The Hazards of Consensus #
The most long-term success tends to come from reducing, to the greatest extent possible, the need for agreement and consensus.
A simple micro-example of this is the lunch-time behaviors of two different companies I’ve worked at: Joyent and The Startup. (“The Startup” shall remain nameless, because I hated it there, but I really liked a lot of the people involved, and don’t want to accidentally blame them for any of the badness. It’s sad that it went the way it did, but I am thankful for the experience I got out of it.)
At Joyent, lunch happens at a reasonable hour. (Since I am in the office most often on Tuesday, it’s signaled with a warning klaxon.) It is a leisurely 60 to 90 minute affair, where we socialize easily. Subjects of conversation do touch on work quite a bit, of course, but equally common are baseball, the politics of Crazy Town, what’s wrong (and right) with Computer Science education (or tech in general), and the latest Hacker News troll fountain.
If we all happen to be in the mood for the same type of food, we’ll sometimes go together. However, more often than not, everyone gets takeaway from a nearby spot, and we gather around a big round table. If we start walking in the same direction, but then I want a sandwich and you want a taco, we part ways, and meet back at the office.
People come and go, extremely low-pressure. We disagree more than we agree, but we easily move on from disagreements and find a new subject to toss around. There is no governance or agreement, no hard feelings. Just a loose group of sovereigns who enjoy one another’s company.
Not only do we not have to all agree on the same thing for lunch, the disagreement can often yield benefits for everyone. If you get tired of the same old offerings, and head out in a different direction, you might find a new menu that becomes a crowd favorite.
At The Startup, lunch was catered. (We provide you lunch! Hooray! Come work here!) The food wasn’t terrible, and I believe that it was well-intentioned by the founders, but I’ve since come to view catered lunch as a huge red flag at a company. My coworkers ate fast, and only discussed work, if they didn’t continue working at their desks while they ate. This was viewed as normal.
One time I decided to leave the office and bring something back (mostly because I just wanted a walk, having not seen the midday sun for a week). I asked a few of my teammates if they wanted to come with me, and the answer was “Um, no, I probably shouldn’t…” When I returned, no fewer than 5 people asked me nervously if I was ok, if anything was wrong, where I went, what was going on. It was SO weird.
Something as simple as a sandwich wrapped in paper showed that I wasn’t part of their unified tribe. To make matters even worse, most engineers there were in the office around 7:30 in the morning, stayed for the catered dinner (!!), worked another few hours, and then stuck around drinking beers in the office together until 23:00 or midnight. Having bags under one’s eyes was a mark of pride.
Maybe I’m just lazy, but I really only have about 5 solid hours of focus in me on a given day, and those hours suffer considerably if I don’t get the proper physical, mental, and emotional rest. So, it was not a great fit. I quit after a few months. A year (and 2 pivots) later, they were sold for a fraction of what their investors had put in, never having reached the general release of their product.
But back to the lunches. There’s an important point there.
Human societies function best when they are structured in such a way as to tolerate–even celebrate–disagreement. If the society can be improved by disagreement, then that’s even better. It allows for freedom, self-determination, experimentation, and a pervasive sense of safety. If you and I can disagree comfortably, then I don’t have to worry that we might disagree, which means two things: I can be candid with my honest opinions, and I can be open to yours. I don’t have to resist your opinions if I disagree, and I don’t have to guess your reaction before I share mine.
Of course, clearly some disagreements are not tolerable. If you and I disagree about whether to put a knife in my chest, there’s no way we can both get our way. Designing groups that can benefit from disagreement is non-trivial, and usually happens by accident. In the political space, this usually means some form of constitutional democracy, but neither “constitutional” nor “democracy” will ensure a healthy and inclusive culture or guarantee that experimentation will be promoted. Sometimes a limited dictator is safer than a powerful voting bloc.
With npm, I actively sought to design a software system in which we could easily use one another’s Node.js modules, without having to agree on which Node.js modules to use, or the best way to organize them in our projects.
I wish that I could claim credit for the “small modules” design philosophy or the many best practices and popular resources that have been created, but in truth, I can’t. All of those things were designed or discovered by other people.
When everyone else is applying force, it’s best to reduce friction. People were writing a lot of cool stuff, but having to agree on conventions and techniques was putting sand in the gears. The downsides of ceremony and dependency hell are very difficult to appreciate ahead of time, and the choices that lead to that place solve real problems early on. With npm, I opted to not solve some problems so that I could solve others. By sticking to the rule of “minimize the need for consensus”, we’ve found much more interesting problems (like how to handle massive growth).
The next time someone says that they like CoffeeScript, or that they don’t like Promises, or whatever other arbitrary preference you believe is “wrong”, instead of picking a fight with them, just be thankful that they’re not the boss of you, and that you can like different things, and still be friends.