In a startup, founders basically have to figure out how to optimize for a relationship that lasts for 10 years. That's a crazy thing to do with someone you might only know for a couple of months or have only known in a work setting.
The thing is, the only models for understanding that kind of relationship actually come from our parents. So let's start with some marriage research.
This is John Gottman. He studied marriages in Seattle and has been featured in This American Life and a bunch of different places. He has a cool magic trick: he can watch a couple fight about something for 15 minutes and predict with 85% accuracy whether they'll be divorced four years from now. If he watches them for an hour and has them also share their hopes and dreams, his prediction rating goes up to 94%.
This is the same video they would show to priests, psychologists, psychiatrists, marriage counselors, and successfully married couples. They don't predict better than random chance. But John has figured something out. There's something about the way we have an argument that determines longevity.
One of the most surprising things he discovered was that successfully married people who last a long time don't never fight. Turns out everybody fights, and we all fight about the exact same things: money, kids, sex, time, jealousy, and the in-laws.
The thing that's interesting is that I think all of these map out to the stuff we're going to fight about in a company. You with your co-founders are going to have these issues. What's nice about knowing everyone fights and that you know what you're going to fight about is that we can make a plan for figuring out how to deal with this.
The other thing that John Gottman figured out is that there are four major things we want to avoid when we're fighting. When we do these things, they create leading indicators that the relationship is in serious trouble.
This is basically when you're talking with someone and you're like, "Hey, I have a serious concern about this bug that we're trying to fix, and I'm really worried about this thing, and I'm not sure that we're going to be able to deploy on time." And someone comes up and says, "Well, you know what I don't like is the fact that you leave a bunch of dirty dishes in the sink."
Criticism is basically this idea that we don't fight on one topic. We start trying to bring all these other issues into play instead of addressing the one issue at hand.
This is pretty easy to understand. It's intention to insult. So basically I say, "Hey, I'm worried about this bug and we're not going to be able to deploy on time." And someone says, "I don't like your face." That's contempt. What you want to avoid is making things personal because we're in a business.
This one is kind of easy to understand. It's when someone doesn't own responsibility about the problem. We can't move forward because someone won't admit that there's a problem out there. We defend that we haven't done anything wrong, and therefore there can't be resolution between two people.
This one is super dangerous. It's when basically you're like, "Hey, I got a problem," and the person just walks away, won't engage, won't talk to you. So there can be no way to create any kind of resolution.
Just as you wouldn't do this without a plan, we want to make a plan. I'm going to talk about four different things that we can do that help avoid and protect us from those four horsemen.
This feels pretty straightforward, but you want to do this early in the relationship with your co-founders and in the early stages of your company.
Here's our list again of the types of things that we might have problems with. In the early stages of the startup, let's say Adora and I are doing a startup together. It's just her and me. Then what you want to do is just kind of say, "Oh, who's responsible for what stuff?"
What this will do is if there's a problem in that category, then that person that we have assigned ahead of time to be in charge will be the ones that will ultimately either make the decision or are responsible. This protects us from defensiveness.
Notice here on jealousy, this is about competition usually. In the early stages of your startup, you should not be worried about competition. Competition is not usually what kills you in the very early stages of your company.
As your company ages, it might change and look something like this. You'll assign things to different positions and heads, and as a result, when problems come up, you know that sort of is delegated.
What happens if things go out of hand even up at those levels? Well, basically what you want to do is decide after you delegate who has ownership, determine what is success and failure.
You want to know ahead of time: "Hey, we've divided up the task, but we also want to know when is there going to be interference with a person that is supposed to be leading these decisions? What is considered success enough that we shouldn't be interfering? Just let them do what they think is best. And what is considered really bad so that we have to interfere and something has to be done about it?"
So in this case, good examples would be like: "Hey, if we've successfully fundraised, we don't need to talk or replace the person that's responsible for that. If we're shipping on time, if we're rated top three amongst our peers, or we've built a referral program that's working, we don't need to be criticizing the person that's working on the stuff. They're doing a good job."
On the corollary, we want to define what are the things that are going to trigger conversations—really hard conversations. Like, "Hey, if we hit this area, we need to put the brakes on and we need to discuss what's going on and actually try to resolve these problems."
A lot of people like to delegate stuff, but they don't have a way of saying, "Hey, when are we going to have a conversation about this when there's trouble?" These are really easy to do, and the reason you want to do these early while you're sober emotionally is because once you get angry and emotions come into play, then you might not be thinking rationally.
Now ultimately, usually it's the CEO in the company who has final say. You as a team can decide differently how you want to resolve it, but ultimately whoever's the CEO usually is the one who resolves it. If there's problems with the CEO, then it's the board. In the early stages of startup, the board is usually composed of just the founders, so you have to ultimately work it out.
This will protect you from stonewalling. What I mean is, what is your attachment style? There was all this research done in the 1960s about how people approach relationships, and basically it was determined that there are three major types.
There's a secure attachment style, and that means basically it's like, "Hey, you know what, I don't have a problem going up to people, relying on them and having them rely on me, and sort of us creating a relationship. I don't mind being vulnerable, and I don't mind other people being vulnerable with me."
There's an anxious style. So there's a type of person that will be like, "You know what, I kind of don't get enough love, not as much as I want. I kind of want to hold on to people, and I kind of want to have people constantly confirm with me that they want to be with me. I feel like it's a little difficult."
And then there's another type, the kind of person just like, "I find it kind of difficult creating relationships with people, and I kind of want to run away sometimes because it's really scary, or I'm worried that I'm going to mess it up."
The thing that's super important here, especially with your co-founder, is you want to know your co-founder's attachment style because that's going to dictate how you are going to be able to resolve and understand your differences.
What it turns out, oddly enough, is that an anxious attachment person and an avoidant attachment person—these are the two most common in the world. There's not that many well-developed secure people out there. They tend to want to be with one another.
So you have someone who needs space to make a decision and to process problems and tension, and someone who needs validation constantly to process conflict and issues. When those two people are together and they don't realize what the other person needs, they don't realize that they're going to have to bend to sort of make it work.
There are lots of good books on attachment styles. There's a wonderful Wikipedia page that covers it. I would recommend watching YouTube video from School of Life. What I would highly recommend is basically understanding that if you're with someone that is of the opposite type, you're going to have to do work either to reach across the aisle.
If you're an anxious person and you're talking to an avoidant person, you just have to realize like, "Oh, that person needs space, but that doesn't mean they're running away from you." And if you're an avoidant person with an anxious person, if someone needs your attention or if you need your space, then you have to let them know. It's like, "Hey, I'm going to be back. I realize that you're going to need an answer for this. I'm going to go away, I'm going to figure stuff out, and I promise a time that we will deal with this."
This will protect you from criticism. When you're emotionally sober, it's the best time to create a process for dealing with disagreements. The reason you want to do this is because once you're upset and angry and filled with emotions, you are not going to be thinking straight. The odds are you might say something you regret, you might say something that you don't mean, and the other person might do so, and then you will have a much different problem than the bug not being fixed and deployed on time.
One example of this comes from a company called Matter. They created a spreadsheet for dealing with disagreements. It's a disagreement decision framework. Basically, it just talks about when we have a disagreement, we should just document it. This helps make things really transparent, makes us understand both sides very clearly, we talk about the different options, we say who made the decision, what the decision was, the date it was done, and then rationale.
When we walk through this process, if we've decided this ahead of time, then it means that we are not afraid when disagreements come up. It's like, "Oh, we have a process for dealing with this, and we will figure it out by filling out Excel."
There are lots of different ways to do this. You don't have to follow their very specific framework. You just have to agree ahead of time what you want to do. So therefore when you are upset, you just go, "Okay, great, we have a process for this, and process says go have a timeout or eat a bologna sandwich or take a nap first, and then we'll figure out what we have to do."
It could be a process where it's like, "Hey, if there's a real disagreement and both sides feel equally strong, we will flip a coin, and then that will be the decision for the test of time. We will let lady luck decide." It doesn't matter. You just have to both agree.
This strategy will protect you from contempt. The way that you avoid making things personal is you have to figure out a way of communicating with another in a way that will not be threatening.
There's an amazing book on this called Nonviolent Communication by Marshall Rosenberg. It helps you be honest with other people without criticizing, without insulting, without putting down other people. The magic comes in the structure that feels somewhat fake for people who are not into being touchy-feely.
Basically, when you're giving some kind of criticism, you want to have it in this format: "When [observation], I feel [emotion] because I'm needing some [universal need]. Would you be able to [request]?"
We're just going to break down each one of these different parts, and they're all tricky. It's a thing that a lot of people will try to do, and you'll spend your whole life trying to get really good at.
You need to make an observation versus having an evaluation. What you want to do is start your disagreement or criticism by anchoring it to something that is concrete. You do not want it to be something that is connected to opinion. It should be something that you actually saw or heard because therefore you can't disagree with something that actually happened.
An observation would be like, "You said that you'd send that document last week and I haven't received it." That is a great observation. An evaluation that someone might say instead in the heat of the moment is, "You're lazy." That kind of feels like an observation, but it's not. It's evaluating the person.
Give you another example: "Your work is sloppy" is not objective. Instead: "Hey, three of the numbers in this report were inaccurate." That's where you want to start.
"You're always late"—you want to be really careful because that's a generalization, it's an evaluation. Observation: "Hey, you arrived 10 minutes late to the meeting this morning."
Evaluation: "You ignored me." Observation: "I sent you two emails and I haven't received a response."
Notice when we start with observation, we start with a fact that can't be refuted, and so we're not going to end up arguing about something else. Notice all those other evaluations, they immediately will trigger an emotion in you, and so that's why you want to be really careful that when you start this criticism, you don't start with one of those.
We have to talk about our emotions. I saw this irrefutable observation and it made me feel something. What we have to be really careful of is not saying thoughts but instead talking about feelings.
An emotion will be "I feel frustrated right now." A thought would be "I feel that you aren't taking this seriously." The way you can tell if something is a thought or a feeling is you substitute the phrase "I think" with "I feel" and it still works.
"I think frustrated" doesn't work, so that's a feeling. "I think that you aren't taking this seriously"—oh, that's a thought.
There are a couple of emotions that we have to be particularly careful of. One is anger because anger is usually tied to a bunch of hosts of other things. When someone says "I feel angry" or you realize that you feel angry, you want to be really specific about what's causing the anger, what's triggering it.
The other tricky emotions are evaluative emotions, and usually what you need to figure out is what underlies that evaluation. I'll give you an example: "I feel blamed"—someone else is evaluating me. The impact actually is "I feel scared." Someone is blaming me, and so I feel scared.
It takes a lot of work to understand that when someone is giving me some kind of judging feeling, what is the core root of it? Other examples: "I feel judged"—the actual impact is "I feel resentful." "I feel misunderstood"—the impactful statement is actually "I feel frustrated." "I feel rejected"—the real impact is "I feel hurt."
Every negative emotion lies an unmet universal need. What that means is that when you're feeling one of these frustrated or blamed or scared or hurt feelings, there's something that's missing that you're going to need.
The thing that's really tricky about universal needs is you have to be careful of realizing: is it a strategy or is it a need, and is it truly universal?
I'll give an example. You might be able to say "I need a sandwich." That is not a universal need. You might say "I need a sandwich to give me nourishment." That's more like a strategy.
You might say something like "I need you to copy me on every single email." But the thing is that's not a universal need. That becomes very specific. A universal need would be "I need some transparency about this process."
You have to be careful of not making needs about something that's very specific to yourself or just that situation because once it's a universal need, then it's something that everyone can agree that everyone should have.
Other universal needs are like "I need support." The way you turn it into a universal need is by saying something like "I need support" because not everyone needs support from Jerry, but everyone does need support. If you say "I need support from you," it stops being universal.
At the very end, so basically we said like, "Hey, I've noticed something that can't be refuted, I've told you about a feeling and how it impacts me, and I told you that it results in some universal need that we all can agree that we need to have." Now we get to saying what we'd like to have changed as a result.
What you want to make is a request, not a demand. The difference is that a request is an invitation to the other person to meet our universal needs. It's much easier to be able to do than to say "I order you to do something."
What we want to do is make it very specific. "I request for you to be more respectful" is not that great because who defines what's respectful? My version of respectful might be different from someone else's. Your request should be something like "I request that you arrive to meetings on time."
Say what you want, don't say what you don't want. A lot of people will say "I request that you don't dismiss other people's ideas straight away." The thing is it doesn't indicate the behavior that you do want, and so it becomes really difficult to act on. A better one would be "I request that when a team member shares an idea, you ask two or three probing questions before sharing a conclusion."
And then stay curious. Sometimes you might make a request and someone might say no. What you need to do is not just freak out that the whole process isn't working. The idea is actually to be like, "Hmm, maybe I haven't put this request in a way that can meet more needs than just myself. Could I do this in a way so that they can understand and be on board for everyone to be involved?"
We all know what technical debt is, right? When we're building out software really quickly, sometimes you're like, "Well, that's not going to scale really well, and it's going to be dirty and quick, but I'm going to get it out the door, and I'm just going to put that in the back of my mind. It's something I have to fix later."
Well, in our relationships with one another, you will incur emotional debt, and unlike technical debt, you really don't want that to go for very long. You want to pay this down every day.
It turns out also in John Gottman's research that it wasn't that people who were really good at being in a marriage only thought about really big things. It turns out they would immediately bring up stuff even when it's really tiny or small. They would never let a small thing grow to be a medium thing and then eventually a big thing. They immediately will talk about it.
It's like, "Oh man, can you close your mouth while you're chewing? Real quick, it's just kind of bothering me right now." And then do it in a way that's respectful.
So when you're with your co-founders and you're in this really sensitive relationship and you're finding stuff that's being really troubling, you can communicate those needs really quickly, and you will prevent those small things from becoming big things.
The best way to start doing this is to practice. We call these 'level three' conversations.
Level one is that informal conversation we have with other people where it's just data exchange, passing information back and forth.
Level two conversations have some emotions, talk about some things that are personal.
Level three conversations are relational. They're engaged with something that's happening right now between two people that is super important. It is a deep dive into what might be really troubling or what might be really mattering to two people.
In a startup, there's a lot of things that are going to matter to all of the people working on the company. Let's go through some examples of things that you can do after this talk.
Some good ones are: "What are our short-term goals for the company?" You'd be surprised at how often people are not on the same page about this. "Are we using the right metrics?" We've got lectures on those. The answer is I hope so by now. "Are we hitting our goals?"
"Who's responsible for what?" This is super sensitive. "Is it clear who is responsible for what?" Just have that conversation. "Do we agree that the current division makes the most sense?" This might be super simple answers, but if there is any kind of disagreement, we want to hash that out.
"Is our workload distributed in an optimal manner today?" "Do we all feel a high level of dedication and motivation right now?" Great thing to just check on every day. "What mechanisms are in place for providing feedback to one another?" "Have we carved out time for paying down emotional debt?" "Do we feel like we can have these level three conversations at any time?" "Do we have a process in place for thinking through this stuff so that we can be honest about where we are in our company?"
How to work together: everyone fights, so you want to make a plan. You need to figure out what's your attachment style, what's your roles, what's your goals, and a process before emotions get involved. Do it while you're sober.
Use non-violent communication to share honest feedback without criticism, and then pay down emotional debt on a regular basis. This is the most healthy way that you will make sure that things will not turn into a giant blow-up.
You can start having hard conversations right now. There's no doubt in my mind that there's probably some issue that the two of you or three of you or four of you or God forbid seven of you are not talking about.
[1] The "Four Horsemen" model comes from psychologist John Gottman's research on marriage stability. YC frequently references this in founder communication training as a framework to spot and address toxic interaction patterns early.
[2] Dividing responsibilities upfront and assigning ownership is a core idea from Amazon's "single-threaded leader" model and widely practiced in high-functioning startups. It helps reduce ambiguity and prevent blame during high-stress moments.
[3] Attachment styles (secure, anxious, avoidant) come from adult attachment theory in psychology. Founders who are aware of their patterns tend to be more emotionally resilient and better at navigating conflict in cofounder relationships.
[4] The "non-violent communication" (NVC) format is based on Marshall Rosenberg's work. It's become a popular tool in founder circles for giving feedback that's direct but non-hostile—especially when tension is high.
[5] The concept of "emotional debt" and level 1/2/3 conversations draws from coaching frameworks like Conscious Leadership and Reboot. Addressing small issues early helps prevent relationship breakdowns that sink startups.
[6] The "define success and failure while you're sober" idea echoes pre-mortem and decision hygiene techniques taught by YC, Stripe Press authors, and startup coaches like Jerry Colonna. It's about setting rational expectations before emotions flare up.
[7] MatterApp and similar tools (e.g., Almanac, Notion templates) are often used to formalize cofounder operating agreements—especially in remote or async teams.