6 ms·
> At first I was puzzled. How could things be fine at 60 startups and broken at 80? It was only a third more. Then I realized what had happened. We were using a
by jefftk 4y ago
> At first I was puzzled. How could things be fine at 60 startups and broken at 80? It was only a third more. Then I realized what had happened. We were using an O(n2) algorithm. So of course it blew up.
This is neat, but isn't actually right. Each partner has to know each startup, which is, yes, O(n2) relationships. But there's no one who needs to do work proportional to the number of partner-startup relationships: each partner only has O(n) startups to keep track of. So probably the reason it blew up was just ordinary linear growth outstripping capacity: 60 startups was an amount most partners could keep track of, and 80 wasn't.
- jkaptur 4y agoYeah, the tech-nerd chestnut to bring out here is Dunbar's number. Not that that's necessarily the mechanism, but it's the thing to mention.
- travisjungroth 4y agoAppealing to Dunbar's number for org sizes is a strong signal for me I should de-weight whatever the person is saying. They're just parroting pop-science with reasoning that doesn't hold up to a second of critical thinking. If humans evolved to hold about 150 relationships in our minds, to say that an org has a tipping point of 150 people assumes the members of the organization know 0 people outside of the organization. Maybe this is approximately true for Amish communities. It is not close to true for startups doubling in size every year. The available "relationship slots" for your company is probably more like 10-25. If you want to say "things get weird at about 150+", sure, maybe that's true. But no need to bring up theories that extrapolate primate cranial capacity.
- bjelkeman-again 4y agoMy experience is that one can have multiple 100-150 person contexts and keep track of the people in them, i.e company, football community, friends etc. But it is harder when they get bigger.
- chalst 4y agoThis is basically right, but to hairsplit, it's an O(mn) algorithm, which is significantly different from an O(n) algorithm in that you can't help manage the number of startups by adding partners.
- pavon 4y agoUnless the number of partners also grew in proportion to the number of startups. Then the partners had 33% more startups to learn, but they were scheduled for 33% fewer meetings with each startup, so the total number of meetings needed for all the partners to learn all the startups was O(P*S) which is O(N^2), since P&S are both proportional to N, so it took quadratically longer for that familiarity to occur.