If you are only used to driving cars, it is hard to appreciate just how huge a force drag can be. The reason is that drag increases as the square of speed, so an object will experience 100 times the drag at 300 mph as it does at 30 mph. Not 10 times.
In Physics Can Be Fun, Soviet popular science writer Ya Perelman provided a dramatic example of the consequences of drag. With drag, a typical long-range artillery shell travels 4 km. Without drag, the same shell would travel 40 km.
Or 10x further. Which brings me to the famous 10x effect in software engineering.
If you haven’t heard of it, the 10x effect is the anecdotal observation that great programmers aren’t just a little more productive than average ones (like 15-20%). They tend to be 10 times more productive. A similar effect can be found in other kinds of creative information work.
Can you transform yourself into a 10x person? If you meet certain qualifying conditions (by my estimate, maybe 1 in 4 people do), I think you can.
The 10x effect arises from the interaction of thrust and drag on the fundamental playing field of your life, your calendar.
If you understand how this works, and learn to manage the interaction, the benefits are enormous. Managing the interaction involves a rule that is about as close to magic as I’ve ever encountered: you just prioritize longer tasks in your scheduling.
The bad news is that while drag management can be learned, thrust generation, I have come to suspect, cannot.
This is more than an analogy. It can be mathematically modeled. I learned this many years ago when I was working on certain scheduling algorithms for telescopes. The problem was simple: you have observation windows of varying lengths, and big and little observation tasks to fit inside them. My scheduler algorithms could very efficiently pack the schedule, reaching nearly 100% utilization of observation time. The catch was that in trying to maximize utilization, the algorithms packed in all the little jobs first. This led to a growing backlog of longer jobs. If a mix of jobs was streaming in continuously, the long jobs would never get done.
You don’t need math to understand why: little jobs can fit into big windows, but big jobs cannot fit into little ones. Here’s a picture.
I fixed the problem by keeping track of the ratio of long and short jobs, and inserting periods where I’d let long jobs jump the queue, sacrificing some utilization in order to drive down waiting times.
That’s one half of what’s going on: our natural tendency to fill up time leads to a backlog of longer tasks.
For the other half, consider the difference in content between long and short jobs in human work. It turns out that there is a very interesting correlation: longer tasks (4 hours or more) are more likely to be the productive, creative ones (like writing or coding sessions), while short tasks are more likely to be busywork. Call the former thrust tasks, and color-code them blue, and the latter drag tasks, and color-code them red.
This is the reason why Paul Graham’s anecdotal observations in Maker’s Schedule, Manager’s Schedule ring so true. Creative work requires longer uninterrupted work blocks. Most managerial work is short-duration busywork. The Maker/Manager distinction is too coarse for my tastes, which is why I use thrust/drag instead. Everybody has elements of both kinds of work and schedule in their lives.
If you do creative information work, I will bet you that your current to-do list will yield a distribution that looks something like this: lots of little drag tasks, and a few big thrust tasks.
How do you tell thrust and drag apart? (merely classifying longer tasks as thrust tasks would be both wrong and beg the question).
Turns out there is more than correlation between the length and creative production of a task.
Thrust items create high value. They are autotelic: they involve a mindful state of flow as discovered by Mihaly Csikzentmihalyi, which emerges when you are just beyond the edge of your current skill level, and have internalized the performance standards of the creative field, so that you are able to continuously monitor the quality of your own output via internal feedback.
When you start a session of thrust work, situation awareness takes longer to develop. Once you get going, momentum keeps building, and you only slow down and stop when you run out of energy. Momentum within a work session does not increase linearly, it increases quadratically or faster. So the peak momentum you can develop in a two-hour session will be at least four times the peak momentum you can develop in a one-hour session (note that you need a minimum window length: super-linear curves overtake linear ones only after an initial period where they lag; by my estimate, the crossover happens around the half-hour mark). This means you can do qualitatively different types of things over one two-hour session than you can over two one-hour sessions.
Some day I will try to mathematically model this effect and compute actual exponents, but for now you can just remember it via the analogy to physics: it’s like a snowball accelerating downhill: if you let two snowballs roll downhill for a minute, they will collect less mass together than if you let a single snowball roll downhill for two minutes. In terms of damage the snowballs can do, the one big snowball will pack a bigger punch after two minutes than the two smaller ones together will, after one minute.
And most importantly, thrust items are deliberate practice items. The more you do them, the better you do them. Your effectiveness increases at a compounded rate. Every coding or writing session is a compound-interest improvement over previous ones. The compounding goes on even when you are not actively working on the skill, as your brain processes and integrates your session-level learning.
This last point is why you can get to 10x type returns: such exponential gains always involve positive feedback loops, and thrust tasks contain two of them: the autotelic loop within a session, and the longer-term deliberate practice benefits that carry over between sessions. These two positive-feedback loops together basically create an addiction.
I’ve written two detailed posts (Daemons and the Mindful Learning Curve and The Calculus of Grit) on the phenomenology of thrust, so I won’t elaborate more here. But the net effect of the characteristics of thrust work is that the higher-value ones take longer. The longer a thrust-work session, the more valuable the output, with biological limits being the ultimate constraint.
Drag items create low amounts of value. And I mean here intrinsic value; not value in the sense of money. There are many drag items like hunting for coupons/deals that can have a lot of dollar value in terms of resources conserved or transferred in zero-sum ways, but don’t create any value.
Drag items are not autotelic because performance standards cannot be internalized (an example is bureaucratic paperwork: they often make little sense, so you cannot internalize quality standards and fill out forms “better” over time). Situation awareness develops quickly once you learn the specific arbitrary details of a drag task. Once you’ve figured out how to navigate post offices in a new country, switching into “post office” mode doesn’t take long, but for something like coding or writing, deep familiarity doesn’t really reduce the time it takes to develop situation awareness.
Switching costs for drag work tend to be high. Whereas writing or coding just requires booting up your laptop wherever you are, going to the post office or to a consulate to get a visa can involve high switching costs.
Drag items are also event-driven. Instead of internal momentum determining starts and stops, external events do. If you wait for an hour in line in some government department, only to find when you get to the window that the service provider is breaking for lunch, there is basically nothing you can do about it.
So you can put it all together yourself now: 10x performance involves two things: generating thrust and minimizing drag. Let’s consider each in turn.
Generating thrust means you have to be capable of some sort of mindful-learning deliberate-practice behavior.
In other words, you need a thrust engine.
If you suddenly have four uninterrupted hours to spare, and a great energy/alertness level, but are at a loss about what to do with yourself, it means you do not have a thrust engine in your life.
Can you discover your engines? This is an open question that I am currently thinking about. My tentative conclusion is that a thrust engine has two parts: a natural talent, and a capacity for meta-cognition with respect to that talent. A thrust engine is a developed strength (though not all strengths are thrust engines). Strengths psychologists seem to think that anyone with a talent can develop it into a strength simply by investing time.
I don’t think that is true. I think a separate capacity for meta-cognition is a necessary ingredient that many hard-working and talented people lack. So they make no progress despite putting in the time. Without the meta-cognition, you don’t get either the autotelic positive feedback loop within work session, or the deliberate-practice improvements between sessions.
Maybe I’ll change my mind as I think more about this, but right now I am pessimistic. I’d say maybe 20-25% of people are mentally capable of creating thrust engines in their lives. About half of them probably never discover their talents at the right times and/or cannot find the opportunity to develop it. So sadly, only about 1 in 10 people luck out and develop a thrust engine.
The good news is that if you have a thrust engine, drag reduction is as simple as a scheduling prioritization rule:
Clear your calendar as much as possible (empty out the “hard landscape” to use GTD terminology), create as many longer windows as possible, and start filling them with thrust work.
This is one of the few very simple rules in life that isn’t simplistic. It is also intuitive, which is why people seem to keep rediscovering it. Besides Paul Graham, Wayne Dyer has also made a similar point (using an evocative demonstration involving big balls, little balls and a large container).
The great thing about this rule is that it is self-reinforcing. The more you fill your calendar with thrust work, the easier it gets, because you are creating value that can potentially bring in resources for drag reduction. You will also find yourself energetically investing in drag-reduction infrastructure (creating this infrastructure is thrust-work, but using it is still drag work, which is why so many people who are great at inventing systems and processes are terrible at actually using them).
That bad news is that if you don’t have a thrust engine in your life, drag reduction will also become nearly impossible, because there is no source of energy. Everything from getting organized to losing weight becomes easier if you have a thrust engine humming. Not only do you have a reason to reduce drag (free up more resources for thrust), you have the means to do so.
Dealing with Drag Backlog
Prioritizing longer thrust work sessions will inevitably create a backlog of drag tasks. How you deal with them depends on whether you are cash-rich/time-poor, time-rich/cash-poor, or cash-poor/time-poor.
- Cash-Rich/Time-Poor: If you are cash-rich/time-poor, outsource everything you can. Hire people or services for every kind of drag work. Managing drag work is also drag work, so hire personal assistants to do it. Yes, it seems rather selfish to fill up others’ lives with drag work, but you can console yourself with two thoughts. First, there really are cases of “drag for me is thrust for you” types of work (though there is unfortunately a large “drag for everybody” category). Second, for many people the choice isn’t between thrust and drag. It is between destitution and drag. It can be a win-win at least in the short term. If somebody doing drag work for you discovers a thrust engine in their lives, be generous. Help them move to the 10x side.
- Time-Rich/Cash-Poor: If you are time-rich/cash-poor, your options are more limited. Procrastinate, let the drag work pile up, and then batch process them whenever you find the energy. At least you’ll save on switching costs. You could also give up some luxuries in order to be able to afford to outsource some sorts of drag work.
- Time-Poor/Cash-Poor: Our lives are structured so that we tend to progress from time-rich to cash-rich, via a horrible transition phase (usually mid thirties to mid forties) when you are time-poor and cash-poor. This is even worse if you have kids. There is really no good solution in this regime. I am in this situation now, and I chose to give up significant cash for significant time.
- Time-Rich/Cash-Rich: If you are both time-rich and cash-rich, surprisingly, things may not all rosy. From meeting a few people who are “lucky” enough to be in this state, I’ve noticed that this state attracts significant anomie. If you suffer from the anomie of time+cash richness, I’ll think your situation through for you if you pay me.
Getting to 10x and Staying There
Being without a thrust-engine is horrible. The periods of my life when I lacked one are the worst ones I remember. Unfortunately, thrust engines have a lifespan (generally between 7-10 years). You have to get through an initial starter-motor phase, hit an ignition point, and then keep the engine running until it wears out. You’d better have another engine starting up by that point, or you’ll be in trouble.
In high school, my thrust engines were swimming, ping-pong and math. In college, I had swimming, theater and reading. Then I went through a bad phase when I had nothing. Then I got to ignition with Matlab programming. When that engine began to die, my writing practice hit ignition. I am now 4 years into a phase with writing as my main thrust engine. I think the engine will burn out in about 3-4 more years, at which point I’ll need to either find something new, or fundamentally reboot my writing.
Getting to 10x when you have an engine humming takes another level of effort, much of it devoted to drag reduction. I’ll be 37 next month. I’d estimate that I’ve spent maybe 8-12 months overall in 10x mode. I am trying to systematically raise the “10x uptime” in my life, but it isn’t easy.
If you want to dig deeper into some of the ideas in this post, there is related material throughout Tempo. Chapters 2 and 3 should be of particular interest.