In episode 43 of To Be Continuous, Edith and Paul are joined by Ellen Chisa, Paul’s co-founder and CEO of Dark, as well as John Kodumal CTO and co-founder of LaunchDarkly. The group discusses pitching, what it’s like finding a startup partner and the road to effective management.

This episode of To Be Continuous, brought to you by Heavybit. To learn more about Heavybit, visit While you’re there, check out their library, home to great educational talks from other developer company founders and industry leaders.


Ellen Chisa is co-founder and CEO of Dark.

John Kodumal is CTO and co-founder of LaunchDarkly.


Edith Harbaugh: What made you decide to do a startup?

Ellen Chisa: I did my first startup a little over 10 years ago. It probably should’ve been a reality TV show instead of a startup, where five of my friends and I decided to have a company.

Except four of them were dating each other, two pairs of two. None of us had ever not lived in a dorm, and it went about as well as could be expected. But it was enough to convince me to do it again.

Edith: Now would be a great time for you to introduce yourself.

Ellen: I am Ellen Chisa and I am Paul’s co-founder and the CEO of

Edith: Were you one of the two pairs, or the fifth person?

Ellen: There were two of us who weren’t in one of those ill-fated relationships. So I guess I was the fifth person or the sixth person.

Edith: So, what made you decide to do a startup?

John Kodumal: I’ve known Edith for a long time. I think it was like around 2007 or so, we were having a Thanksgiving dinner. Edith was at my house. I cook a mean Thanksgiving dinner.

Edith: Turducken.

John: Yeah, it was good. And I think we were just like, “We’re going to start a startup,” right?” Like, “You know that, right?” Took a little while for the timing to shake out, but we knew from that point it was an eventuality.

Paul Biggar: Was that your first startup?

John: My first startup, yeah. This is only my third job ever in life.

Edith: Now would be a great time for you to introduce yourself, since I know you quite well.

John: I’m John Kodumal. I’m Edith’s co-founder and CTO of LaunchDarkly.

Edith: I already have so many questions such as, this is only your third job? I mean, you graded in college.

John: Oh, okay, I suppose that’s true. I graded for a CS AD at Harvey Mudd, that’s logic. I spent two hours a week grading homework assignments.

Edith: And when you were getting your PhD?

John: I was just poor. I got a PhD in computer science, much like Paul. You just get paid very poorly for a number of years. I had a stipend, so I guess theoretically as a grad student you’re both an employee and a student.

But literally, I was in the Bay Area earning 20K a year for six years. So yeah, it’s not fun.

Edith: Were you a TA?

John: You can teach. I taught for one semester and then basically, the way it works out, is that happens instead of your stipend. So you don’t get paid more for teaching, it’s just something you have to do.

Paul: It’s crazy when you go to your first job afterwards, and suddenly you have seven times as much money.

John: Yeah.

Ellen: Why doesn’t everyone drop out immediately and get a job that pays seven times as much money?

Paul: Because if you don’t stay for the PhD, you’ll only get a job that makes five times as much money.

John: Yeah, and then you can walk around and people call you “Dr.”

Paul: Do you actually use “Dr.”?

John: I tried to once and it didn’t go over well. He asked me, I was at a bank and my credit card says “Dr.”

Paul: Yeah, I had that one, too. It was super awkward.

John: Yeah, they’re like, “Dr.?” Oh, what kind of physician are you? And I’m like…

Ellen: Computers.

John: Computers…? And they were extraordinarily disappointed.

Edith: Did they try to show you a wart or something?

John: Yeah, after that point I was like, no.

Paul: Showed you a bug in their software. They’re like, “Can you fix this?”

John: Yeah, it’s like, “It looks like a buffer overflow.”

Edith: Which is funny because your wife is actually a medical doctor.

John: She is, yeah. She’s an ophthalmologist and she has the same sort of insecurities around her title as me. It’s like, I’m a doctor of something useless, but she also gets confused with optometrists. Optometrists are different from ophthalmologists.

Ophthalmologists are medical doctors and surgeons, but if she tells people she’s an ophthalmologist, a lot of them think she’s an optometrist and they think she does the “cover your eye and read the big ‘E'” thing.

Edith: You know, what they both have in common is that nobody can spell them.

John: It’s true.

Edith: Ellen, you had this, first, reality-world startup, and then you decided to do another one.

Ellen: There was more in between the two. I had the first reality-world startup and I learned many things I was bad at. Then I decided to go learn a bunch about all of them. Eventually it would be the right time to do another one. But I didn’t want to do it before I felt like I knew the right set of things.

Edith: How did Paul persuade you?

Ellen: I think Paul had happened to meet me at the exact moment where I knew I was ready. because the job I had before this, I’d been in EIR in an incubator and I started working on a project. Then eventually the incubator went away, we kicked out of the companies we’d invested in, we canceled all the other internal projects and started building it.

I’d just been through this experience of seeing a team scale from just me to 50 people. And go from, basically, where we were at at the beginning, which was, it was a project within an incubator to a serious B stage company.

I’d seen the entire process and it felt like there were no excuses not to do it anymore. If there was time to do it, it was now. 

Paul: It was good timing on my part.

Edith: Did you have hesitations, given the way stuff had gone south before?

Ellen: No, there were 15 minutes where Paul walked me through a lot of slides about X and L complexity, and I was like, “Yeah that seems like a problem.”

There were a couple slides in the middle where I was like, “Hm, I don’t think this is going to work.” Then about 20 minutes later, I was like, “Oh no, this is definitely going to work. Yeah, we should do this.” But it was very much a within-the-first-half-hour decision.

Edith: John, it took us 10 years.

John: Yeah, we weren’t even sure what the idea was for a while.

Paul: You were doing consumer stuff or consumer ideas before you decided to do LaunchDarkly?

Edith: We had a joke that we should make like a graveyard of all our ideas that we tried for about a day.

John: Oh, there was the static analysis startup.

Paul: Oh yeah, that would’ve been good.

John: Yeah, it would’ve been really good.

Ellen: Did you have a process for how you tested all of the ideas, or was it just every day something new?

John: Mostly we convinced ourselves that they were terrible ideas as quickly as we could.

Paul: And they took a day?

Edith: The terrible ones would take an hour.

John: Yeah, the static analysis one we worked for six or eight weeks on it, maybe even longer.

Paul: Oh, shit.

Edith: No, like eight months.

John: Oh, eight months. It flew by so fast because we were having so much fun.

Paul: But what was it going to be?

John: I had this observation that DVCS and Git and the way people do development doesn’t work really well with most static analysis tools. because they’re not built to be incremental in any way.

The thing that static analysis doesn’t have is any notion of time, like how the code base is changing over time. 

So I had this observation that if you married the types of static analyses, that map easily to graph reachability problems, to get meta data and you layered get meta data on top of that static analysis data, you could basically use data log as a query engine to ask questions about code base and static analysis questions that marry time and commit history with the facts that you’re interested in.

It would be incremental by design. When you made an incremental change, you would just re-derive the necessary facts, based on that change. So it’d be much faster and you could do things with temporal understanding.

Paul: Did you come to the conclusion that this was a terrible idea, or no market? How did it go from, “Oh that sounds pretty interesting,” to…

Edith: We worked on it for a solid year.

John: We worked on it for awhile.

Edith: Because when I look at dates…

Paul: This is ramping up, we’re six weeks and eight months, now it’s a year.

John: We spent like 15 years on this.

Edith: Well this is development estimation versus…

John: It was a combination of two things. The first and most important thing was, Edith was talking about the idea with a bunch of prospective customers and the level of excitement we were hearing was not that great. There were a lot of attempts to commercialize static analysis right around when I was finishing grad school. And they all ran into that hair-on-fire problem.

Paul: You worked at Coverity, right?

John: I worked at Coverity for three years.

Paul: That one got commercialized and it exited.

John: Yeah, it exited. Did a really good exit. It was a great company, some amazing technology. But I think a lot of the stuff that they were doing, it was really applicable to legacy code bases or a lot of like embedded systems and stuff like that.

Then modern software development, like, modern web software development, say as a subclass in there, a lot of techniques there weren’t interesting and Coverity as a tool wasn’t interesting for them.

Paul: So you just talked to all the people, and then…?

Edith: I got kind of a begrudging, “Hey, maybe I’ll look at that as a favor to you.”

Paul: What was the process of going from that to, “We’re not going to do this anymore?”

John: It was a combination of that plus I realized at some point I’d spent nine years of my career working on this stuff and I was done.

Paul: Oh, yeah.

John: It was just too much. I knew how far the rabbit hole went and I desperately wanted to get out of that rabbit hole.

Ellen: At this point was it just the two of you? Just working on ADS together and trying to figure out what to start?

John: Yeah, there’s a bunch of code that I wrote. There’s a prototype of an analysis system.

Paul: Did you release it? Open-source it?

John: No, I really should, it’s like something I’ve been meaning to circle back to because I did a prototype analysis which was like a clone detector based on like some interesting research that had come out of Davis, and it was like really good, actually.

It worked really well, there was a lot of stuff there that would’ve been interesting but taking that interesting research idea and turning it into a real product, was a lot of work. 

That was that last-mile problem in static analysis, going from, “Here’s an interesting research problem,” to, “Here, a polished commercial product that a company can use on their code base and derive value out of,” that’s a huge leap.

That was something we learned at Coverity and something that I rediscovered in this process. Those two things together, it was more than enough information to let us know that we needed to work on something else.

Paul: If he has to answer why are we making our own language, Ellen, this is the thing.

Edith: Are you really making your own language?

Paul: Yes we are, but the reason is if we don’t make our own language then we have to deal with someone else’s.

Ellen: I don’t think that’s the only reason.

Paul: It’s not the only reason but like…

Edith: I thought the reason was that Paul likes languages.

Paul: No,

it’s just easier to build a new language that is designed for exactly what you want than to try and prove properties about someone else’s language. 

John: I think the other interesting thing is, another reason that it’s hard to commercialize a lot of the research that I did. It’s much easier for people to adopt new languages now. I think the thesis when I was in grad school was like, “Oh, it takes like 30 years for a language to hit mainstream.”

And then you saw Java where I think the increment was like maybe 10 years or something like that. I might have my history a little bit wrong on that. But then new languages like Go, all of LaunchDarkly is written in Go and Go is not that old a language.

Edith: John, was it hard to give up the idea? I mean, this was your research, this was something you spent a lot of time on.

John: I still think it’s an interesting idea, the specific approach we were taking to it, and I hope that somebody pursues it. because I think software development ultimately will be better off if we had that kind of information, the kind of thing I had envisioned.

But it was easy to get rid of it as an idea, because going down those familiar paths of writing those parts of analyses that I’d done before at Coverity and in my thesis work, I was like, there was too much familiarity with it. And I really wanted to do something new and different.

Edith: I thought it would be a harder conversation when we had that conversation, because it was like, “Hey, maybe we should push a couple more months and see if we can productize this, I don’t want to give up too soon.”

Paul: You said that?

Edith: I remember it, because I was like, “I don’t want to. We’ve put so much work on it, I don’t want to give it up too easily. Let’s make sure we’re on board with giving it up.” Do you remember that?

John: Yeah, and actually that was right around when I left my company. And it was really hard because I went for about, I guess it was about a year, I’d forgotten the timing of it, but I went for about a year knowing I was going to leave my last company and feeling like we had something that I was anchored to that we were going to work on together.

And that was going to be the thing, and then I knew I was going to leave. I put my notice in and we made this decision to pivot and do something else. We were going to scrap that code base and it was a little bit hard because

 I was leaving and I couldn’t really articulate what I was going to work on next, because I didn’t have a concrete idea. So it felt even riskier to be leaving a good situation. 

Ellen: What did you tell people?

John: I had all sorts of hair-brained ideas. I think I told people that I was working on a VR-related idea, which I wasn’t.

Paul: It’s interesting, the thing of not wanting to do what you’ve done for so long that I feel once I’m done with something, I’m done. I’m not going back to it.

I think you were saying the same thing about the travel startups.

Edith: I do not want to do any more travel startups.

Ellen: Me neither.

Edith: I was at Tripit for a long time.

John: Oh, that’s so funny, two static analysis founders and two travel founders.

Edith: By the way, my middle name is Ellen.

Ellen: I did not know that.

John: My middle name is Paul.

Edith: And many people in this room are wearing LaunchDarkly shirts.

Paul: Many.

Edith: Nearly all, and Paul, looks lovely on you.

Paul: Thank you, thank you. It always does.

Edith: Ellen, did you have to make a similar leap of faith to join Paul?

Ellen: Yeah, actually. I don’t think I could’ve picked worse timing to have a startup. At the point in time when I met Paul, actually,

a mutual friend had emailed me saying Paul was looking for a co-founder. I said not to introduce me because I was afraid I would want to do it. 

My husband and I had just bought a house in Boston and I had this company and I’d just built this team I was very proud of. We’d just raised a bunch of money, and I was getting married in August.

I think there were some other things going on. It just was not great. But I think when I looked at it I said, “Oh, this is exactly what I wanted to do in three years. I don’t have to wait for three years. I can just do it right now.” And that seems like the better answer.

Edith: I think we did the reverse, we just kept postponing and postponing it.

John: Then we ended up picking a similarly bad set of timing. I’d just had a kid and left a comfortable situation. I think when it’s the right time, it just feels like the right time, regardless of like how it might seem externally.

Paul: Your wife’s a doctor. I feel you’re going to be okay.

John: Yeah, that’s true.

Ellen: The way I did it, I think it was a more slippery slope where I was like, I will regret it forever if I don’t do this. Probably it won’t work. How can I do the things to make it fall apart the fastest?

Paul and I had met on a Friday, and by the next Friday you had already flown to Boston and were living in my house and we didn’t know each other at all. 

Edith: Did he bring his LaunchDarkly shirts, though?

Ellen: He did.

– Yeah, like the first day we hung out for, I think, five hours. And that was a lot.

– Yeah.

Paul: I had gone through this process with a bunch of people before and I had not got to the point of five hours with someone actually feels like a pretty good experience.

Edith: I guess we were the reverse. We just kept postponing and postponing it. Some of it was me and some of it was you.

John: Yeah, because we were both in relatively comfortable situations in a lot of ways.

Paul: If you hadn’t postponed it, would you have come up with feature flags?

Edith: I don’t know.

Paul: It feels like feature flags are, not new but starting to catch on around the time that you started building.

John: I think if we’d come earlier it would’ve been too soon, for sure, for the way we built it. I think if we’d started the company five years earlier…

Edith: People were still shifting to the cloud at that point.

John: Yeah, I think we probably would’ve had to build a completely different product that was not SaaS-focused, probably like an open-sourced play or something like that.

Paul: You probably wouldn’t have realized that feature flags was the thing to build, either.

John: No, absolutely not. It was the experiences that Edith and I had both had, the size organizations that we had, it drove us to realize where the need was. And so without that experience, I don’t think we would’ve built as good of a product as we built.

Paul: You’d probably have tried to build “CI as a Service.” It was very popular in 2012.

Edith: It’s funny because we knew we wanted to start a company and we’d always be bouncing ideas off each other. I would say, “All I really how is to build software. I don’t have any deep domain expertise in,” I don’t know, “autonomous vehicles.”

Paul: Something about plants, right?

Edith: I was at a plant company but that was a bad one. But then it turned out that knowing a lot about software was actually really valuable.

John: Yeah, because

not only do I only really know about software, I only know about how to build software for other software developers. 

Paul: Edith, did you talk to any other potential co-founders?

Edith: No.

Paul: Ever ever think about it, either?

Ellen: How did you know you wanted to start a company together?

Edith: I was just like, John would be awesome.

Paul: How many years did you know each other by that point?

Edith: A long time.

Paul: Like 10?

John: Ish.

Edith: Ish. Well, we had briefly that other guy who I won’t name.

John: We thought about bringing a third person in around like a specific idea we had. He was a really brilliant person, but at the end of the day he wasn’t invested in the same way that we were. I guess I didn’t talk to anybody else, either.

Edith: It’s weird because I know that, Paul, you interviewed probably 50 people.

Paul: 49 people.

John: Were you literally the 50th person, Ellen?

Paul: Or maybe the 49th, I’m not sure.

Ellen: We can look at the spreadsheet.

Paul: There’s a spreadsheet.

Edith: See, I had the opposite where I had been working with a lot of people for a long time. And I could never find enough conviction to start a company with any of them or any of the people I met socially who I knew who started things.

I had come to a conclusion of, basically the shape of the person I wanted to work with and what I wanted them to be good at or not good at. So then I could pattern match that pretty quickly. Where you had the pattern from having talked to 49 people, and I had it from knowing a lot of people and knowing it fit.

Paul: Well, I first went through the people I knew, and some of them are in the 49 but I realized that either there wasn’t anyone who was ready to go now or there wasn’t anyone who was the right fit.

Ellen: Yeah, I didn’t talk to anybody else.

Paul: That’s the way you’re supposed to do it, right?

You’re supposed to have someone that you’ve known all your life and they’re the perfect co-founder for you. 

John: I think Edith is probably the only person that really believed in me as a potential co-founder.

Edith: Well, of course, you were perfect.

John: Well, thank you.

Paul: Aw, you guys.

Ellen: Very cute.

Edith: No, I had this deep conviction. I’m like, “John is the guy.”

John: But you were the person with that deep conviction, so that’s why we’re here.

Edith: I was like, “I will pry him out of his company and he will come do this.”

Paul: So did it take much convincing?

John: No, not really.

Edith: Some.

Paul: It sounds like you were coding on this idea part-time while you still worked. So you didn’t quit everything and then start doing it, by any means.

John: There were timing reasons for that. I think there was a conviction that I was going to leave. I think when I talked to some of the people that Edith talked to afterwards, I realized that she was worried that I wasn’t going to leave.

But there was never any point where I wasn’t going to leave. I was going to leave as soon as… We actually incorporated a little bit earlier, before we really started working on anything. When we signed the incorporation paperwork I knew I was done, I was going to leave. And the timing was just around other external factors and stuff like that.

Edith: He’d just had a baby, so I wanted the baby to be at least six or nine months old.

John: Which is really good because having a two-month-old and a startup, it’s crazy. It’s not fun.

Paul: I like the other thing, the thing with the process where you find the ideal person. Because you do find the ideal person that way. Well, I think most people don’t.

Ellen: Thank you?

Paul: There’s a compliment in there somewhere.

Edith: What has it been since, you’ve been working together for six months now?

Paul: A little over.

Edith: What’s funny is, even after John and I started the company and we’ve known each other for decades, we kept document of stuff we didn’t know about each other.

Paul: Oh, we just found out some stuff ourselves.

Ellen: Just funny stuff.

John: This seems like a really good venue for the two of you to share some of those things.

Paul: My one that we just discovered this afternoon, or that she didn’t know, is that

on a scale of one to ten, I basically live between four and seven. I never hit ten, but I never go below a four, either. 

I’m generally just sitting in a five or a six, just generally content.

Edith: Sometimes you seem really morose.

Paul: It’s probably a four.

Ellen: I learned that I’d been interpreting the four as a one, and vastly overreacting.

Edith: Sometimes you seem really down, Paul.

Paul: That’s a four.

Edith: I’ve never seen a seven.

Paul: Yeah, sevens aren’t around that much. I’m usually at a five or a six.

Ellen: Term sheet arriving.

Paul: Oh yeah, that was good one.

Edith: Is is true that you’re in The Commitments?

Paul: Ah, yes.

Edith: I wasn’t sure. Is this just an Irish cliche?

Paul: No, that picture is this thing Ian tweeted. That was actually me at seven years old.

Edith: Here you are slumming it as a developer when you’re a movie star.

Paul: Yeah, I know, my 83 pounds went exceptionally far.

Edith: How did you end up in The Commitments?

Paul: Oh, that’s 83 pounds, that’s what I was paid for the thing. Because we used pounds in Ireland back in those days. Sorry, what was that?

Edith: How?

Paul: How did I end up there? My drama teacher, I think she was married to the casting person for the movie. Obviously it was filmed in Ireland and they needed people, we were at, me and this other guy, were going to be page boys. And then they cut the wedding scene, so they just threw us into some other stuff.

Edith: You should put that in your LinkedIn profile.

Paul: Yeah, “Star of The Commitments,” that would be good.

Edith: I was on Reading Rainbow, which is an American TV show.

Paul: What were you doing on it?

Edith: I was talking about my idea of a perfect person.

Ellen: What was the perfect person?

Edith: I said it was somebody that could eat a taco without spilling anything.

Ellen: Yes.

Paul: John, obviously a question here, can you eat a taco without spilling anything?

John: Not even close, no.

Edith: Actually, no, that wasn’t what I said, it was my friend Ian said, and I found it more memorable than what I actually said. But somewhere there’s a video clip of me. Do you know what Reading Rainbow is?

Ellen: Yes. LeVar Burton visited Kickstarter when I worked there, it was great.

Edith: Do you know what it is?

Paul: Vaguely.

Edith: John, explain what Reading Rainbow is.

John: That’s a kid show, LeVar Burton, it’s great. There’s a catchphrase.

Paul: Did he wear his visor?

John: No it was like pre–

Paul: Oh, pre.

John: Yeah, pre-Geordi LaForge. But I think that it was pretty popular when we were younger. I actually didn’t know that you were on Reading Rainbow.

Edith: I was around seven or eight.

Paul: But what else went in the spreadsheet? What else do you know about each other that you did not before?

Edith: Allergic.

John: Oh yeah, I’m highly allergic to everything. This started in my mid-20s or so. I was not highly allergic to anything and now I’m highly allergic to many things.

Edith: That he actually started a degree in design.

John: Well, in human and computer interaction. I was a PhD student when I went to Berkeley. First I was in human and computer interaction, and then I shifted over to PLs after that.

I spent two years doing HCI stuff. It’s really hard, actually. I think a lot of people have this misconception that HCI is like the softer CS, but it’s incredibly difficult and incredibly hard work.

Edith: John also has a minor in statistics.

John: Yep, you had to do a lot of course work at Berkeley for the PhD. So I ended up having to take a bunch of grad-level statistics in math. And that’s where I learned that CS is very easy compared to math, especially grad-level math.

Paul: It’s interesting because all of the PL papers, there’s no statistical vigor at all. None. We put an average, we put it on a graph, we’re done.

John: Yeah, that’s because when people like you and I go into like those actual math classes, they get blown out of the water by brilliant mathematicians.

Edith: This was handy because I actually didn’t know that he’d gotten this minor in statistics until we’re starting LaunchDarkly and there started to be more harder math with bucketing and putting people… He’s like, “I got this.” I’m like, “You got this?” He’s like, “Yeah, I got this minor in statistics.” I’m like, “That’s pretty cool.”

Paul: I keep forgetting that Ellen went to HBS. We were looking at some finances I don’t remember… Oh, it was the 409A. I was explaining what I had done from my previous 409As, and she’s like, “No, no, I know how to drive this model from scratch. It’s okay.”

Ellen: I believe it was, “This is how you do a discounted cash flow,” or, “This is why there’s a discounted cash flow.” And I was like, “Yes, we can make one right now.”

John: I don’t understand what any of those words mean.

Ellen: I didn’t understand what any of the static analysis was, so it’s fine.

Paul: You just wear the pants, it’ll be fine.

Ellen: I did have a moment at HBS where they decided one year that every single HBS student should have to start a startup as a class, in groups of five, which I feel does not model proper co-founder behavior. But fine.

I ended up having to pitch this fake startup with five people on stage to 1,0