in the face of competition from open source? Do you think that they will be able to create durable revenue over time? Do you think that they will be able to hold up the large valuations have been pegged to them? Or do you think that the models will become so commoditized and that open source will be such a fierce competitor, that they will find it difficult to maintain the enterprise value that they’ve been assigned today?
19:28
So I think the future will be hybrid. And what I mean by that is, I think there will be a role for proprietary models, because the proprietary models tend to be one generation ahead of open source. What we’re seeing is that, you know, if it takes open ai $100 million, and I’m just making the numbers up to train a model, the fast follower is able to do that for 20 million for a fraction of the cost because so much of the learnings disappear Through the ecosystem with even in a proprietary model. And so the most cutting edge models in the near future, I think, will continue to be proprietary, whether it’s from Google, or Microsoft, or open AI or elsewhere. And at the same time, the n minus one models, you know, will often be open source, or there’ll be many, many alternatives. And open source models will inherently give companies much more flexibility and being able to sort of tweak the weights, being able to sort of do further customization and fine tuning. And so I think there will be many, many scenarios where you will prefer to use an open source model, in forget about the philosophical debate, but where an open source model makes more sense, because even though it’s one generation behind, or half a generation behind, it has so much else, so many other advantages. So I think that’s one thing, the future will be hybrid. I think the second is, I do think that a lot of the companies that are model building proprietary models will die. You know, our our belief at Foundation, and my belief, is that being in the proprietary model business, is is, is going to be very challenging. For anyone that venture funding. We already saw what happened and inflection point is an amazing product. I love that product. I prefer to use probably over everything else. But still, like they couldn’t, you know, they couldn’t make it work. I think, you know, we talked about opening is very unique relationship with Microsoft. But you know, there’s a lot of press in the last week about Satya saying, Hey, we have the IP we own the modelling with or without open AI will be totally fine. And then of course, they have a partnership with Mr. Oil, and they acquired an inflection. And so you’re obviously building and they have a lot of in house research going on as well. So I do believe that it is going to be very challenging times for the vast majority of venture funded mortal companies. I wouldn’t want to be an investor one of those. And that doesn’t mean that, you know, that all fail. But I think the vast majority of them will struggle. Yeah.
22:19
And one of the things that stuck out to me, I was listening to this somewhere, the switching cost of these models is so low among CIOs who are evaluating these models, where they’re using them in their testing environments, and then ultimately running cost analysis between all the different model providers, and whatever, wherever they’re able to acquire the cheapest model for the prompts that they’re using and tune it accordingly. That’s the model that they’re going with. So the differentiation seems very low across these model providers, aren’t they all may be running up the same hill. So it’ll be interesting to see the way that this unfolds. I. Another thing I’d be curious to your take on is this notion of the b2c models and models that are going to be used and consumer applications differing from those that are used in b2b Because if you’re looking at a legal tech startup, they they don’t need all the parameters, they need the parameters that are localised specifically for the applications within LegalTech. Do you think that the b2b side these b2b models will be hyper fragmented for all these individualised use cases and verticals? Or how do you compare b2b models versus B to C? And just general general thoughts? Because it’s something that’s not often delineated between the two at least from the the outsider looking in? So from someone like yourself who’s deep in this space, really curious to just hear from your perspective, the the challenges that may be posed between the two segments of b2c versus b2b in that dichotomy? So
24:00
look, I think I think there will be a plethora of use case specific markets, and whether those use case specific models are built by third party vendor and then make available to a large number of startups or individual startups, you know, do fine tuning for their own models is fine tuning becomes easier time will tell. I think there will be a handful of critical reasoning models. And we will build compound systems that combine the two. And I think that’s true for both b2b and b2c. Because b2c also is very broad. I mean, today when we think b2c models, we think about sort of, you know, the apps that you’re using, whether that’s pie or you know, or pick your favourite, but there will be you know, b2c models you there, a b2c model is being used in advertising. They’re being models being used in a variety of use cases, they’ll get integrated into all kinds of consumer applications, which will also require customization And similarly, there’ll be models that get integrated a variety of b2b applications. So I don’t think of the cart as b2c versus b2c, I just think that you will have, think of it more as horizontal models that are there for reasoning. And then for use cases, there’ll be use case specific models in again, some which will be, you know, provided as an API by third parties, and many that will be fine tuned by individual, you know, technology companies or users of those models. Yeah.
25:31
You wrote another article too, talking about whether incumbents stand to gain more than startups and where the opportunities may lie. But thinking about the Magnificent Seven in this is a bit of a shift from what we’ve been talking about, which which of the Magnificent Seven not named Vidya do you think is best positioned to succeed with the platform shift that we’re experiencing today?
25:54
I think what Microsoft is doing is remarkable. You know, for a company, the size of Microsoft, the pace of innovation, that we are seeing has has not been seen, I think in the history of large companies. I’m very bullish on on what Sofia is doing there. I’m also very bullish on what Facebook is doing. Those are the two I would pay. But those companies, you know, look, there’s a lot going on. And I think anything is still very early days in the ecosystem. I’m a little less bullish if I had to go to the other end of the spectrum. I’m probably less bullish on Apple and alphabet. I think both have been slow to respond. But you know, again, this is where the first innings of you know, of the game. And I would definitely not count them out at all.
26:48
What Apple’s always sneaky, I feel like we’re they, they usually work on things behind the scenes for a long period of time. And then they release with a big bang, it feels like them owning the the interface, you know, the iPhone, with the consumer, that they’d be poised to be in a position to really leverage that platform, and completely remake Siri the way that I think most would hope that it operates. It’s a very suboptimal product from a user experience perspective. It’ll be interesting to see what happens with Apple, I guess, who do you Who do you think is the most at risk just in terms of the way that their business model is worrying today? Obviously, each of these companies has many different revenue lines. But who which company do you think is?
27:33
The answer there? I think is simple. The clearly alphabet is more most depressed. Yeah. Look, Google has, you know, give or take 100% of its profits come from its search business. In its search business, you know, there’s aspects like video search that, you know, are relatively immune. But, you know, the blue, traditional blue links are likely to be reimagined with AI. So when when its core product is going to be reimagined. I think it’s a real question whether it’s going to be reimagined by Google, or somebody else. In the challenge, always, when you have this attacker incumbent dynamic is one is there’s flux and economic model. And so the incumbents are naturally resistant to changing, you know, their primary source of profit. But there’s also flux and consumer behaviour. And so the incumbent is dealing, I mean, Google has billions of users every day, for them to make a change. Four to 5%. Early Adopters is very hard. Dynamic. And this is, you know, I’ll move away from the big difference in seven to the broader opportunity for startups. But the core dynamic that is so interesting right now, is there is the pace of change of technology. And what AI was enabling is so rapid, that consumer adoption is going to lag for the next three to five years, not three to five months. So the next three to five years, we’re going to see early adopters, you know, the five or 10% that are willing to take, you know, completely reimagine the world will have access to a product that is very different from what the mainstream does. And if you’re any software company, b2b, b2c, with hundreds of millions of users in the case of b2c, and in the case of b2b with 1000s, of customers, even you have to have a product that works good for the mainstream. You can’t give it your product for the early adopters, because you’ll have an existing mainstream customers. So how do you manage that? How do you manage that dynamic that what you have in market today is the product for the mainstream. But if you don’t have a product for the early adopters, you don’t have a good testing ground. Google historically actually used to be pretty good at doing that by having products that were you know, they Got lots of these alpha beta products that they were testing, in some form to call their Google Labs at some point different things, but never in their core business. And so it will be interesting to see how Google reacts. But equally importantly, I think every software company between 100 million in revenues and a billion in revenues is at risk. It doesn’t leave, you know, the entire sort of hierarchy of software companies is going to be rematch like anyone who has, as I said, 100 million to a billion in revenues, you’d have less than 100 million, you’re a small company in the grand scheme. So banks, you know, my, you know, my hope is that they will react quickly, and they will, they will embrace the future. But it’s harder as you get across off that 100 million dollar mark. And I think the vast majority of those companies will, will that.
30:46
Yeah. I mean, Google is clearly facing an innovators dilemma right now with search, even the gross margin profile on search is changing, given that the revenue per search is going down, and it’s more expensive to run. What would you do if you were running Google? Like, what do you think, from the outside looking at? Do you think, what is the move for them? You
31:07
know, I’m not reading Google. And I don’t I don’t envy Sanders position. I think he has any he’s he stuck between a rock and a hard place in so I have lots of ideas, but I’m sure sitting there have all of those ideas and more, but also have a lot of strategic constraints. Yeah.
31:27
Yeah, absolutely. It’s, I don’t have any that position as well. No, it’s gonna be interesting to track what they do. A shoot if we could feature anyone on the show, who should we interview and what topic would you like to hear them speak about?
31:38
There are so many amazing people. I mean, it depends on what you’re wanting, if you’re looking for investors. You know, I think Quentin and general catalyst was a great guy. I’ve really enjoyed working with him as a fellow board member. And as many great entrepreneurs you know, Asha, actually my namesake, Asha, Gargan, eightfold, who is gatefold is I don’t know if you’ve heard of eight volt is one of the pioneering companies aiops are close to 100 million in revenues. Now there were one of the pioneers in forking Burke, and applying that to so are really taking MLMs and scaling them fight, I think actually would be a great guy to talk to as well, if you’re looking for a more technical more founder on surveillance. Absolutely. You
32:24
You lead their seed round. Is that right?
32:25
I’ve been involved with eight fold since the start. Yes.
32:28
I said, what what book article or video? Would you recommend to listeners, either something in recent memory that you’ve found informative or inspiring? Yeah,
32:37
I read all kinds of things I happen to be reading outlive, mostly because I’m trying to try to focus on my health right now. But there’s there’s incredible stuff, trying to think of books that I would recommend. There’s just so many interesting things. You know, I the, the one I’ve been reading, one of the things that I’ve always enjoyed a lot. Is this book on? Well, the invention of yesterday. I don’t know if you’ve heard of that. That’s a fascinating book. I’ve really enjoyed that. Prisoners of geography is another one. Sorry, prisoners of geography. The other one is the genius, the geography of genius. If I were to pick one, I would say actually, the most fun book to read is the geography of genius. It’s a light read, it’s a really fun read, I would highly recommend that.
33:26
Why is it 100
33:28
The geography of genius really goes into if you go back goes back in time over the last couple of 1000 years, and looks at periods of time and geographies, where we’ve seen, you know, an explosion of talent, an explosion of innovation, not just in technology, in you know, in, in commerce and in, in in socked in music in philosophy. And then it looks like what led to that what were the capitalists who were the key characters and so it’s it’s it’s like seven short stories, seven, eight short stories, which are real this is this is nonfiction. But really almost like fiction. Yeah.
34:09
Interesting. Interesting yet, studying, studying history and finding the patterns that emerge over time is super interesting. The last but not least, what’s the best way for listeners to connect with you and foundation?
34:22
You don’t send us an email. We I we look at our we look at all our emails. I look at all the emails that come to me now it has to be interesting. I don’t reply to 99% of them. So something has to stand out in your email for it to be the one and 100 I would reply to, but I do read every single one. Of course it helps if you know someone in our portfolio, and you there’s a warm introduction, but I don’t think that’s necessarily do you remember the best cold email you’ve ever gotten? No, no, none stand out. But there’s definitely no there’s I mean, there’s no nothing like good or bad best. I mean, there’s there’s cold emails I respond Do something that caught my attention. In the five seconds I spent on it. Yeah, there’s got to be something in the first paragraph. Whether it’s a person’s person, it’s typically the person’s personal story. Yeah, good catches attention, or it’s something about the idea. I think a lot of people have emails that are very abstract. Whereas what, you know, when someone’s spending five seconds and ready decides to respond or not, what you have to do is say one thing that stands out, yeah, that’s worth opening the presentation. You You know, your attachment. And then your attachment has to do one thing that makes it worth responding. And so on and so forth. Yeah, absolutely. Because they say the goal for every meeting is to get the next meeting.
35:43
Exactly. Exactly. Well, I appreciate the appreciate the advice on the cold email. And thank you again, for doing this. This is a real pleasure.
35:51
Thank you for having me on the show, Nate.
36:00
All right, that’ll wrap up today’s interview. If you enjoyed the episode or a previous one, let the guests know about it. Share your thoughts on social or shoot them an email, let them know what particularly resonated with you. I can’t tell you how much I appreciate that some of the smartest folks in venture are willing to take the time and share their insights with us. If you feel the same, a compliment goes a long way. Okay, that’s a wrap for today. Until next time, remember to over prepare, choose carefully and invest confidently thanks so much for listening