how to do stuff

process notes i

the last few months have been very productive! some highlights include: building ddr, reading two long russian novels, writing fiction, fixing up my website, getting more clarity at work. a few observations from recent project lifecycles:

i. strong vs weak links

(this terminology comes from science is a strong-link problem)

early on in most projects i assess the quality of my work based on potential value, and towards the end there’s a shift to assessing quality based on actualized value. this roughly corresponds to the difference between exploring and finishing. one consequence of this shift is that, in the beginning, the quality assessment is dominated by my strongest skills (which tend to be the reason i started the project), while towards the end the quality assessment is dominated by my weakest skills (which are usually the bottleneck to being done)

as a concrete example - i think with ALIGNMENT it was pretty clear that my strengths were prose / narration and my weaknesses were pacing / worldbuilding. my earliest drafts had no semblance of story and were basically just a series of metaphors i was excited to write, so that was what i spent all my time developing and there was nothing else to evaluate. meanwhile, during my last few days of editing, the plot and characterization issues felt so jarring to me that i didn’t have time to worry about anything else. doing a single thing badly can ruin enjoyment (eg. a badly-dubbed tv show) even if everything else is done well

or for a more technical example - in my current work i’m much better at ml engineering than frontend development, so when i’m building features this manifests as having a lot of promising-looking data early on and then spending a long time struggling to integrate that data into good interfaces (in this case the right thing to do may be to hire more people and specialize)

ii. decisionmaking

i roughly bucket skills as either a) domain-specific knowledge (execution) or b) higher-level, like planning and decisionmaking (deciding what to execute on). a) is usually a prerequisite to b) and many people get stuck on a) so they never have a chance to really understand b)

it’s become very clear in the past year that i am a competent executor but rarely pick the correct thing to execute on. once i noticed this i also started noticing that almost everyone spends the majority of their time executing on obviously incorrect things. common examples include: prioritizing smaller tasks to feel productive, distracting oneself with low-level execution to avoid feeling big-picture confusion, copying design patterns without thinking through how applicable they are

(to elaborate on the last point: one pitfall i ran into was designing websites based on how i thought websites “should look” instead of thinking about the exact experience and reaction i wanted a user to have; another was structuring meetings / 1:1s based on meeting structures i was accustomed to instead of working backwards from the state i wanted each person to be in at the end of the meeting)

one thing i appreciate about some of my current collaborators is that they’re quite good at constructively questioning why we are doing something and using that questioning to clarify our plans. that was ultimately one of the main reasons i decided not to go to grad school - with few exceptions the people i met there seemed like they were gaining lots of technical experience and execution ability, but not that much feedback on how to really decide what to do. unfortunate because the latter was the only thing i really cared about anyway

sometimes people ask me how committed i am to working on ai in 5+ years and if i’m worried about skills not transferring. to which my answers are: not committed at all (i’m happy to stay if it continues feeling interesting and important, also happy to jump ship if the field hits a definite wall), not really worried either since i am usually good at acquiring execution skills and expect higher-level skills to transfer. from a growth perspective i don’t think it really matters what you work on as long as it’s aligned with your values and you’re really pushing your boundaries (though of course growth should not be the only thing you care about)

iii. focus

life has been much better ever since i wrote what is the point of blogging? and then proceeded to quit blogging. the (self-imposed) expectation to post regularly was always lurking in the back of my mind, and whenever i had free time or new thoughts i would wonder if i should be writing something. i didn’t realize the extent to which this kind of background process was harming my ability to focus until i cleared it, and i think it’s safe to say most of the things i finished recently would not have gotten done if i’d continued trying to write in the same way

i also developed a few other strategies for focusing. two of my favorites:

• only keep 1 browser window with a small number of tabs (say, 5-10) open, one of which was required to be whatever project i was working on at the time. the project should always be clearly visible every time i open my computer

• i only ever had time to work on ddr during weekday evenings when i was already pretty tired. instead of committing to working on ddr every day, i committed to going to the garage and touching all my wooden tiles every day. some days i would be too tired afterwards to do real work, but most days once i started touching the tiles i would decide to fix something and then an hour would go by