Hammertime Postmortem

Link post


A bit less than two months ago, I set out to write about in­stru­men­tal ra­tio­nal­ity ev­ery day for thirty days. In this post, I will quickly eval­u­ate how well I felt I did along each of my four stated ob­jec­tives. I will si­mul­ta­neously eval­u­ate all the Ham­mer­time tech­niques and ideas by their effec­tive­ness to my life.

This pe­riod was my dead­line to 8020 in­stru­men­tal ra­tio­nal­ity. Thus, I do not plan to blog any more about it for a while. How­ever, I do want to ex­press my strong in­tent to write a fourth cy­cle of Ham­mer­time in the early months of 2019, if only to check my long-term progress.

1. Ham­mer­time Re­port Card

I will grade my­self on the four goals I stated in the first In­ter­mis­sion thread:

My rea­sons for writ­ing this se­quence were, in clear or­der of im­por­tance: (a) to prac­tice writ­ing, (b) to re­view CFAR tech­niques for my own benefit, (c) to en­ter­tain, and (d) to teach in­stru­men­tal ra­tio­nal­ity.

On re­flec­tion, these were equally im­por­tant goals and I only listed them in that rel­a­tive or­der be­cause I be­lieved the later ones would be harder to achieve. I will grade ev­ery­thing out of 100, count­ing up from zero. Only the rel­a­tive sizes of the num­bers mean any­thing.

Writ­ing Prac­tice: 90100

This worked out quite well. I pro­duce con­tent about three times faster than I did at the be­gin­ning of Ham­mer­time, with per­haps the slight­est de­crease in qual­ity. Speed I value as much as strength, so this was an amaz­ing im­prove­ment. There are things like or­ga­ni­za­tion and style I should have played around with more, and a Yoda Timer of copy-edit­ing af­ter each post would have benefited the writ­ing qual­ity greatly.

Per­sonal CFAR Re­view: 95100

Through this pro­cess I was forced to re­flect on, try out, and push the bound­aries of al­most ev­ery sin­gle tech­nique in the man­ual. Other than a hand­ful of tech­niques that don’t click with me at all, this two-month pe­riod has been the perfect amount of time to throw at ded­i­cated in­stru­men­tal ra­tio­nal­ity prac­tice. The long-term value of the learn­ing I did at CFAR at least tripled be­cause I did this.

En­ter­tain­ment: 65100

Hit or miss. Hand­ful of posts that were re­ally fun to write, and still look fun to read. I no­ticed a num­ber of clear limi­ta­tions in my writ­ing toolkit that don’t seem to be fix­able in a day or two (but might be if I ac­tu­ally tried). De­spite my best efforts, I’m still not Eliezer or Scott.

What am I miss­ing? I plan to ex­per­i­ment more with di­alogue, which I’m awful at writ­ing but seems to make some of Eliezer’s and Scott’s fun­niest stuff. Also, de­tailed and en­ter­tain­ing ex­po­si­tions of sci­ence are sorely miss­ing in my writ­ing – this seems like a gold mine as well.

Teach In­stru­men­tal Ra­tion­al­ity: 50100

Not sure this se­quence is any bet­ter as ped­a­gog­i­cal ma­te­rial than just the CFAR Hand­book, which is a mod­er­ately dry refer­ence man­ual. Per­haps that’s good enough. A hand­ful of peo­ple seemed to benefit quite a bit, but my sense is that even among the peo­ple who read ev­ery post, few did any of the ex­er­cises or got any mileage out of this se­quence over learn­ing what the con­cept han­dles are. In the end, I always made de­ci­sions in fa­vor of “write what’s in­ter­est­ing for me” rather than “write what I think would be most use­ful to the reader.”

Per­haps an in­ter­ested reader would like to take a cou­ple hours and re­assem­ble the most use­ful parts of Ham­mer­time into a cleaner sub­se­quence. As a re­source on in­stru­men­tal ra­tio­nal­ity in­struc­tion at most half of the posts in Ham­mer­time are of high value.

Over­all: 75100

Very im­pressed with my­self that I fol­lowed through with this pro­ject with only minor de­lays. Every­thing went ap­prox­i­mately as well as could be Out­side-View ex­pected.

My main take­away is to con­tinue throw­ing my­self head­long into medium-term pro­jects with­out think­ing too much about them, and trust my in­stincts. It’s not ob­vi­ous that more plan­ning or struc­ture would have helped in net – it may even have soured the whole Ham­mer­time pro­ject and caused me not to finish at all.

2. Ham­mers by Power Level

I will go through the core tech­niques I cov­ered in Ham­mer­time, and grade them each based on effec­tive­ness in my own life.

I’ll sort them into three tiers of awe­some. Note that the tech­niques in Ham­mer­time were already pre-se­lected from a larger pool of tech­niques based on how good they seemed to me just af­ter CFAR.

S/​A Tier

Fo­cus­ing: 100100

Doesn’t always work, but when it does … life-chang­ing in­sights. Prob­a­bly had three or four over the course of Ham­mer­time. Would recom­mend.

Yoda Timers: 95100

Timers and dead­lines re­ally up my game. I think I’ve always shied away from us­ing them be­cause “con­test math,” “speed,” and “com­pet­i­tive­ness” be­came low-sta­tus af­ter high school, but man am I built for this. Some­times I think that if grad school was struc­tured as a se­ri­ous of olympiads ex­cept with open prob­lems, I would get a lot more work done.

De­sign: 90100

Amaz­ingly un­der­rated tech­nique. Amor­tiz­ing ev­ery­thing, al­low­ing my­self to re­move triv­ial in­con­ve­niences, spend­ing time mak­ing my phys­i­cal space bet­ter. Sub­stan­tially im­proved my baseline qual­ity of life: sleep qual­ity, over­all com­fort, aes­thet­ics. If I gave up ac­tively us­ing in­stru­men­tal ra­tio­nal­ity right now, the effects of the De­sign choices I made in the last two months would still last for years.

B/​C Tier

Bug Hunt: 80100

Very use­ful to prac­tice ev­ery so of­ten. Ups your notic­ing game quite a bit for a long time.

CoZE: 80100

Another solid tech­nique. Gave me the tools to push through many minor un­en­dorsed aver­sions and try things in­stinc­tively. Doesn’t work as well by it­self on the big­ger aver­sions – in my ex­pe­rience, these re­quire the aid of Fo­cus­ing and Fo­cus­ing is the one do­ing the work.

Silence: 80100

I feel as if com­bat­ing the ten­drils of nihilism in ev­ery­day life is one of the biggest prob­lems to solve. Silence was my first at­tempt at fram­ing the prob­lem and offer­ing a par­tial solu­tion. As always, peo­ple need to al­low them­selves to bab­ble more.

TDT for Hu­mans: 75100

Im­por­tant prin­ci­ple that fi­nally al­lowed me to un­der­stand the ap­peal and util­ity of virtue ethics/​de­on­tol­ogy. Re­quires more iter­a­tion and work to make it ac­tion­able.

Friend­ship: 75100

Notic­ing the value of and set­ting up long-term iter­ated con­ver­sa­tions with friends was ex­tremely valuable. Ex­per­i­ment­ing with this also led me into a hand­ful of awk­ward so­cial situ­a­tions and un­pro­duc­tive con­ver­sa­tions. I’ve up­dated to­wards there ex­ist­ing even fewer peo­ple than I thought with whom I can have in­ter­est­ing con­ver­sa­tions on a reg­u­lar ba­sis.

D/​F Tier

Mur­phyjitsu: 65100

It feels as painful and difficult to prac­tice as read­ing abil­ity in Go – life is too chaotic. For now, it’s only use­ful on the five-sec­ond level: what are the ob­vi­ous things that will go wrong? Per­haps af­ter I col­lect more data about com­mon failure modes Mur­phyjitsu will be more use­ful. As of now, I feel woe­fully un­cal­ibrated.

On the plus side, did in­spire my longest work of fic­tion to date.

TAPs: 60100

Weird and un­nat­u­ral to prac­tice. Hand­ful of use­ful things I thought I in­stalled rapidly faded with time. TAPs seem to last about a week for me with­out some other reg­u­lar re­in­force­ment mechanism.

In­ter­nal Dou­ble Crux: 50100

Too many steps. The only real value seems to be as a method for gen­er­at­ing Fo­cus­ing tar­gets. This is pretty valuable, but still.

Aver­sion/​Goal Fac­tor­ing: 30100

Tried a few times, didn’t stick. Much weaker than Fo­cus­ing. Usu­ally, what I need to do is “find out my true main mo­tive and aver­sion to­wards the thing,” and once that is done the path for­ward be­comes clear.