Attempt to distill a working philosophy from this:
Or, in habryka's own words:
Do not spread yourself thin. And even more importantly, do not forget to explore.
Another category of what I suspect are increasing returns come from ceiling effects: if there is a ceiling, and you hit the ceiling, a rational observer will infer that your true value is noticeably higher than the ceiling. The lower the ceiling, the more the relative inflation. So you get an effect where returns are diminishing from 97% -> 98% -> 99%, but then at 99% -> 100% there's a sudden spike.
I've wondered if this is responsible for an apparent effect where adding a little bit of polish to a nearly-perfect design has an outsized impact compared to the naive diminishing-returns expectation of that polish being invisible.
This graph shows how many productive work minutes you get when adding an additional person who works 40 hours a week, to a team that spends 2-60 minutes per week on each team member to keep them in sync with the rest of the team. If you spend 30 minutes of everyone's time per person to keep then in sync with everyone else, then if you try to scale that to 80 people, you have gained no productive hours, as it takes 40 hours just to keep everyone in sync with your last employee:
I think the effect you’re describing is even stronger than this model/graph implies, because the 40 hours you’re gaining are marginal-employee-hours, while the coordination costs are paid in average-employee-hours.
An n-person meeting requires n updates, each with n people listening, resulting in n^2 minutes of time spent in the meeting.
This seems incorrect. Each of the n people having 1:1 meetings requires (n^2)x minutes spent in the meeting, but an n-person meeting only requires nx minutes since the speakers only speak once to all people. Of course, this is excluding back and forth, which will scale with the number of people, but not quadratically, since many people will have the same questions/comments.
40 people attending a one hour meeting spends 40 hours of total employee time. 20 people attending a one hour meeting spends 20 hours of total employee time. Total employee time is the main metric that matters
To tie the two comments together explicitly: an n-person meeting, with each person giving an update, requires nx clock-minutes, and each clock minute spends n person-minutes, leading to (n^2)x person-minutes in an n-person meeting.
Context: Post #5 in my sequence of private Lightcone Infrastructure memos edited for public consumption. Much of the advice in this might not apply to your situation, read with that context in mind.
Most things have diminishing marginal returns. I often repeat the Pareto Principle to others: "You can get 80% of the benefit here with the right 20% of the cost", which is a particularly extreme case of diminishing marginal returns.
But I think for much of the work that Lightcone does, the returns to effort are generally increasing, not decreasing.
To explain, let's start with the simplest toy case of a situation in which trying harder at something gets more valuable the more you are already trying: A winner-takes-all competition.
If you are in a competition where the top performer takes all winnings, then doing half as well as the other contestants predictably gets you 0% of the value. Indeed, inasmuch as you are racing against identical candidates that put in 99% of the possible effort, and your performance is a direct result of the effort you put in, all the value is generated by you going from 98% to 99%+. If you stopped at any point before then, you would gain nothing and all your effort would be wasted.
Winner-takes-all markets are particularly common in software, where the cost of distribution and the time to scale up distribution are approximately zero. If someone ships a better product than you, your users can often shift suddenly and all at once. But really almost every market is structured so that being the very best at something is a lot more valuable than being the second best. Inasmuch as performance is correlated with effort, this means the marginal returns to effort are increasing.
But winner-takes-all markets are not the only dynamic that causes increasing returns to effort. The next biggest factor is that coordination costs increase (roughly) with something like the square of the number of people involved.
Maintaining context on a project you are working on alone is easy. But when you bring someone else into the fold you suddenly have to think about how to split up the work, and coordinate on that work. This burden of coordination is often quite large. If you can avoid bringing in an extra person, the hours that you worked to prevent that are thus high return.
To make this point a different way: Imagine two people working 20 hours each, vs. one person working. A workforce of people that work 40 hours per week can tackle a much wider range of projects before they inevitably have to pay the overhead costs of coordinating. This means the labor of someone working 40 hours per week is more than twice as valuable as the labor of someone working 20 hours per week, i.e. returns to effort are increasing.
At larger scales, my guess is that the cost of naively coordinating a group of people roughly scales with the square of the number of people. Most naive coordination strategies involve keeping everyone in the loop on what everyone else is doing (resulting in n2 message-reads needed to achieve that). To illustrate this more vividly, think about the costs of a daily or weekly all-hands/standup. An n-person meeting requires n updates, each with n people listening, resulting in n2 minutes of time spent in the meeting.
This creates a strict upper bound on how large your teams can be without changing how you coordinate with each other:
This graph shows how many productive work minutes you get when adding an additional person who works 40 hours a week, to a team that spends 2-60 minutes per week on each team member to keep them in sync with the rest of the team. If you spend 30 minutes of everyone's time per person to keep then in sync with everyone else, then if you try to scale that to 80 people, you have gained no productive hours, as it takes 40 hours just to keep everyone in sync with your last employee:
0.5hoursemployee∗80 employees=40 hoursThis is the obvious reason why we structure large organizations hierarchically and into departments, where each level of hierarchy and subteam formation reduces the amount of time we have to spend coordinating with them (while producing many other issues which I have talked about previously which cost you in different ways but usually also gets worse with each marginal employee). Evaluating the marginal coordination cost per employee for an organization that from time to time undergoes reorganizations and changes its corporate structure would be a lot more difficult, so let's for now grant the (admittedly naive) assumption that coordination costs increase continuously for each marginal employee.
If we grant this, then this quadratic cost of course correspondingly creates increasing marginal returns to increasing the productivity of each employee (holding a given workload fixed) as with each additional productive hour worked, you get to reduce your team size, which staves off the costs of coordination for longer:
These graphs show how much value a business of 400 employees produces (for each one of its employees, per week), assuming that they can translate one hour of effective employee effort into $100 of economic value. As you increase the number of hours each employee works, value produced increases superlinearly, producing ~$300 of value for the first 30 hours, ~$700 of value for the next 10 hours, and ~$1,667 dollars for the next 20 hours. Our assumptions imply that this specific business would not be able to pay its employees anything at all if they worked less than 20 hours as coordination costs would exceed available time.
This model suggests (and I do believe it) that if you can trade how intensely you work across different periods of your life, that you will produce more value if you make more extreme choices. Working extremely intensely for your 20s and 30s, but working very little later on, is a better (economical) choice than working normal hours for most of your life (with many many caveats, some of which I will cover in the rest of this post).
Paul Graham covers some related ground in "How to Make Wealth":
And at this point I do have to issue a warning. Lightcone does not appear to be the kind of project that becomes exit-ready for you in 4 years. While I think the value we provide for the world is enormous, giving your all for 4 years here will not leave you with a nest egg that will last you the rest of your life and allow you to relax. Such is the fate of working in charity instead of business. This should obviously reduce your willingness to invest in Lightcone the way you would invest in co-founding a startup.
Meta: The below is more directly aimed at Lightcone employees, but I actually think it generalizes quite well! Beware of people telling you that it is of great importance to invest more into their specific project and to abandon your other responsibilities. There be many skulls on this path.
I do believe the things I am saying here, but I would be reckless if at this point I didn't highlight the obvious fact that I, as your boss, will tell you that working more intensely, and investing more intensely into the company, will be good for you. Ultimately, how much you decide to invest into Lightcone is a negotiation between you and the organization, and of course I, as the person with the most control, will have many reasons to arrive at the conclusion that the right choice for you, my dear friend, is to give it all up for the company.
And indeed, if an alternative to working at Lightcone comes around, or some other priority in your life starts looming larger, I think the arguments here bite against working here and to invest more into those other things. Do those things with the intensity that you brought to working here. Do not spread yourself thin. And even more importantly, do not forget to explore.
While locally each unit of effort might pay off more than the previous unit, you need to invest enough into search and exploration to find the right place to invest your efforts, which at a global scale, is still a much stronger predictor of a life well-lived than how hard you worked (even though both are IMO crucially important). So if you find that working intensely at Lightcone for a long time is making you worse at knowing that this is really the right place to invest so much in, then I encourage you to work marginally less, and spend more time exploring.
For example, I actively support creating space for people to work trial at other organizations while you work at Lightcone. You of course shouldn't expect that you can spend >20% of your weeks trialing at other places, but honestly, 5% (i.e. 2-3 weeks a year) doesn't seem crazy to me.
Now, it's also important to clarify when it is not the case that marginal returns to effort are increasing, as often they are not.
Some common situations where you should expect marginal returns to effort to be decreasing instead:
Within your work at Lightcone, you will often want to half-ass a project. And sometimes even the total of all responsibilities given to you at Lightcone will not benefit that much from marginal effort. The times during which that happens are often good times to relax, explore more, and stock up energy for the times when the returns to effort do be increasing again.