This Founder is Making 1B+ Excel Workers 20x Faster | Meridian, John Ling

1. 0:00 – 1:30

   Intro

   1. JLJohn Ling0:00

      I think I just really enjoyed learning new things. Doing more work was just, like, more opportunities to learn. Oh, there's like 50 problems. In each problem, you probably, like, learn a little bit more about something completely different. The way I would just think about it, I would just go try. And if you fail, that's okay. I don't believe any person on the planet spent 1,000 hours trying to build financial models with AI. Okay, I'm gonna do nothing except for, like, instruct the AI, and I'm gonna try to build this, like, LBO model that I would otherwise have to do for work. If anything about the bankers, they're just like, "We're just gonna do it by hand." And then if you don't know how to do it, you probably just don't know how to do it. I think that there's probably some kind of, like, decomposition that you can do where, like, models can do different parts of this workflow very, very well, but you just don't know because you haven't really, like, spent the effort to do, like, the investigation. And I was like, "We should go solve this problem." My name is John, co-founder and CEO of Meridian. We're essentially building AI for spreadsheets. We think about, like, Microsoft Excel as the most distributed programming language in the world, and our goal really is to say, "Hey, how can we help all of the people that spend a lot of time in spreadsheet software today just move 20 times faster?" Prior to that, I spent about a year and a half at ScaleAI. Before that, I started a couple companies. We've raised slightly more than $15 million. Our seed round was led by Andreessen Horowitz and the General Partnership. I mean, that's kind of where we are, relatively early, but hopefully we can continue to grow. [upbeat music]

2. 1:30 – 2:59

   How he became a top 1% performer at ScaleAI

   1. JLJohn Ling1:33

      I think I just really enjoyed learning new things. I think more than anything else, I felt like doing more work was just, like, more opportunities to learn. Oh, there's like 50 problems. In each problem, you probably, like, learn a little bit more about something completely different. And I think Scale was one of those places where if you wanted to learn about a different side of the business, you could go do that. It wasn't like, "Hey, your job is, like, X. You can only do X." It was like, "Your job is X, but, like, if you do X and you realize that, like, Y and Z and A, B, C could also be done," there was the opportunity to essentially say, "Hey, I'm gonna go learn and, like, expand my personal sort of, like, knowledge space, and, like, go do these things." Being willing to sit down and, like, dig into research, for example, was extremely valuable. I think especially in AI, it, it becomes relatively easy to get lost in, like, the execution, meaning like, "Oh, okay, we're just gonna do, do this because, like, we need to get this thing done." It's actually really valuable to take a step back and say, like, "Why are we doing this?" And then the way you learn is that you probably just go read all these research papers. Well, let's just, for example, take, like, quality of data. Like, what does it mean for data to be high quality versus low quality? What do researchers care about? What specifically makes this data point valuable? Like, I sat down and I read, like... I went through, like, so much of our data across so many domains, and I think that's, that's one way to learn.

   2. SPSpeaker2:58

      I met John through mutual

3. 2:59 – 7:26

   Why I Bet on This Founder - a16z, Kimberly Tan

   1. SPSpeaker2:59

      friends at Scale, where I had consistently heard that he was really a top 1% performer at Scale. I heard this across the board from many, many people. He didn't allow the confines of Scale, which was already a growth stage, larger startup at that point in time, confine, like, what he thought was right or not right to do in the business. And so he really took a very first principles approach in thinking about what would the right thing for Scale be, and he was unafraid to voice those opinions to people and then actually move mountains to make them happen.

   2. JLJohn Ling3:30

      Why go over to Scale? But I do think, like, the biggest reason was definitely, like, I felt like it was a very unique place to observe AI development. I think they were very convinced, obviously, that the next wave of, like, types of, like, large language models are going to very dramatically change trajectory of what the world looks like. For myself, I think selfishly I've always wanted to start another company. I think that not knowing what LLMs can do or, like, not really immersing yourself in sort of, like, this rapidly developing ecosystem or technology, however you wanna think about it, is, like, a mistake. I would be much better off spending, like, the next four years, uh, at least at that time I thought I was gonna be at Scale four years, really, like, learning as much as I can about how large language models worked and how it was developing, what was the trajectory of the technology and, like, how people are, like, implementing it, et cetera. Well, a lot of my job was making sure that, like, hey, the data that Scale ultimately produced was valuable. Spent a lot of time thinking about, like, benchmarks and evaluations. Also spent a lot of time thinking about, like, hey, how can we internally, like, leverage LLMs to make our internal processes more efficient? Um, so I think that for me was, like, really, really, really interesting. I started using Cursor a lot, um, over the last couple months. Or like, you know, the last generation of models where, like, hey, coding, like, really felt very real. Zero to one actually went from two weeks to, like, 30 minutes or, like, half a day. I had a moment where I was just like, "Wow, this thing is, like, magical, and I want, like, everyone to, like, go use it." You know? I was just like, "Everyone on this team must vibe code, and if you don't know how to vibe code, I feel like you're just gonna be lost or you're gonna be left behind." But, like, ultimately, I think it was just, hey, here- here's, like, a new calculator, but it's, like, not a... It's, like, a super, super powerful calculator. But I think, like, more tangibly, 'cause I live in New York, like a lot of my friends work in finance, and I think that, like, the energy is just, like, completely not the same, right? Where, like, you're in San Francisco. Everyone is, like, super, super excited about, like, okay, here's, like, the latest vibe coding, like, unlock, right? Where, like, oh, you have all these, like, skills that you can leverage for, like, Claude, for example. Or like, here's how you can do these, like, crazy architectures. It feels like the ground or the, the, the number of tools sort of, like, is increasing, like, exponentially. And then, like, you come back to New York and, like, that's just, like, not true. When I- Talk to like our team, when I talk to like candidates, or even like investors, I think, I think a lot about the idea that I don't believe any person on the planet spent 1,000 hours trying to build financial models with AI. I don't think anyone has been s- sitting down and been like, "Okay, I'm gonna do nothing except for like instruct the AI, and I'm gonna try to build this like LBO model that I would otherwise have to do for work." 'Cause if you think about it, the bankers, they're just like, "We're just gonna do it by hand." And then if you don't know how to do it, you probably just don't know how to do it. But I think that there's probably some kinda big decomposition that you can do where like models can do different parts of this workflow very, very well, but you just don't know because you haven't really like spent the effort to do like the investigation. In contrast to that, when you think about code, I think that a lot of these coding tools are built by the people who use them, so they have a much clearer idea of like what does success look like? What are the different use cases that I care about? I can very clearly articulate where the model's failing. But I do think when you take that and you apply it to a domain where you're like not really an expert, it's, it's pretty easy to say like, "This model is wrong," but it's pretty difficult to really identify exactly why, like the number is not the number that you would expect it to be. But yeah, that's kind of how I thought about it, and I was like, "We should go solve this problem."

4. 7:26 – 10:08

   Bias Towards Action

   1. JLJohn Ling7:27

      I think if I look back, my first job out of college, I think that most salespeople would probably also tell you this, right, is like if you don't try to talk to someone, like you will never know. And I think that's something that I've like always done. I would say like don't be scared to reach out to people. Don't think that like, hey, Satya Nadella will never respond to your email. I mean, if you think that way, he's obviously never gonna respond to your email. But if you reach out, you might be surprised. Maybe he'll respond. That's like something that, you know, that I thought was really, really interesting. It's really easy to fall into this narrative that like, oh, these things are like impossible, but you actually don't know. And I think like, you know, most entrepreneurs sort of just have that belief. I think it requires like a, an enormous amount of like suspension of disbelief, right, where you can, where most people would just be like, "You're crazy," but you can actually go in and just be like, "I don't know what you're talking about. Sounds totally doable," right? And then you would go try to do it. You also learn by like trying things that you've never tried before. And like if you fuck up, you fuck up. It's okay. Nothing wrong with that. But at least you know, right? And you can build reps internally. You know, our company as a whole actually promotes and allows people to like try to solve things their own way. And if you fail, it's okay. You just go support them, right? You're like, "Hey, you tried this thing. Maybe we need to push back the deadline by a few days, and then we'll like find other people to support you," right? Everyone in the company will come support you. And I think you have to build this like environment where it's okay for people to like experiment and not succeed. I mean, I think like obviously you always wanna build something that is like, like a masterpiece, right? Like, I think our goal for like starting a company obviously is to like build something that we can be really, really proud of, that we think is going to transform a lot of people's lives, that is going to be like, hey, here's a company that we can look back on in like five years, and it has like dramatically impacted the lives of like a lot of people.

   2. SPSpeaker9:19

      As we think about how knowledge work is gonna change with AI, there's almost no bigger category of knowledge work than the spreadsheet and Excel worker. And as someone who worked in spreadsheets and Excel as a banker for a brief period of time and then as a consultant, um, I could just viscerally understand, one, like why this was enormous market, um, and probably in some sense like one of the largest, uh, software markets out there, and two, why AI was gonna fundamentally change how we did work on spreadsheets. And so I think that, uh, Meridian's vision to really augment this form of knowledge workers, similar to how a lot of the, the coding companies have augmented the work of the developer, I think there's just so much potential here to actually be able to infuse the work done in spreadsheets with meaningful intelligent and automation.

5. 10:08 – 11:57

   Spend 10,000 hours with AI - Own your unfair advantage

   1. JLJohn Ling10:10

      The more time you spend with the technology, the easier it is for you to have an intuition around like what is possible today. And if you do this over like a very sustained period of time, you also build an intuition of what is going to be possible in like three months or what is going to be possible in like six months or a year, right? And I think like that in and of itself is extremely valuable. I would just spend as much time as you can playing with it, right? Like, I think it will be advantageous to be one of the people that have spent... Let's say you're interested in finance, right? That have spent, you know, like 10,000 hours trying to do finance with AI. I think that prompting is still a very, very valuable skill. Like when you apply to like Y Combinator, they actually tell you that like doing the application in and of itself is super valuable because it forces you to sit down and think through these aspects of your business that maybe is not as well articulated in your head as it is until you write it down. I think that when you explain a task to a large language model, in a similar vein where you learn how to be relatively specific about your ask, you learn a lot from that process, right? Like trying to explain to an LLM like what you re- really want it to do actually gives yourself a lot of clarity around what you really wanna do, and I think that part of it is actually very valuable by itself. [upbeat music]

Episode duration: 11:58