Episode 435 of The VentureFizz Podcast features Kojo Osei, Partner at Matrix.
The venture capital industry originated back in 1946 in the Boston area, when The American Research and Development Corporation was founded by MIT president Karl Compton and Harvard professor Georges Doriot.
From that point, an industry was born, and firms were created, but very few have had the lasting impact and track record as Matrix, one of the OG firms that helped institutionalize this industry.
The firm has a tremendous track record through the years making investments in legendary companies like Apple and FedEx… plus category creators like HubSpot, Zendesk, and Oculus… or current investments like Canva, Suno, GOAT, and Flock Safety.
Matrix is investing out of its 12th fund, an $800M fund for seed and Series A investments, and the firm continues to have a strong presence on both coasts.
Kojo is a Stanford grad, who was the Head of Product at Sirona Medical, the AI operating system for radiologists, before becoming a VC. Now based in NYC, Kojo has been an investor with Matrix for over five years. Some areas of interest these days for investments include agent-driven commerce, how the software infrastructure is getting rebuilt for agents, and model application integrated companies. His investments include companies like Channel3, BoldVoice, Ampersand, and LM Studios.
Topics Covered:
- A conversation around agent-driven commerce, a term that Kojo actually coined.
- Growing up in Ghana and what he learned from his entrepreneurial mom.
- Studying at Stanford and his early professional journey.
- What led Kojo down the path of getting into venture capital and what the first year looks like as a VC.
- All the details about Matrix, plus a deeper dive into what Kojo is targeting for investments.
- The art of a cold email to an investor.
- The importance of “why now” when evaluating entrepreneurs.
- Thoughts on GTM strategies for developer products.
- And, so much more!
Podcast Sponsor:
This podcast is brought to you by one of the strongest longtime supporters of the local startup ecosystem, Silicon Valley Bank, a division of First Citizens Bank. With more than 1,500 bankers and relationship advisors and $44B in loans as of Q4 2025 – SVB delivers expert guidance, specialized products and a team that knows the innovation economy inside and out. Learn more at SVB.com.
Transcript:
Here is the cleaned-up transcript with repetitive words, extra spaces, and the ⁓ symbols removed for better readability:
Keith Cline (02:44)
Kojo, thanks so much for joining us.
Kojo Osei (02:46)
Glad to be here.
Keith Cline (02:49)
I’m excited to talk to you because you’re a partner at Matrix. And you know, Matrix has this history, one of the anchor VC firms over multiple generations of technology. And I’ve always appreciated that Matrix has had a presence in not only Boston, New York, and San Francisco, but they look at opportunities across the whole spectrum. And we’re gonna talk a lot about the VC industry and Matrix. But as I was doing research on you, I found out that you actually coined a term, which I think is very notable. It’s almost like an honor when you actually create a term that people start to leverage. And that’s agent-driven commerce. So what does that actually mean? And how does that happen? How does one get coined like that phrase becomes something you originated?
Kojo Osei (03:43)
Yeah. About two years ago I wrote a blog post, actually a series of blog posts on agent-driven commerce, and the kind of high-level idea was that if you broke down the journey of buying anything—software, consumer, clothing, whatever—there is a huge chunk of that journey that was currently constrained by time, right? You as an individual have to go and do this huge discovery and search process. And as a result, the transaction cost of finding things to buy was really high. However, agents are, you know, compute-bound. And so you could basically trade this bottleneck in the commerce process, which is time-bound, for something that’s compute-bound, and that would lead to some pretty interesting second-order effects. It took me basically two years to find an investment that fit that thesis, as all things in venture go, and yeah.
Matrix and my personal investing style tends to be very thesis-driven. But what you’re referring to is probably that series of blog posts, and folks picked up on it. Two years later, I found an investment that I was excited about to kind of consummate a thesis, and I continue to be excited about it and to look for investments in it.
Keith Cline (05:09)
So what do you think are the steps where this becomes kind of like a consumer reality, like mass adoption?
Kojo Osei (05:16)
Yeah. So step one is in the discovery process. Whatever you buy, there’s some amount of discovery that you do to purchase that. I think presently there’s a lot of passive discovery. If you think about what these recommendation algorithms on social media are, you’re scrolling on TikTok, you’re scrolling on Instagram, and you get recommended an ad. It’s a passive discovery process. If you wanted to do an active discovery process—like, I’m in the market for a new backpack, just a work backpack—I have to go out and read Wirecutter, read a bunch of stuff, read a bunch of Reddit threads, and I’m just kind of time-bound, right? I don’t have as much time to do that. What would be great is if an agent that has some context on me and some context of what I’m looking for could take over the discovery process. So the discovery process is the first step. I think we’re actually a little bit closer to that. It’s early days, but it’s happening, right? So that’s the discovery process, sort of step one.
The next step in that process is what I would call the sort of filtering or monitoring process. So there are things like monitoring price changes in whatever it is that you want to buy so you can buy at the exact right price that fits your budget. And then kind of a final step is actually consummating a transaction. So giving an agent full autonomy to actually make a purchase for you—that has not happened yet. I think there are a couple of pieces that need to still click. Stripe has its agent SDK where you can kind of preload a virtual credit card. Companies like privacy.com have a way for you to preload a virtual credit card to give to an agent to buy. There are a lot of interesting ideas around using crypto for this type of transaction. I don’t quite think we’ve settled on a kind of economical way to do this, and it’s still a bit of the Wild West, but I think that’s where things are headed.
I will add that this is also not just for consumer goods. Think about software, or I always joke that product-led growth—which is kind of like the 1.0 beginning of SaaS ways that people would market—is now becoming agent-led growth, right? If you’re using Claude Code or any of these AI coding tools, sometimes the agent will suggest a library for you to use. There’s a company called AMP that famously said, “We will offer a free tier of a coding agent system on one condition: that you accept ads inside of your coding agent.” So your coding agent can now say, “Hey coder, I think you should use this library instead of that library, and buy this subscription instead of that subscription for your project.” So this parallel track that’s happening in the consumer world, I think will also happen in the enterprise world, where agents will also start to recommend specific software tools that you should be using. And for developers, specific libraries and tools that developers should be using.
Keith Cline (08:32)
Interesting. Yeah, so it’s not just consumer, it’s B2B. I want to talk about your background, but what was the investment that you made in this space?
Kojo Osei (08:41)
A company called Channel3. Channel3 is tackling that kind of first step, which is the discovery process. It was quite surprising to me as I dug into this space that there was no canonical way for agents to query for a catalog on the internet of products. Channel3 basically is a product API for your agent to just query for any product on the internet. It’s the most up-to-date, the most fresh data source that your agent can connect to to ask, “Hey, Kojo is looking for a new backpack. Can you find me backpacks that match the following criteria?” And then of course, the business is going to build the other parts of the journey that I talked about.
Keith Cline (09:29)
Fascinating. So it’s gonna be in the future, people will be like, “Remember when we used to have to do research on Google to find things and go and buy things on Amazon?” where it’s just all gonna just happen and it’ll be instantaneous. So it’s an exciting future ahead. We’ll see how it all unfolds. All right, well, let’s talk about your background story. So, where did you grow up? What were you like as a child?
Kojo Osei (09:46)
Exactly. I was born in Ghana, in West Africa, and ended up in the States for college. My mom was an entrepreneur and so I sort of grew up around the idea that entrepreneurship is a very aspirational career and it is also a way to kind of create self-sufficiency and improved economic outcomes, not just of individuals but of entire communities. And so, I watched my mom build a business, I watched her wake up every morning and go out and try to make something, and it really had an impression on me. I ended up going to Stanford for college and after college, did a couple of things—had a stint at various internships in finance and so on and so forth.
But I began my own journey on the founding team of a business called Sirona that was building automated diagnostics in the medical imaging space. And that was sort of my first real full-contact sport with entrepreneurship. It was a venture-backed company, at the time pre-generative AI. And so we were building bespoke image detection and object segmentation models to help radiologists diagnose medical images.
Keith Cline (11:23)
How did you get involved with that company?
Kojo Osei (11:25)
So I had studied math and computer science in undergrad and then bioinformatics in grad school, and I had actually taken a class with one of the foremost thinkers at the time on applying object detection and image segmentation to radiology. Through that, this professor was on the founding team of the business, and through that kind of felt I could contribute my skills towards a venture. It was sort of a classic Stanford founding team story.
Keith Cline (12:03)
And it was also creating the whole like radiology operating system, is how it was phrased from what I gathered, and you started as an engineer originally, but ended up leading product.
Kojo Osei (12:09)
Exactly, exactly. As all startups go, there’s so much work to be done and the amount of work when you’re a small team cuts across engineering and product. In some sense, I think I was early to the realization that engineering and product should be far more tightly coupled. I think today lots of teams would agree with that in the way that you kind of staff and build an org. But in those early days, we were all doing everything. I was writing code, I was talking to customers, eventually kind of more formalized head of product, the product org and so on and so forth. But it was a huge learning process for me to do both engineering and product.
Keith Cline (13:03)
Well, just like the team, the founding team, very uniquely qualified to work on this problem, right? Yeah, just very, very extraordinary. And the company has since gone on to raise, I think, over a hundred million in funding.
Kojo Osei (13:08)
Yeah, exactly. Yes, exactly.
Keith Cline (13:19)
So what led you down the path of getting into venture capital?
Kojo Osei (13:23)
I always say I somewhat got into Venture on accident, and I think specifically Venture and Matrix felt like a very different opportunity than Venture anywhere else. I had met the Matrix team in the context of their talent development initiative. So if you think about Matrix, it’s an investment firm, obviously, but a significant amount of time is spent on talent development, network development, and community building. I met the firm in the context of that and it really stood out in the way they interacted with young talent at the time. The care and thought they put into learning about folks, understanding people’s motivations—I found it quite aspirational.
And when I was ready to kind of think about the next thing, I also learned that most of the people on Matrix had a similar background in terms of being a founder or being early at a startup, or kind of starting from the operating side and then transitioning into venture. And so it felt very mission-aligned, so to speak, and the folks were great.
Keith Cline (14:37)
So a quick side question. For people who are exploring opportunities to join a venture capital firm, what advice would you give somebody on that career path?
Kojo Osei (14:51)
Yeah. I would say, number one, every venture firm is very different. I think there’s kind of the meme of all venture capitalists not being differentiated on the outside. But I think internally, people really have different opinions about the world. People practice venture capital very differently, and it really comes out in the way they engage with talent in the community. And so step one, I would say, is try to get to know venture capitalists as early as you can in the process of figuring out what’s next. I think that will give you so much information about what each firm is like because each firm is just so different. The recruiting process is different. The folks are different. The way they practice venture capital is different. I find it surprising that the term venture capital in my mind is too broad a term to describe what’s going on in the ecosystem today. And I’m sure we’ll get to that at some point. So step one, try to get to know folks and the people behind the firms, the people inside the firms, as early as possible.
Step two is that venture capital firms are a function of the opinions and networks of the people inside of them. So try to have something that you believe in that is additive to the existing firm, right? It could be a thesis of what you’re currently doing, it could be something you believe about the future that you have high conviction in. I find that operators, especially if you come from an operating background, are probably in the weeds of something that’s not as obvious to other people that are not in that industry or that category. And that can be an asset for operators that want to get into venture capital.
And then the last thing is I always kind of do the “try before you buy.” You know, if you can angel invest, great. If you can’t, try to have a kind of phantom portfolio that you follow, that you’ve written about, that you’ve demonstrated—maybe online on a blog or something—just so you can get a sense of whether or not it’s the kind of job you’d enjoy doing. And also so people can see the depth of your thinking when you’re going through that process.
Keith Cline (17:05)
And that’s great advice because I think there are people out there like, “Should I be an operator, a founder, or an investor?” So when you start in the role as a venture capitalist, what does the first year look like?
Kojo Osei (17:17)
Yeah. My guess, just based on exposure to other folks, is that it really differs. I will tell you what my experience was. I think in that first year there is a lot of learning, and learning happens by osmosis in venture capital. It is such a relational job and the number of shots on goal are so small that the only way to learn is really by shadowing folks who’ve been doing it for a while. So I got to shadow some of the amazing folks at Matrix and that was a huge part of my first year.
After that, again, you probably came in with some thesis or some opinion or some network that you’ve been cultivating and working on. You continue to develop that, then you start to make investments, and then you kind of start to build your own portfolio of investments inside of a firm. But I would encourage anyone at any firm to really just take advantage and take their time in shadowing people who’ve been doing the job for much longer than you have, because it is such a learn-by-seeing and learn-by-doing job, and there’s no way around the apprenticeship model in my mind.
Keith Cline (18:41)
And most VC firms, not all, but most, operate a pretty lean organization, right? Like there’s partners and maybe associate, senior associate, principal, whatever their different levels are. And then there’s maybe an operating team. But for the most part, they’re pretty lean organizations. So all the advice you’re giving is dead on. When you’re joining a VC firm, it’s not like joining a company where we’ve got three rounds of interviews and we’re gonna offer you a job. It’s almost like a marriage, right? You’re probably gonna get to know all the partners really well before that position is offered. And by then, you probably know whether or not you wanna do it, if it’s the right chemistry between the partners.
Kojo Osei (19:14)
Exactly. Exactly. And I will say it’s also quite similar to the process of finding, winning, and leading an investment, right? You get to know a founder, you figure out if a thesis is aligned, you figure out if you can actually help this founder, you demonstrate that you can actually help this founder. It’s very much like dating when it comes to even meeting founders and finding an investment that you are passionate about and have high conviction in. So the process of getting into venture capital mirrors the process of getting into a deal.
Keith Cline (19:58)
Yeah, absolutely. That’s a great analogy. Okay, I’m gonna talk about the history of venture capital for a second here. There is a road in Waltham, Massachusetts called Winter Street. And this road back in like the nineties was where all the VCs were. I actually think it might have been called Mount Money, I think—don’t quote me on that—but this is where a lot of the VCs originated. And these are blue-chip VCs that still exist today. It’s like Greylock, CRV (which was called Charles River Ventures), and Matrix. And what I was trying to talk about earlier as far as alluding to the fact that Matrix still has a strong presence in Cambridge, you’re in New York, San Francisco. A lot of those VC firms that I mentioned, they just moved their operations out west, but they originated in Massachusetts or Waltham.
And they were the OGs of the internet 1.0 era and still continue today to do extraordinary work as investors. Matrix being one of those firms, right? So I just want to give the kind of history of venture capital because Matrix is one of these pillar firms that if you go to your website, it’s like, “invested in Apple,” right? Like, wow, that’s insane, right? But then each technological shift—HubSpot, Oculus (which was acquired by Meta), one of my favorite investments is The Echo Nest (which was acquired by Spotify)—it came out of the MIT Media Lab. And every time I use Spotify, I think of Boston, MIT, and The Echo Nest because of the personalization that originated for Spotify from that acquisition. GOAT, Zendesk, and then obviously current investments, right? Like Suno just announced a $400 million Series D round of funding. So there’s so much behind Matrix. I just felt like I needed to talk a little bit about the history of venture capital, but how Matrix has continued to keep kind of that bi-coastal relationship.
Kojo Osei (22:05)
And I think one of the things that makes the firm unique is that the folks here are quite adaptable and humble about participating in each technological shift, right? There’s not as much dogma—you know, like we don’t say we only do hard tech, or we only do fintech, or we only do… there’s not as much dogma. I think folks are very open-minded. We will go where the founders are. We think there are great founders on both coasts, and we just remain very open-minded about where great ideas and great companies come from.
Keith Cline (22:43)
Now what is the typical stage of investing for Matrix?
Kojo Osei (22:46)
We will invest anywhere from first check to Series A. I, for example, have done several first-check deals and have done Series A’s. That’s very typical for the fund.
Keith Cline (22:58)
And what are you concentrating on these days as far as the investments that you’re making? We talked about the commerce aspect, but what else are you looking at?
Kojo Osei (23:07)
One fascinating thing that’s happening is that the entire kind of software infrastructure stack is getting rebuilt for agents. And so I have a company called Ampersand that builds the integrations layer for agents. Previously, the user of an integrations tool was a human—it was a human developer. They go, they read the docs, they try to figure out the API schema and data schema for some version of Salesforce, and then they implement it. All of that now is shifting to agents, right? Your agents write integrations, your agents maintain your integrations pipeline. And so that general theme of software infrastructure being built for agents—every primitive of software infrastructure being built for agents—has been a big focus of mine, Ampersand being one of those companies.
A second focus is what I would call model application integrated companies. At the application layer, I have a company called BoldVoice that builds a voice coach, and it’s integrated with their own model stack. They built their own model from the ground up, they have the largest dataset of phoneme-level transcriptions and related audio, and they serve their end user with this amazing product that is only made possible because they are integrated with their own model. In some sense, you could think of Suno the same way, right? The model and the application are integrated. In some sense you could say the model is the product. And so I’ve focused a lot on companies that are able to—in other modalities out of text and LLMs, this is a little bit easier—control their own future and control their own destiny by having some piece of their intelligence and their model intelligence controlled in-house. Those are two of the biggest theme focus areas for me right now.
Keith Cline (25:01)
And I was looking at your blog, and so you started with Matrix in 2021 and your writing in 2022 was definitely where we are today, right? It was definitely dead on as far as how software engineering was evolving and how AI, machine learning, and LLMs are evolving and going to affect the infrastructure and the tech stack. So it seemed like you were thinking about this before the moments started to happen.
Kojo Osei (25:33)
That’s a big feature of working at Matrix—most of our conversations internally are about where the world is heading. And most of our conversations with just people in our network or founders really focus on that. If you think about the number of investments we make, each person here is probably doing one or two investments a year. That’s not a lot of investments, but a lot of the time is spent getting to conviction. It’s spent getting to those one or two deals that you’re going to do a year. And writing for me has been a way to really explore various theses. It’s been a way to crystallize my own thinking, and it’s been a way to kind of put out an opinion in the world—and of course, change your opinion if it turns out to be wrong—but to put out an opinion in the world and to track how that opinion evolves over time. My blog has been kind of the public demonstration of that, but internally, a lot of those conversations are happening and mimic things that you’ve probably seen on my blog.
Keith Cline (26:44)
All right. So you talked about conviction and a low volume of investments per year per partner. So how does a founder get on your radar and how are you looking to kick things off, like that first meeting, first conversations? What do you envision that process looking like?
Kojo Osei (27:03)
Yeah. One of the things that I really encourage founders to do is to just come inbound. I personally put my email out on the internet and anybody can email me. It’s kind of a magic of modern technology that you can just email anybody, and I think when I was younger I didn’t quite appreciate how useful that was, but I really encourage founders to just email folks. People really do read your emails, especially when it’s related to something that they’ve written about, related to something that they’ve said online, or something that resonates based on your profile.
Keith Cline (27:46)
Can I dig into that a little bit? Cause the cold email I think is an art, right? If you’re not getting a warm intro through somebody that you know mutually, which is always like the preferred—well, they usually like to have a warm intro—if it’s a cold email, what should they say? How do they keep it short and sweet? What do you recommend that’s gonna give you the mindshare to engage with that email?
Kojo Osei (27:59)
I think the first thing is finding why this person—why am I the right person to take a look at this deal? And I think it comes down to resonating with something that I have either written about, an investment I’ve done, or something in my past. I think that is kind of the easiest anchor for a cold email, and I think it’s quite important to establish that resonance early in a cold email.
And I think the second thing is just making it easy for people to process what you’re doing. There are several examples of great cold email templates and so on on the internet, but I think for me, the less the better. Just making it easy for people to grok what you’re doing in one paragraph, two paragraphs max. If you pair that with a resonance of why this investor is the right person to look at this deal, I think you probably end up in a very good spot.
Keith Cline (29:17)
And I’ve never raised venture capital, so I’m not the authority, but I have interviewed a lot of venture capitalists, so I do have somewhat of an authority based on hearing it multiple times. So my recommendation is make sure you’re reaching out to the right VC because they do have an area of focus. Make sure they’re the right stage that you’re looking to target as far as the type of investments that they make—seed, series A, pre-seed, whatever. Write the email. Do not use AI and generate AI slop because they’re gonna notice it right away. Everybody recognizes the AI slop emails now. If you’re gonna reach out to that person and take the time, and you’re hoping that person’s gonna take the opposite time of reading it, make it authentic, right? And shorter, I think in today’s world, is better. So there are the Keith Cline tips on how to get cold emails into your inbox and read. So okay, the first steps. We email, you’re like, “Cool, let’s meet.” So what are you hoping happens next?
Kojo Osei (30:18)
Even before we get into the business that founders are building, I like to spend time understanding the motivations behind why they’re doing what they do. I think a lot of people skip that part. And frankly, VCs can tell if the motivations are somewhat shallow. Building a business is a 10-year journey minimum. If you get lucky and it goes faster, that’s fine, but the fuel for continuing when things get tough is that internal drive and that internal motivation. And I like to start every meeting with some discussion about that internal drive and motivation. It is, in my opinion, almost equal weighting to what you’re building—the why of why you’re building this, why go down this path. It is an extremely difficult and challenging journey, so why do it? I like to spend some time getting at that. And then naturally you get into what is the business, what are people building, and so on and so forth.
I always try to spend a lot of my time getting at what is the one thing or the two things that this founder knows and understands about a market, or about a technology, or about whatever they’re building that is not obvious to everybody else—or at least that’s not obvious to a lot of people. And I think being able to articulate those two things—why are you doing what you’re doing? What is your personal motivation? And what is some non-obvious insight that is driving you, that’s giving you, the founder, conviction on the business?
I think a lot of founders kind of think that the VC has to get to conviction in the idea, but I would actually reframe that as: venture capitalists need to get to conviction on how you got to conviction. You, the founder, are the one that’s gonna be building the business, right? If the process of getting to conviction for yourself is shallow or doesn’t come out clearly, it’s going to be quite difficult for a venture capitalist to believe in you or believe in the idea.
Keith Cline (32:36)
I mean, that goes back to when you started your career with Sirona Medical. I highlighted the founders, right? It’s like they have this background that makes them uniquely qualified to tackle the problem they’re solving, which speaks to why they care and why they have conviction. Again, I’m not a venture capitalist, but if I was, I would follow the same path where it’s like, did you just randomly stumble on this idea and somehow became passionate about it because you were at a whiteboard? Or did you actually live this problem and not stop thinking about it because it was something you had to solve?
Kojo Osei (33:09)
Yeah. And there are some whiteboard ideas that do end up working, but I would argue that they work because the founder at some point had to convince themselves and have a level of irrational conviction in the idea—you just cannot shortcut that. Maybe the process of finding the idea was somewhat random, but the process of getting to conviction in the idea, I think, is almost akin to a spiritual journey. And I think you can tell in the meeting when the founders have really gotten there versus when they have not.
Keith Cline (34:17)
What about—I think you wrote about this either on your blog or on LinkedIn—there was this like “why now” philosophy, and the importance of history and failed ideas in “why now,” right? So what does that mean?
Kojo Osei (34:33)
Yeah. I love to spend some time with founders on everyone that has come before them, and what those people have done in a category or in a space, and just try to understand why do they think those people failed, right? It is rare to see an idea that nobody else has thought of before. Maybe there are some ideas like that, but you know, they’re few and far between. So it’s quite likely that someone else has thought of this idea, someone else has tried, some other venture capitalist somewhere has backed someone to go do this before.
And I think the best founders have some opinion, some intuition on why others have tried and failed. Where did they go wrong? And that gets at the why now. Sometimes the technology is just not there. Sometimes the market’s just not there. Sometimes the incentives of the business model are not there. And I find it quite refreshing when founders are able to articulate which of these combos or some other thing is making the opportunity exist today. And this “why now” conversation is one of my favorites to have with founders.
Keith Cline (35:52)
So I love Lenny’s podcast and he had Tony Fadell as his most recent guest. I was listening to it this morning and he talks about the iPhone and the success of the iPhone, which couldn’t have happened if it wasn’t for the iPod. And even going back, I didn’t know about his history of a company called General Magic that was working on non-keyboard interfaces with touch, but the technology wasn’t there yet for what they were building. So all these things wouldn’t have happened unless finally the technology was ready to do what the iPhone was ready to do with Wi-Fi or 3G adoption. Like, we take things for granted that Apple had this vision, but there were so many steps for them to get there that had to happen technologically to eventually—and in his experience, uniquely qualified—it just like I think of that as a prime example too.
Kojo Osei (36:45)
Totally, I agree. And that’s true for—well, I’ll use the current moment that we’re in, right? You know, folks have been writing about language models for years. What is the why now? The availability of compute and data, the fact that you had the first common crawl datasets, the fact that you had this huge corpus of the internet of humans just writing stuff and creating this huge corpus. Coinciding with accelerated compute that was built for what? For gaming and for simulations and for crypto. Everyone forgets the availability of compute is really a function of crypto and gaming and kind of all these things that came before it that made this advancement possible. People have been talking about this for as long as I’ve been alive, and I think the best founders are able to intuit what is the thing that is happening today that will unlock the next huge inflection point for their business? What is the thing that happened in the past that they can ride to unlock a huge inflection point for their business?
Keith Cline (38:00)
Okay, so we’re in some timeframe. What inning of AI are we in? Like, you know, you mentioned there there’s history, right? Marvin Minsky was working on AI at MIT in the 50s, right? So it’s like AI isn’t necessarily new, but obviously the technology of what we’re able to do with it is very real in today’s modern era. So what are your future predictions and where do you think we are in terms of the innings of AI?
Kojo Osei (38:30)
I think we’re quite early. I think there are a few buckets. Let’s start with just what happens to foundation model companies and what happens to the availability of intelligence. I have a lot of friends that are not in tech—they work in more traditional industries: law, business, finance, consulting, whatever. And when I juxtapose my experience working with founders every day with their experience, it’s very clear that AI adoption is just so early. It’s just a fraction of what it could be. So that’s number one.
Number two is, where do people get their intelligence from today? People get their frontier intelligence from a handful of providers, right? You get your frontier intelligence from Anthropic, or OpenAI, or maybe Gemini. And the current assumption is that that is just the future, and I think that could not be more wrong. I think that the inference market will likely become fractured into multiple markets. I think number one, you have the personal inference market. I have a company called LM Studio that does work on quantizing and running models on your MacBook or on your iPhone to help you get intelligence. I think that will be a huge market. There’ll be the enterprise market, which will be some combination of frontier models and open-source models and in-house models with a router that’s probably routing queries out to the right model. And then it would be perhaps what I would call the industrial inference market, which is you are trying to solve some crazy Earth problem, and you need the most frontier model with the largest amount of compute. So I think this current state is a temporary state and the inference market will fracture into at least three gradations, and users will get their intelligence from a wide variety of places. So that has a lot of implications for kind of talent density today, it has a lot of implications for where talent density is going to go in the future, and has a lot of implications for venture capitalists and how they fund businesses. So that’s kind of the second thing.
And then the last bucket that I think a lot about where we’re headed is the implications for society, for productivity, and for the labor force. I think that a lot of the assumptions around AI or models being substitutive with human labor is based on perhaps an underappreciation of the relational nature of work. I think quite a lot of work is relational and not particularly analytical. And I think people, if you’re in tech, just underappreciate that, right? If you’re an engineer, you go to work and you write code and you get paid for writing code and you’re like, “My God, the models can write code now. So obviously everyone else’s job is as substitutive as my job.” I think a lot of our economy is far more relational than people appreciate, especially if you’re in a technical role. So I think the current assumption of the jobs apocalypse and this kind of widespread 10% unemployment fear is frankly wrong and overblown. And so I think we’re in this moment we’re just early, the inference market will change, it’ll fracture, and there will be a lot of opportunities for people to do all sorts of new jobs that we just frankly have not thought about yet.
Keith Cline (42:29)
Absolutely. Yeah, the labor market’s gonna be really interesting to see how it all unfolds. So you’ve been a venture capitalist for five years now. What are the common things that founders do that prohibit their success? Like the common mistakes that they’re making?
Kojo Osei (42:48)
That’s a great question. I think it comes down to—you know, there are like tactical things, right? Hiring, or product, go-to-market, and so on and so forth. I actually think all those things pale in comparison to what I would call self-awareness, and just looking at yourself in the mirror and being honest about what’s going on in the business. I think that the most important trait in founders that I found correlates with success is they’re just brutally honest about what’s happening to their business and what’s happening to themselves, because that’s the first step to changing anything. If you can’t admit that something is wrong or you’re not performing at a certain level, it’s just really tough to make any changes. And in some sense, I think of my job as a venture capitalist when I work with founders as being a bit of a mirror to just mirror back what’s happening in the business so they can actually see for themselves and come to the right conclusions. I think it’s one of the things that, frankly, is one of the few things that venture capitalists can help with outside of giving you money—it’s just helping you see the realities and coming to the truth of what’s going on in the business as opposed to kind of going on this journey blind and without the right level of self-awareness.
Keith Cline (44:23)
Well, I feel like there’s a lot of products being built for the developer community. So how does a founder actually go to market with those products that drive adoption? It seems very crowded, and then obviously Anthropic and OpenAI are coming out with tools fast and furious. So how does one build a company and drive that adoption?
Kojo Osei (44:46)
Yeah, I think the developer market has changed significantly. I think step one is you really should think about the agent being your user, as opposed to two or three years ago where there was a room full of developers that were using your library or your tool or whatever. That’s step one. I think that is one way to unlock growth—agent-led growth—if you’re building a tool for agents to be able to implement easily and leverage easily. Maybe the second thing I will say is there are emergent second-order consequences of agents writing code that we still don’t quite understand. And I have found that the most interesting ideas I’ve heard recently really lean into that. They assume that a significant amount of code will be written by agents, and then they ask the question, “Then what? What do I need to build? What do I need to do to have a market in this space?”
Keith Cline (45:57)
So we’re rolling off of Tech Week, and it was New York last week, Boston the week prior. To focus on Boston—I know you’re kind of location-agnostic as far as the investments that you make—but what observations did you have about Boston?
Kojo Osei (46:15)
I think Boston still has a phenomenal supply of talent coming out of MIT, Harvard, and various colleges in Boston. I do think that talent is quite mobile and tends to vote with their feet to other cities. However, they originate from a lot of these colleges. And so I find it refreshing to meet early talent in their journey and be of help. Some of our best companies have come from Boston. It is a great place to build a business. But I do think that the kind of tiering of cities—right? You’ve got San Francisco, New York, Boston, LA, I would put in that bucket as well—I think it’s competitive. And if you’re a city or if you’re in local government, I think you really ought to think about why would young people choose to live here and build their businesses here? Why would companies choose to hire here? And, you know, if I was local government in any of these cities, including San Francisco or New York, I would be thinking about this.
Keith Cline (47:28)
I mean, there’s a resurgence in Boston, which is good and I think needed, with like the Mass AI Coalition, and then if I can point to a sticker right here, the TNT with the accelerator of what they’re building for Harvard and MIT startups is awesome. Rob Blaine is just doing a phenomenal job and we need more of this activity in Boston. Y Combinator originated in Boston and Cambridge back in the day, so there you go. I think there’s just been—you know, we had Techstars, but even Techstars is not operating in Boston right now. I hope they do come back. But we need more of that type of presence and building blocks for these founders to take advantage of. So hopefully we’re on the right track again. So how are you leveraging AI in your day-to-day workflow?
Kojo Osei (48:11)
Exactly, exactly. I use various tools, as well as some homegrown tools. I read and write a lot. I use—this is gonna be kind of an old-school tool, but I’ve been quite impressed with how they have embedded AI into it—ReadWise, which is the kind of clipper of various articles. I read a lot of papers, I read a lot of articles, and it’s just amazing to see how well they’ve embedded AI. I do a lot of research and so I love the deep research features in ChatGPT and Gemini especially—they have great deep research features. And then the second thing is what I would call like a custom CRM, which is a tool we call Radar internally. It’s a custom CRM, very customizable, allows us to keep in touch with our network and do this kind of talent discovery and talent engagement, which is a huge, huge part of our day-to-day jobs.
Keith Cline (49:17)
So that’s a custom app that was built for—wow, very cool. Yeah. Okay, so I think you already mentioned maybe one of these, but I was gonna ask you: three apps you can’t live without, and it can’t be Slack, email, or calendar.
Kojo Osei (49:19)
Yep, a custom map, from scratch. Yeah, exactly. I cannot live without ReadWise because I read so much. And I actually wish there was a kind of physical-world ReadWise tool that I could carry with me for the physical books that I read. Another app that I would say is LM Studio. I use LM Studio every day. I think I hit my limits on Claude and ChatGPT almost every day, and so it’s just great to have unlimited inference. And if you think about intelligence saturation, right? There are certain tasks for which there is saturation with an open-source model and you don’t need a frontier model for it. So I open LM Studio pretty much every day. The last app I would say is an app called Obsidian, which is in the same vein of reading and writing—it’s kind of a networked thought tool. So as you write, you can connect ideas to various ideas. It’s kind of like Apple Notes plus a connecting ideas tool, and I’m in it every day as well.
Keith Cline (50:41)
Well, usually I ask for a podcast or book recommendation for entrepreneurs, but I wanna tweak it a little bit. How are you staying current with all that’s happening in AI or the developer community? Like, are you on X mainly? I know a lot of that content has moved back to X from what I’ve gathered, or Reddit—where are you finding your information to keep up to date?
Kojo Osei (51:03)
Yeah. I really like primary sources and so I read a lot of papers. And I keep track of—you know, I love RSS feeds. I’m a bit old school in that sense. It’s amazing. It’s so easy. It just works.
Keith Cline (51:14)
Me too, bring back the RSS feed. It was so easy, it worked.
Kojo Osei (51:24)
Exactly. I wish someone actually would kind of rethink social networks from the perspective of RSS feeds. But I love RSS feeds. I have a list of RSS feeds that I follow and they’re all tracked in ReadWise, and so whenever someone publishes something, I just read it. I have a bunch of primary source RSS feeds—these are researchers, these are software engineers that I think have great taste, write well, and are on top of things. I read the primary sources with papers.
I also read a lot of public markets commentary. I think that investing in private markets today—you cannot invest in private markets today without appreciating what’s happening in the public markets. And I also think the opposite is true—I think you cannot invest in public markets without appreciating what’s happening in the private markets. So there are a number of public market writers and commentators that I read often, and a handful of companies whose earnings calls and CEO letters I think are quite important to pay attention to. And I actually encourage founders, if you’re a founder and you’re in some industry or category, just pick a handful of CEOs that you respect and admire and just follow their earnings commentary, follow their CEO letters. You learn a lot from it.
Keith Cline (52:34)
Great advice. What do you like to do for fun outside of work?
Kojo Osei (52:39)
I play a lot of tennis these days. It’s a great sport. It’s one of those sports where the mental game is just as taxing as the physical game. And I play guitar. And so if I’m not working, I’m playing tennis with friends or jamming out with some friends.
Keith Cline (52:59)
Are you playing pickleball too, or no?
Kojo Osei (53:02)
I have tried pickleball. I think pickleball is also good—it’s not tennis, but it’s good.
Keith Cline (53:07)
Right, yeah, that’s fun. I agree. I used to play a lot of tennis. I don’t as much anymore, but I play the guitar and the occasional pickleball. Well, Kojo, thanks so much for taking the time to walk us through your background story, obviously all the great work you’re doing as an investor at Matrix, and obviously all the great advice.
Kojo Osei (53:27)
Thanks, Keith. It was great to be here.
