A little context on what I’m so excited about this interview.
A couple of years ago I attended an event where a former anti-trust policymaker spoke at Brown University. In making the case for breaking up big tech, he said markets had never disrupted a technology monopoly and ascribed American technology dominance to antitrust actions in the 1970s.
While I think the later 1970s and early 1980s are an especially fascinating era in modern American economic history, I challenged the point. I got up and asked, “Is it possible that the structure of American capital markets contributed to the success of the American technology sector?”
The speaker looked at me confusedly.
I continued: “For example, we have a much more robust venture capital ecosystem here in the United States. Is it possible that the ability of technology disruptors to get funding helped the American technology sector succeed?”
I didn’t really get an answer, so I sat down.
Fast-forward to the current day. I sat on a panel with several young start-up founders. And by young I mean 20, not 35. Very inspiring. I saw next to Teddy Solomon. Teddy is 23, and co-founded social media app Fizz when was 19. Fizz has received more than USD 40MM in venture funding.
The venture sector’s willingness to fund young entrepreneurs, outsiders and disruptors contributes mightily to technology innovation and to American economic success. Boldstart ventures, which Ed Sim leads, invests in the domains I am most passionate about -- enterprise infrastructure and cybersecurity.
Three takeaways for me from the discussion
< 1 > Venture capitalists don’t just allocate capital -- they are mentors and coaches who help entrepreneurs with disruptive ideas build companies. (Though Ed cautions us that not every VC engages with portfolio companies in a helpful way.)
< 2 > Procurement and other corporate governance function are not preventing enterprise tech organizations from doing business with disruptive firms. Some of the larger firms are explicitly investing in technology innovation teams to engage with the startup community.
< 3 > Everything is going to start moving more quickly, but platforms will be critical. Large companies are getting serious about AI-enabled development, but they will need the platforms that provide guardrails and enable resiliency, security and performance at scale.
James Kaplan: Alright, can you hear me, Ed?
Ed Sim: I can, James,
James Kaplan: Excellent. Alright. Welcome, Ed Sim of Boldstart, to the Prosaic Times podcast.
Ed Sim: I’ve been loving your newsletter.
James Kaplan: I’m glad. Why don’t you introduce yourself — tell us a little bit about yourself, and particularly your journey to becoming a venture capitalist.
Ed Sim: My name is Ed Sim. I’m the founder and general partner of Boldstart Ventures. we started that, or I started that back in 2010. So I’m on year 16
James Kaplan: Congratulations.
Ed Sim: Yeah, thank you. I mean, year 30 being an enterprise venture capitalist, believe it or not. So I started back in in the mid-nineties, and as I like to say, being the partner to founders from the very beginning, from the inception, and what that is has changed over time.
But that’s what I’ve done and I’ve seen a lot of trends and seen a lot of different things.
James Kaplan: I was on on a panel yesterday with a couple of founders who were between 18 and 23, so I’m just real stoked to be talking to somebody who remembers the mid 1990s.
Ed Sim: I got into it because I actually learned about venture capital when I was in college. I played lacrosse at Harvard, and someone on the team — their brother was a venture capitalist who came in and spoke one day.
I was enabled with the idea of, of like, you could either be a founder or you could actually back multiple founders, having your eggs in one basket or in a few, but then also having the ability to be a player coach. And really the idea was helping founders accelerate the process of, of realizing their dreams.
James Kaplan: If you could speak a little bit about that — I think most people have the perception that venture funds mostly allocate capital and don’t realize how essential they are in helping companies become successful. So I was wondering if you could speak a little bit about that.
Ed Sim: You know what’s funny? I won’t say that we help them become successful because my running joke is that I think more VCs have been responsible for damaging companies than helping them succeed, to be honest. But I think the idea of partnering at the beginning of the journey — being an early true believer, the first investor, writing that check to help them get started and back the business so they can focus on building a product and getting to customers — to me is the most exciting part of the journey. Because you have no idea, things will change a hundred times over. And having the ability to believe in a founder and their vision, and then also having the patience to live through the thick and thin — that to me is the best part of the journey.
It is the most romantic part, is the hardest part and is the most fun part.
James Kaplan: You mentioned — and you joked, not without justification — that some VCs help companies and some VCs may not be as helpful with the companies they fund. What’s the distinction between a helpful VC versus a less than helpful VC?
Ed Sim: For me, I tell the founders when we partner with them that my job is to help them accelerate their path to product-market fit. I don’t actually tell them what to build in the product, but I’ll make some introductions to Fortune 500 executives, for example, who might be buyers of the technology, so they can provide feedback. It might be introducing them to a key hire. It might be challenging them on their strategy. What I like to say is that you want to feel like you’re an adjunct to the team. The best founders are the ones that pull from us versus us pushing on them. If you’re constantly pushing on a founder, then you’re really annoying them. So the best founders, in my opinion, actually pull from you, and they know exactly what they’re getting from you and how you can help them realize their goals faster.
James Kaplan: I recall Marc Andreessen said that in founders, he looked for a certain degree of disagreeableness — the willingness to get the door slammed in your face 50 times and then say, no, no, no, this 51st time we’re going to get that deal, we’re going to get that logo. What do you think makes for a good founder?
Ed Sim: Yeah, I have boiled it down. I like frameworks. I’m a framework thinker, and I think you are too, James, in certain ways.
It’s the five Ps. When I look at companies and founders, the first thing I look at is what is the product. The founder has to have the ability to zoom in and understand the day-to-day life of a user, which could be different from the buyer. And if they cannot understand that product perspective with a deep, unique insight — in terms of why they’re starting the company — that’s a problem. Two is they have to be passionate about it. They’re usually obsessed about this. They wake up in the morning thinking about it. They’re in the middle of the night thinking about it.
They’re in the shower thinking about it, and that passion carries them through hard times where they get 50 nos and get the 51st yes, because they believe so deeply that this should exist in someone’s hands — it’s really going to help that person achieve better goals or outcomes. Three is I think about the art of the possible. In the game of venture, to be clear, our job is to return capital to our investors. And you have to ask the question very simply: how big can this be? When you’re looking at emerging technologies like AI security — when we invested in that a few years ago, there’s no market size TAM, there’s no investment bank or Gartner research on it.
It is just like, hey, will there be more ML and AI in the world? Yeah. And are they going to drive better decisions? Yeah. And do we need more security? Yeah. Okay. That TAM could be massive. I just don’t know when. But that art of the possible is super, super important. And there are a few other Ps — two other Ps that I have — but one is the people. The other thing I look at, James, is: where did you come from, and who else is coming with you? If you’re not a pied piper of people, that means perhaps in your prior roles — and look, there could be some situations where you didn’t get along with someone — but in your history of working, you probably have people that want to join you, that will follow you. And that people aspect is absolutely massively important.
James Kaplan: where do you stand on the technical versus non-technical founder debate? There’s some people who say they only want technical founders. They some, some people who say can be technical or non-technical. What’s, what’s your view there?
Ed Sim: We only invest in technical founders. If you look at our tagline, it’s partnering with technical founders from inception, building the autonomous enterprise. And that to me is very important. But let me say one caveat, Technical founders. I’m not talking about the person just sitting behind a desk all day not talking to people.
Part of being a great technical founder and a founder, itself is being a great storyteller. I like to say that this person, whether it’s the founder or the co-founder of both of them together, have to have an ability to tell two stories. One is zoom in what I talked about earlier and tell the life of a product user and why they’re going to be 10 to a hundred times better using your product. And then two is they have to zoom out and tell a bigger story of why. I should buy your product. Why I should join your company. So that has to be a bigger story. Like this is how we’re going to go on this mission to build something bigger, to change the world, to, to do whatever. And you have to have both, because if you’re just on the product piece itself, then no one may want to join you because they’re like wondering how big can this be? You have to be able to do both.
James Kaplan: Back in the 1990s, somebody told me the most effective people could operate at lawnmower view and jetliner view and switch back and forth between the two fairly quickly.
Ed Sim: I love that that is exactly what I’m talking about. And that context switching is rare in, in one person. but those are the best kinds of founders. and you know them, James, you’ve met with many founders before, and there’s this trusted relationship, particularly in the early days you built with them.
Like, I think this founder’s just going to figure it out. I trust that person to figure it out and they have enough tech technical depth to make it happen.
James Kaplan: What makes a great storyteller and what makes a great story to your thinking? What are you looking when someone says, okay, gee, I want to go do this, or I want to go say this to customers.
Ed Sim: yeah. first of all, first and foremost is clarity. Clarity of vision. And the way I break down a story is what’s the problem succinctly? Is it a hair on fire problem? what is your product that actually solves that problem? And then three is, how big can this be if we all go down that path?
Three, three very simple sentences around that. So clarity. And for me, a founder usually doesn’t have it all the time, right at the very beginning, that. That, I’ll call it the blurb, like that statement that you put in an email or the introduction three sentence statement that evolves over time.
And part of being a great founder means how fast do you learn? How fast? What is your learning velocity? How fast do you take input? How fast do you figure out signal from noise? How fast do you adapt that blurb, and change it? And how open are you to changing it over time? Because you can’t be rigid, you can’t take all the input period, right?
You have to be disagreeable in certain aspects and you’ve, and then you come up with your story. And I think that is, that is part of the importance around it.
James Kaplan: Tactically flexible and strategically disagreeable, perhaps, or tac, strategically fixed and tactically flexible.
Ed Sim: It is you have to have some stake. There has to be meat in the bones, but there also has to be some sizzle.
James Kaplan: Yeah. I mean my, my observation is the best storytellers, even though those three lines can, imbue it with some drama, it’s not that you have a problem and fix it, but you make people care about the stakes and the problem.
Ed Sim: I think what you’re saying is that passion jumps off the screen, person you feel the energy, like that energy transfer when they walk into a room and tell you a story or when you share, there’s a dynamic give and take that is I think, going to be even more important in the world of ai.
There’s being a lot of slop out there, but that is why you choose one versus another.
James Kaplan: And how much of a coach are you to the founders? Because I imagine there’s some folks who have that vision but haven’t quite figured out how to light up the room yet. How much time do you guys spend trying to teach people how to do some of that? Do you, do you see what I’m saying? To what extent are you looking for people for habit, and to what extent are you looking to help them get it?
Ed Sim: I think it’s a combination too, right? When you’re paying for a second or third time founder and you’re paying a higher valuation from the very beginning, even in inception, you’re going to expect them to have all of that. And then there might be some new technical founder creating some brand new market that they only raise a million and a half and they’re young and inexperienced, but they’re also stupid, crazy, right?
They have that list where they’re trying and you’re like, I’m willing to work with this person as well to help them along. And I will also introduce other founders. Who’ve been there before, who can help them along as well. And so I think it’s both, and I enjoy doing both.
It’s actually a lot of fun. And I’ll give you one last perspective. I think the best investor as a, I’m on boards, like we decide to take board seats early, but I call it the three Cs. You have to know when when to cheer. Usually when you cheer is when things are actually going bad.
Like if a founder came to you and said, I just lost that million dollar deal, and they’re about to jump off the roof, you want to pull them back and say, that’s just one. We’ve got three others. Let’s just keep going. need to know when to challenge them, and that’s usually when they feel invincible. Hey, let’s go spend three times as much on this.
Because things are going so great and we’re never going to lose, and my product is so good. You have to beat them up a little bit to make sure that they, they don’t get hubris. And sometimes James, you just have to know when to chill. You just have to know when to leave them alone and let them do their thing and not get in their way.
And I think those three different opposing viewpoints, I think is, is very important to bring to the, to the game.
James Kaplan: One of the best pieces of advice I ever got in my life was from my dad when I was, it was seventh grade or something. I was in, I was in the little league and it was , two outs bottom of the seventh inning. I come up and I was just having a great year and I, um, you know, and we, my team was behind and I asked.
Absolutely knock the daylights out of the ball. Hard line drive right at the third baseman’s chest, six inches to the left. It’s a standup double six inches to the right. It’s a standup double right where it was. I lined out and ended the game and you know my, I was really dejected, right? My dad’s like.
Nothing you could have done. All you can do is hit the ball, just keep hitting the ball with authority. And most of the time it’s going to land. And that’s, to me, that’s your, you know, sometimes you, you do everything right and someone else gets the deal and you have to not let it get inside your head, I’m sure.
Ed Sim: I love that. Yeah, that’s so true. That’s what the journey is like, and people think it’s just glorified. It’s always up into the right, and it’s not like that. A lot of times it’s two steps forward, four steps back, five steps forward. I mean, it, it, it’s a journey and you, you can get beaten up a little bit.
James Kaplan: Tell us a little bit where Boldstart sits in the VC ecosystem. What domains where in the life cycle, where do you guys tend to focus and why?
Ed Sim: I like to say that in today’s world of venture capital, there are only two ways to win in my opinion. One is to specialize and the other is to go big. It’s like the traditional adage in any business: you go big, you go niche, or you go home.
We decided to specialize, and we specialize by stage and by outcome or product that we look at. So stage — I call it inception. This is not seed. It’s not pre-seed. It’s two people, one person with an idea, and they’re in the idea maze, the idea jungle. And we want to help them accelerate the process to saying, let’s start a company. And we’re there at incorporation and leading the round.
And that check size could be anywhere from, 200 or 300 K up to 15 million right out the gate, depending on who you are and what your background is and what the opportunity is. And so we
James Kaplan: is that an alternative to Angel, or is that a compliment to, how does that relate to, the angel round for want of a better term?
Ed Sim: is usually we had, we bring some angels on, we usually leave some allocation for high profile angels, former CEOs, former executives, but like getting an angel round done and you’re raising a million dollars, versus when you need five at once, right?
You want one place to go when it’s at the idea stage. And the problem with preed and seed in all these other categories is that a lot of these. Investors want to actually have more proof of traction. And it’s just a, a, a weird dynamic. And so, the other aspect would be the multi-stage firms do everything.
They write checks at our stage, they write checks, in series H at, 50 billion valuations, right? Or 500 valuations. So, I think our specialty is working with these founders at that stage, knowing the ups and downs and actually getting them to a place where we can bring the other people into the. Cap table very early, and types of companies, it’s, this is why we’re.
We’ve met each other through all the things that we do in IT infrastructure and cloud and security. I call it the autonomous enterprise. That’s when we launched our fund back in June — our Fund Seven, our 10th fund. And the idea is that we back technical founders building the autonomous enterprise. My running joke is that the goal is that one day we want that one person, thousand-billion-dollar company.
James Kaplan: the, the solo unicorn, right?
Ed Sim: But the reality of it is, is that there’s a lot of plumbing that has to change.
There’s a lot of last mile technology that needs to, to be delivered. There’s a lot of security. And so the aspiration of that is we want to fund everything underneath that, all the plumbing, the security, even to the agents being built on top. and I think that’s a, a great aspiration.
James Kaplan: So when I talk about the tyranny of the use case — just looking at vertical use cases only gets you so far. The companies you fund are the solution to that problem. They build the underlying infrastructure that allows — I also find myself saying technology is easy.
Technology at scale with security, resiliency, compliance, and maintainability and efficiency is hard. You do all that second bit.
Ed Sim: Yeah. And it’s really interesting because there’s been a lot of debate, as you saw in the last couple weeks, in the stock market about — oh, Claude released a plugin, right? So, and my running joke is, if you’re a startup can be reduced to a plugin or skill, you’re not long for this world and probably not a good company to start and bank your life on. the underlying infrastructure, you see, the Mongo dbs, the Datadog and all those, and Databricks, right? They’re all still growing very fast. You still need a lot of that. I’m not saying that we’re going to back RAG, vector databases or things like that because I think that’s a already subsumed into what the model providers are providing.
But you nailed it, James. I mean, we talked a lot to lots of Fortune 500 CIOs and CISOs and you do as well. And man, that last mile problem, there’s so still so much more to do and build and everything else like that and I can, walk you through things that we’re looking at. But I would love to hear your thoughts on that.
James Kaplan: Before we go into, the evolution of the space, I was wondering if you could speak a little bit about the purchasing dynamics for Fortune 500, which is, it’s a different world than it was 20 years ago, right? We have, it’s more systematized. It’s more consolidated. The procurement department is a lot, stronger than it used to be.
the security function is a lot, fortunately is a lot stronger than it used to be. We have approved vendor lists each. To my perception. Tell me if this, a few companies have official innovation programs, but that’s relatively small. My perception is it’s harder to get a foothold than it would’ve been 20 years ago.
Is that a correct or an incorrect perception? And how do you know, how, how do you, and how should companies manage some of those challenges in selling to behemoths with increasingly complicated purchasing cycles?
Ed Sim: I would say it’s yes and no because I work with a lot of them themselves. And I do work with like a lot of the, what call it the emerging tech programs, innovation, lots of great ones out there. As you know, there’s JP Morgan, Morgan Stanley, bank of New York, has a new thing
James Kaplan: So that, that’s tends to be certain companies with a billion dollars in spend it up. If you’re like
Ed Sim: Exactly.
James Kaplan: to a billion in IT spend, you’re less likely to have that.
Ed Sim: Yeah. Yeah. And so these larger companies, like here, let’s, let’s just take a step back.
James Kaplan: Yeah.
Ed Sim: I’ve found certain companies that I invest in reaching out to, not only them, but even people without formal programs being as interested in moving so quickly, at least to take first or second meetings.
James Kaplan: Interesting.
Ed Sim: And the meetings, having a check from a vetted investor who’s done this a number of times. We’ve had lots of exits and companies have grown to multi-billion market cap, I think helps get the awareness. So when we do send something, I mean, for example, James, I send you something, I only send you a few things a year, but I want to make sure that it’s going to be interesting stuff and hopefully you’ll have the chance to take a look at it.
But I think it’s easier to get a meeting in certain ways. but I do think that there’s also. So much noise out there, it may be harder to close deals, right? So not only you get those meanings, you’re going to have to separate. You’re going to have to be that person that we talked about, the dynamic individual that sells a story, builds trust. And then you also have to have incredible time to value. you have to be able to pitch incredible time to value so that they can get, a quick ROI from it. And then B is tell a bigger story about why they should standardize you versus 12 other people
James Kaplan: Right.
Ed Sim: the same product. Because AI has just really reduced the moat for many people.
James Kaplan: The, well that’s, that’s encouraging because I, will, on reflection, compared to say 2007 or 2010, we have less of a technology of legal opin. If you go back to 2010, three enterprise. Application vendors. one of two storage vendors. one of three server vendors, one of two network switch vendors, what have you.
And we just have a more distributed world than we used to, and I think we have more CIOs and CTSA, gee, I can’t rely on. My two or three, two or three behemoths to do my technology strategy for me. So that, I suppose it’s the positive thing. And then the flip side of that is the, you know, just managing the, you know, the procurement process, which can be quite comp and getting on a approved vendor list, which can be kind quite complicated.
Ed Sim: [Enterprises] have no idea what they’re getting at, because they still don’t know that on-prem may need, may be important, hybrid may be important. And so when they hit that like, oh man, I thought everything’s cloud-based. Well, no, when you have data and privacy and all those things that you worry about and you’re feeding ai, they’re, you don’t want to pass your crown jewels over to a model provider, sight unseen.
James Kaplan: Yeah, one of my, one of my favorite enterprise IT stories is, some from a, a cloud. related business. Went to meet with the head of infrastructure for one of the major, major banks, and they come in and they do the whole pitch, automation and auto scaling and this, the whole thing, right?
And this, head of infrastructure who is a bit of a dry sense of humor says this is, this sounds fantastic. I get just one question. I get data that can’t enter Switzerland. I get data that can’t leave the United States. I get data that can’t enter Singapore. How do you guys, how do you guys, how do you guys deal with that?
And he swears, the folks from this sales team look back in and say, why would you ever want to run your business that way? To which he replied, you seem like nice boys. Come back when you have this figured out. Which to me was the, the disconnect between, the, the startup environment and the enterprise right there.
Ed Sim: Yeah. But, but to that point I do
James Kaplan: They did, they did figure out. I will note,
Ed Sim: because of how fast the technology’s changing, I do see larger orgs being more willing to
James Kaplan: absolutely.
Ed Sim: of next generation things.
James Kaplan: So sometimes I will admit, I do see tension between the C-I-O-C-T-O organization. And the procurement organization.
Ed Sim: Yeah.
James Kaplan: A horizontal there is any to be horizontal platforms as opposed to vertical use cases. So are they see less of the tension between the center and the application development teams than perhaps some others do?
Ed Sim: absolutely that that is the case. But look, if you’re selling automation technology, there’s always that though, whether it be automating cybersecurity operations, automating it operations, there’s always a tension of, once it goes below the CIO. Are you automating my job away? Right? And so there’s always that healthy tension and the answer of course is no.
We allow you to do more with less. We allow you to grow faster with less headcount. We allow you to do the more interesting things. But there is that tension. Let’s, let’s not, kid ourselves on that.
James Kaplan: So now, now let’s talk to how the market’s evolving. What do you see the themes and the major, what areas are you excited about over the next two or three years, and how has that evolved compared to the past three or two or three years? Give us a little bit of the, of the movie as it were.
Ed Sim: Yeah, so that I think about is that when I laid out our new fund strategy back in June, we talked about the autonomous enterprise. And in my thinking, we are 10 to 15 years plus away from that happening. And that we should be prepared for this long cycle, for enterprises to move from pilot to production.
So that was our thinking. However, I think there’s a sea change in the world in mid-December when people went back for the holidays and started playing around with Claude Code. If you think about it, I think about what developers do first as a tipping point before what regular humans do.
James Kaplan: I’ve used both Cursor’s and Claude Code’s agents, and I’ll say — without getting into details — that my coding agent experience has been transformative.
Ed Sim: Engineers are saying, “I don’t need to look at the code. Let me just have the agent crank away and fix itself and keep fixing itself.”
So that was a major step change. And then two, because the developers were using it, and they’re using it for other things like personal productivity — they launched Claude for Work.
James Kaplan: Yeah.
Ed Sim: I’ve played around with all that stuff. And fast forward, you have Goldman Sachs announcing just last week that they had been working with Anthropic and that they want to automate the back office for KYC and accounting. And then basically what happened was that Anthropic had been there for six months working with the folks at Goldman and they made this statement. And so my point is saying, wow, when have you ever seen something come out so fast? A bank, a regular bank, coming out and saying, “I’m going to do this.”
That doesn’t mean everyone’s going to do it, but I can tell you this, I do really feel like, based on everything I’m seeing, is that this agentic technology, or the autonomous enterprise as I call it. It is moving faster than ever because the technology is getting better. people, perhaps are trusting it more, but that also means more opportunity to secure all that stuff and everything else.
And we can talk about all that later.
James Kaplan: Can I challenge one of your assumptions or one of your framings?
Ed Sim: please.
James Kaplan: Where you said developers aren’t looking at code — let me suggest something to you. I suggest we call Cursor or Claude Code a compiler. Which is to say, 30 years ago we had this debate: how come people aren’t looking at assembly language?
Right. And I would suggest to, engineers are still writing code, they’re just writing. Declarative code via spec-driven development, right, rather than writing procedural code and declarative code is an infinitely more efficient way to communicate a set of instructions to a computer. Does that resonate at all?
Ed Sim: Well, it does in the sense that English, English is the language of coding right now. Right. In
James Kaplan: Right.
Ed Sim: explain things and that they’re writing things. So Yes, it does. But my point really is that when
James Kaplan: I’m not disagreeing with you. I’m agreeing with you, quite frankly. I’m just —
Ed Sim: Yeah. No, I agree in the sense that engineers are still writing — they have to use the English language to actually write something, a framework, and how people think about it. But man, when you have founders you’re sitting with that can’t wait to get back from dinner, saying “I have 30 agents waiting for my instructions” —
we live in a different world,
James Kaplan: The value in AI enabled development is less that we get to use English language. It’s more that we can be declarative rather than procedural, because we don’t have to do all the error checking ourselves. We can just say, these errors must be checked.
Ed Sim: Yeah. In other words, how about this? In other words, building things to abstract it away is, is more fun again, for a lot of these engineers. Right. And they can
James Kaplan: Exactly, so. Exactly, so.
Ed Sim: I think that that is removing the grunt work. And, but I think that applies to, so basically you see what happens in devs and that will apply to a lot of knowledge work as well.
We’re just going to be managing a fleet of agents. All of us. We’ll be managing a fleet of agents doing work for us, and then we get to spend more time doing podcasts and going to steak dinners and things like that where you build relationships.
James Kaplan: I just think we may wind up spending more time managing agents. When you have a 500-person team working for you, that’s a lot of work.
Ed Sim: This is why some of my founder friends were like, “Oh man, I’ve got to get back from dinner.” They were getting anxiety because they felt like the agents weren’t working —
James Kaplan: Exactly. Or they’re going off in the wrong direction. I think you touched on something else very important there, which is I suspect the line between software development and non-software development will get very blurry very quickly.
Ed Sim: Completely agree.
James Kaplan: I sometimes compare it to what happened on the trading floor in the 1990s, where people started building algorithms in a spreadsheet, and then started building algorithms in code, and it just wasn’t clear when somebody who had been a trader became a developer.
Does that make any sense?
Ed Sim: Yeah. Yeah, absolutely.
James Kaplan: It’s just — “Oh gee, I came here to be a trader” — or vice versa, when someone who was an engineer wound up helping define trading strategies.
Ed Sim: A lot of these are going to end up as a skill. Things will end up as a .md or markdown file. There’ll be a registry of these skills and people will be able to tap into them. It could be for knowledge work, for algorithms, for anything you want. You can run it on your own machine, you can run it in the cloud, but it’s going to be really interesting as these become packaged as skills — that’s going to be the standard.
James Kaplan: And then there will be a skill — a human skill — in building skills.
Ed Sim: Yeah. And then, and then there’s going to be a skill package manager and a bunch of other things, and I won’t name companies, but I have one of those that just launched today who you’ve talked to.
James Kaplan: Right. Yes. One thing I’ve been thinking about a lot is the rules and skills we place around agentic development and the version control we place around those rules or skills.
Ed Sim: You nailed it. And so that’s why I think you need a registry, you need a package manager, you need to vet it. So you need evals and all the skills, and every time there’s a change, it needs to update automatically because these are living, breathing documents that actually can execute.
That’s the difference between having a spec in the past where people usually never even updated it. Right. And then
James Kaplan: Right. Yeah. Now we have living tech.
As, I mean, we used to joke about, not joke about, we used to talk about security as code a few years ago, which eventually became cloud security posture management. Now we’re going to have spec as code, right?
Ed Sim: And now they’re just, I think they’re just blow down to a skill.
James Kaplan: Yeah, and part of those, some of those specs will be functional, go deliver this functionality. And some of those specs will be non-functional, here, I don’t know, perform this, scan the code for this type of vulnerability before it’s allowed to go into, go into production. Or even something like, we’re going to use this set of semantics to name variables and not that set of semantics.
Ed Sim: And some of those skills may be abstracted away so that they’re actually end user skills, for example, prioritize my buckets and do things like that. Right. So that’s what makes skills so interesting. I think basically we’re all going to be living in one large markdown file.
James Kaplan: You happen to be right. I, the, the future is marked down.
Ed Sim: Isn’t it crazy? The oldest tech — .md. GitHub — you write your website in MDs.
James Kaplan: Everything was a text file. Right.
Ed Sim: Yeah.
James Kaplan: Which is great because, text is, text is the ultimate universal spanning layer,
Ed Sim: I am spending more time in text edit right now in my machine than I ever did.
James Kaplan: It’s fascinating because you can edit the same file with two different tools. Everything’s readable if you need to. It feels like everything old is new again.
Ed Sim: And then the question is: it’s all cool for a startup, it’s all cool for individuals, but then how do you deploy enterprise-grade? How do you point that to 500 developers, to a thousand, to 10,000 developers? How do you actually do that? How do you build the security guardrails in the version control system, as you said? How do you have an enterprise-approved registry where all these skills are secure and have been tested and they actually work?
So, and what happens when there’s an upgrade, right? you don’t automatically use the latest skill because it has to be tested. It has to be so these are all the things that people forget will have to be built in the enterprise.
James Kaplan: There is one thing that gives me a bit of hope, which is people are really bad at interpreting rules. But AI isn’t. And you can start to encode a set of rules around: we name variables this way, we break up code blocks that way, we scan for this type of vulnerability. And you can imagine some of that being foundational and built into the underlying tools and some of that being idiosyncratic to an individual technology shop.
And something that the architects or the technology strata, strategist for that institution defined.
Ed Sim: I agree. And then also the key will be then it will learn continuously on each individual person. And over time it will be dynamic, because nothing’s ever static in life. Any organism’s not, static. And so that will also adapt over time with how your organization, does things.
James Kaplan: Well, and that I think gets to the, some of the real complexity about doing this in scale because when do you want to let it adapt and when do you not? Right? Because there’s certain things like, okay, let me give you a simplistic example. social security number or credit card number should never go out outside the four walls of the organization.
You don’t want to adapt that.
Ed Sim: No, that should not.
James Kaplan: You don’t want to say, oh gee, the context is different, so it’s okay to send this. So
Ed Sim: we’ll call that an anomaly, if that happens, right?
That there’s enough, action items that I can learn from.
James Kaplan: We’ll need to figure out, okay, what are the enterprise class rules? What are the enterprise class rules that. that never change and can change over time. What are the either business domain or sort of segmented rules and what are the idiosyncratic rules to a particular project or particular developer and managing that hierarchy, I imagine will be both interesting and challenging and a tremendous source of opportunity for a range of companies,
Ed Sim: I agree. And the question is, is how can AI help you write those rules, right. And keep them updated,
James Kaplan: right? And manage right, maintain.
Ed Sim: because that’ll be interesting, right? because there’s a big debate right now about, systems record and large companies like a ServiceNow or Salesforce, just crud, right? And there’s lots of rules built into it, right?
Lots of money spent and creating triggers and rules and how things work. Hundreds or thousands of them are encoded into the
process in these systems. So they’re going to be hard to rip and replace
over time, the question is, will agents get better to
those over one at a time, two at a time,
James Kaplan: and I think the, the SaaS space is, the business application spaces. I think it’s going to be choppier than the infrastructure space over the next few years because I think there’s a lot more, one size fits none at the business application layer than there is at the infrastructure and security layer.
Ed Sim: Agreed. For sure.
James Kaplan: All right, so we’re coming up at, coming up at 45 minutes or so. What, what have I not asked about? What should, what should viewers know about
Ed Sim: I guess where do I see holes out there, or
James Kaplan: please?
Ed Sim: some of
James Kaplan: sounds great. Yeah, that’s, yeah, that’s a fairly obvious question. That’s one I should have asked. Thank you.
Ed Sim: Look, at the end of the day, we need more security. There just needs to be more security around agents. For example, when you have Claude Code running around and doing stuff on people’s machines — let’s say you have an organization with people using Claude Code.
How do you make sure that the machine is locked down and not giving up credentials and other important things? I think that’s going to be very important. The other thing I keep thinking about is that in a world of agents doing work, we’re going to need fine-grained authorization and authentication.
Authentication is: is the agent who the agent says it is. Authorization is: what fine-grained control should they have? What should the agent be able to do, and what should it not be able to do? And I think agents have to have their own credentials because you have to have an audit trail.
Particularly with banks and regulated industries. Then on top of that is what happens if there’s a hijacking of an agent during a tense. Step workflow. How do you make sure that cryptographically, that agent, is what that agent is and continues to be throughout the process. So there’s a lot of plumbing that people don’t talk about that’ll need to get fixed.
Another thing to think about is the OpenClaw movement. We didn’t talk about that.
But the idea is that I’m definitely playing with putting it on a virtual machine on Digital Ocean — its own Gmail account, I created its own Telegram bot account, I have it training on my newsletter “What’s Hot in VC” to understand how I think.
And then I forward emails to it that are interesting and important, but I’ve gated it. I think that architecture as a gateway — this gateway that can do a lot of things — is really cool and should be very enterprise-ready. At the same time, who’s going to lock it down? Because we can’t get the full value. You said earlier, “I need a personal assistant” — that could be your personal assistant, but you’re not going to give it access to everything.
James Kaplan: I’ve been experimenting quite a lot with agents for myself. The administrative stuff works much better than the content-based stuff.
Ed Sim: Correct.
James Kaplan: Yeah. Let me ask this question. Rich Isenberg and I talked about this the other day — non-deterministic security. Here’s what I mean by that.
We currently have what I would describe as a very deterministic security model. You have this control. It’s either implemented or it’s not. Either it applies to this or it doesn’t. Social security numbers? Never, ever, ever. The DLP tool will never allow the social security number to exit the environment under any circumstances.
But if it’s a strategy presentation. No, most of the time you don’t want it, but if you’re doing, some sort of funding, funding round you, maybe you want to send it to your investment banker. Right? So is there, will we see more contextual or more cybersecurity controls to take context into account?
Ed Sim: Oh, context is key.
I won’t tell you which one, but I have a portfolio company that literally says, “Hey, you’re going to build agent technology to automate some of the work you do in security.”
But it’s very hard to do it without context. Do you want to do a one-off? Or do you want to have a platform that literally connects to every one of your security systems, connects to your identity systems, connects to your cloud, connects to who James is and what he does — and then that’s the context. Then we’ll watch people do work and say, “Hey, these are some alerts, but I think you can automate these processes.” I think context is absolutely king. And I’ll give you one other interesting aspect — we just invested in a network security company, name to be determined, that’s reinventing network security. But it has to be reinvented.
Why? in a world of agents
Agents are going to do 99% of the traffic. You have to understand the business context of what the agents are doing in order to block traffic or open up traffic.
It can be deterministic in some ways and non-deterministic in other ways.
because it has to be flexible enough to understand, hey, if I read this. Business context and I read what the rule is. I think it’s important that we let it through. Because based on everything I know about the business —
Or wait, it seems like an anomaly, it seems like someone’s trying to hack and do something different.
So that’s where I think you hit on in the non-deterministic deterministic framework. The context is everything moving forward.
James Kaplan: Perfect. Anything else?
Ed Sim: The only thing I can leave with is: you and I have been doing this for a couple of years.
James Kaplan: Yep.
Ed Sim: This is probably the most exciting and the scariest time there’s ever been in doing this.
James Kaplan: We can do things that were unimaginable 18 months ago.
Ed Sim: And I wake up even like two months ago. I wake up every day wondering like, what am I missing?
There’s this constant fear in a way. I felt some of it during the internet boom in the mid-nineties, but nothing like this because back then we had dial-up. I mean, the speed
at which everything disseminates and happens is absolutely insane. And by the way, the winners that we have today may not be winners tomorrow.
So that’s what is, is thrilling and it’s scary at the same time.
James Kaplan: Exactly right. I could not have said it better myself. Thank you so much. This was fantastic.
Ed Sim: Thanks. I enjoyed it.
James Kaplan: Right.
