Attention is your scarcest resource

Link post

Like many peo­ple, I have most of my best ideas in the shower.

This is some­times an­noy­ing: I could use more than one shower’s worth of good ideas a day, but I’d rather not end up as a shrivel­led yet in­sight­ful prune. Mostly, though, shower ideas are the in­cen­tive that keeps me smelling okay, so I grudg­ingly ac­cept the con­straint.

The time when it was most con­strain­ing was the first time I be­came a man­ager. I only had a few re­ports, so man­ag­ing them wasn’t a full-time job. But I was very bad at it, and so it should have been what I spent all my shower in­sights on.

Un­for­tu­nately, I was spend­ing my non-man­age­ment time on pro­gram­ming. And even if I tried to use my show­ers to think about my thorny and awk­ward peo­ple is­sues, my mind some­how always wan­dered off to tackle those nice, juicy soft­ware de­sign prob­lems in­stead.

Man­age­ment At­tempt #1 didn’t go very well; be­ing dis­tracted by pro­gram­ming, I made lots of em­bar­rass­ing mis­takes that made my re­ports’ lives harder, which in turn made me feel stressed and guilty. I ended up bounc­ing back to be­ing an in­di­vi­d­ual con­trib­u­tor af­ter a few months.

That ex­pe­rience of mine res­onates strongly with Byrne Ho­bart’s ob­ser­va­tion about fo­cus in knowl­edge work:

The out­put of knowl­edge work­ers is ex­tremely skewed based on fo­cus. The pro­duc­tivity tiers seem to be:

  1. <10% fo­cused on the job at hand: mean­ingful risk of get­ting fired.

  2. 10-50% fo­cus: “meets ex­pec­ta­tions,” gets reg­u­lar raises.

  3. 50%+ fo­cus: su­per­star, 10x en­g­ineer, des­tined for great­ness.

“50%+ fo­cus” is roughly when some­thing be­comes the top idea in your mind. It’s when you start car­ing enough to think about it in the shower. It’s when you start ha­bit­u­ally ask­ing “how could this go faster?” It’s when you get re­lentlessly re­source­ful. It’s around when you start an­noy­ing your cowork­ers and/​or sig­nifi­cant other, al­though that part is avoid­able with prac­tice.

Most im­por­tantly, you can only be 50%+-fo­cused on one thing at a time—or zero, in bad cases. That makes it crit­i­cal to con­serve your at­ten­tion, so that you can spend it on what mat­ters.

I’ve found a few differ­ent strate­gies that help me with this:

Care viscerally

It’s hard to re­cruit 50%+ of your brain to work on some­thing just be­cause you feel like you “should” do it. Your su­perego might listen to this new­fan­gled “logic” and “rea­son” stuff, but the mon­key-era parts of your brain can’t be fooled so eas­ily. To be re­li­ably able to fo­cus on some­thing, you need to be in­tu­itively, emo­tion­ally in­vested in the out­come.

For me, that’s not au­to­matic—for in­stance, even though I’m in­cred­ibly ex­cited about what Wave is do­ing, my day-to-day mo­ti­va­tion gets worse if I work re­motely for too long with­out vis­it­ing our op­er­a­tions in Sene­gal. Un­less I can pe­ri­od­i­cally see for my­self that peo­ple are us­ing the things we’re build­ing, it starts feel­ing a lit­tle bit fake, like the so-called “real world” is a simu­la­tion that I can only com­mu­ni­cate with via Slack chan­nels and git push.


As a pro­gram­mer, I tried to make sure that I was only ever work­ing on one thing at a time. Even if I got stuck on that one thing—say I was blocked on wait­ing for a tech part­ner to give me API doc­u­men­ta­tion—I’d let my­self stay stuck in­stead of slid­ing off to work on some­thing else.

In the short term, this made me less effi­cient, be­cause I’d spend less time pro­gram­ming and more time star­ing va­cantly at the ceiling. But if I stared va­cantly for long enough, I’d even­tu­ally get mad enough to, e.g., re­verse-en­g­ineer the part­ner’s API in a fit of rage. This re­sulted in me ship­ping my most im­por­tant pro­jects faster, hence get­ting faster com­pound­ing growth.

Evade obligations

As a man­ager, it be­came im­pos­si­ble to “only work on one thing:” there were too many small tasks and too many pro­jects go­ing on in par­allel. But I have a similar prin­ci­ple: I can only have one obli­ga­tion to other peo­ple. “Give ad­vice on this pro­ject?” When­ever you want! “Re­view my 1,000-line change?” Only if re­view isn’t block­ing you from ship­ping it. “Check in ev­ery so of­ten to make sure this doesn’t go off the rails?” Sorry, I have to, uh, pow­der my nose. “Figure out what the pro­cess should be?” I’ll get back to you in a cou­ple years. (The best pro­cess is no pro­cess, right?)

Dur­ing Man­age­ment At­tempt #1, my pro­gram­ming pro­ject dis­tracted me so much be­cause other peo­ple cared a lot that I finished it—if I ever had to de-pri­ori­tize it, I would make those peo­ple sad. This made it take up my at­ten­tion (by wor­ry­ing about let­ting peo­ple down) even when I wasn’t work­ing on it. Un­for­tu­nately, since man­age­ment work comes in ur­gent bursts, I’d of­ten end up in a situ­a­tion where I had to de­lay ei­ther the pro­gram­ming or the man­age­ment—i.e., choose be­tween mak­ing two differ­ent sets of peo­ple sad—and in those cases it was very hard for me to con­trol my fo­cus.

Now that I’ve no­ticed this, I al­low my­self to con­tribute to mul­ti­ple things in par­allel, as long as they don’t tax my ob­ject per­ma­nence skills—if I to­tally for­got about this pro­ject to­day, would any­thing bad hap­pen? As long as the an­swer is no (usu­ally be­cause some­one else is the one tak­ing ini­ti­a­tive to move things for­ward), I can be pretty sure that pro­ject won’t try to colonize my brain in the shower.

Time­box bullshit

Even if I have only one ma­jor pro­ject that other peo­ple de­pend on, I some­how always end up with a bunch of minor chores to do—re­spond­ing to emails, deflect­ing sec­ondary obli­ga­tions, buy­ing house­hold ne­ces­si­ties, pow­der­ing my nose, filing ex­pense re­ports, etc.

For these, I em­ploy a “bul­lshit time­box”—a one-hour pe­riod a few times a week when I do all the things life is too short for. Dur­ing the re­main­ing 165 hours per week, I do my re­s­olute best not to think about any of the bul­lshit, no mat­ter how much peo­ple bug me about whether a 1.5” tung­sten cube is re­ally a le­gi­t­i­mate work ex­pense.

In or­der for bul­lshit not to dis­tract me for the rest of the week, I try to min­i­mize my num­ber of “open loops”—pro­jects or pro­cesses that I’ve started but not com­pleted. Much like how I sin­gle-task on pro­gram­ming un­til it’s done, I’ll make phone calls in­stead of email threads, or, God for­bid, texts or in­stant mes­sag­ing. It took a while for me to train my friends not to in­stant mes­sage me, but by now they re­mem­ber my rants about how dis­tract­ing it is, prac­ti­cally word-for-word.

Man­age­ment At­tempt #2 didn’t go great for me ei­ther. But I fi­nally did an okay job at Man­age­ment At­tempt #3. Part of this was be­cause I avoided re­peat­ing the par­tic­u­lar mis­takes of At­tempts #1 and #2. But part of it was that on At­tempt #3, our en­g­ineer­ing team grew quickly enough that I went from “no man­ag­ing” straight to “full-time man­ag­ing” with no tran­si­tional pe­riod in which to be dis­tracted, screw up and be mis­er­able.

Once I was full-time man­ag­ing, I had no shiny dis­trac­tions and was able to spend my show­ers fo­cus­ing on how to be a bet­ter man­ager. And once I was 50%+ fo­cused, well, I haven’t be­come a “su­per­star 10x man­ager” yet, but I quickly stopped be­ing 0.1x.