Knowing What's Important
Photo: (ahem) ‘Boys in Bristol Photography’ via Pexels (source).
An emergent skill of mastering any craft—using software, cooking, even playing a game—is knowing what’s important.
This is often true in modern technology. In a complicated interface, the 80/20 rule applies: 20% of the controls are used 80% of the time.01 Learning which parts of the interface are actually used for the majority of tasks is a big part of learning the work.
But there’s also something deeper about what’s important beyond just an interface: discovering which qualities matter most in the execution of a craft. Which aspects or processes do you have to get right, and which are just icing on the cake? I bet this is true everywhere: woodworking, music, plumbing, machining, heck maybe even raising kids.02
I think this is best shown with examples so I’ve collected a bunch of them. Presented below, interlaced with expanded related topics.
Blender
Blender is 3D modelling software. In his famed donut tutorial series, Andrew Price (Blender Guru) first offers the advice to not freak out at the interface, because you’ll learn the important 20% to get what you need done:
First tip: do not get overwhelmed. If you go clicking around, you might find lots of buttons and some of them with very scary names, and it can all get very overwhelming very quickly.
The thing is, you know, Blender can do a lot of things, which means it has a lot of tools. But you don’t need to know every tool in order to do what you want to do. If you’re familiar with the 80/20 rule: there’s really only 20% of Blender’s features that you end up using 80% of the time.
— Andrew Price (source)
Logic
Music production software like Logic have a million menus and plugins and settings and parameters. Even a single aspect of the program, like a single synthesizer, may hold hundreds of hours of potential for mastery.
I did a terrible job at looking this look either comprehensible or complicated, but you get the general idea.
But the hardest part is somewhere at the intersection of the tool, the interface, the task, and the craft. Because often you’ll be confronted with a very high level problem—usually, my song sounds bad. And you need to figure out how to fix it. Experienced practitioners can immediately identify a whole slew of issues, and they know high-impact tricks to apply to improve the sound. So “what’s important” is not just the core features of the program that you’ll use, but but when they’re appropriate to apply.03
APIs
Roughly speaking, a software API just means a list of different functions you can call.
API documentation tells you what exists, but usually doesn’t tell you what’s actually used most by developers. In general, you just get a big, uniform, flat list of options.
One idea I would love is for API docs to show a histogram of usage in the wild.04
Made up numbers for Python’s regular expressions module. Most API docs have their methods in the left-hand side of the page. It’d be fun to see little usage frequencies annotating each method.
Pandas
Pandas is Python’s de-facto library for tabular data.
As with nearly all API references, it’s presented as a huge list of methods on the left.
This is typical, I’m only using Pandas as an example because I find its ergonomics so hard to use.
How do you figure out which of these methods are the most important to use for core operations?
I found Pandas’s getting started guides almost too comprehensive. They’re not opinionated enough. What’s worse is they don’t handle tricky cases. As a result, I never learned most of the important ways of properly working with data. I found one slow and easy way to work, and just always did that.05
Opinionated guides are probably the best because you get to see someone’s distribution of actual API usage. For example, I read someone’s blog posts which, for first time, revealed that .pipe() may actually be a cleaner way of doing many of the things I’d been doing “by hand.” The comparable API page was so laughably barebones as to not even operate on actual DataFrames; it just showed the generic ideas of function composition.06
Python Regex
Python’s re module, for regular expressions, is another example where the API docs are quite comprehensive, but don’t tell you how you’re likely to actually use it.
I think some people would claim this isn’t the job of API documentation. For example, there’s a theory that “there are four kinds of documentation,” so it’s really not the job of API docs to suggest the most useful few methods.
From Daniele Procida’s 2017 talk, What nobody tells you about documentation (source).
I’ve found this idea to be intellectually appealing but untrue in practice. When I’m working, I’m likely to be in a mixture of the above modes: I might need to know some detail of a function’s arguments, wonder whether there’s a better alternative, and refresh myself on some aspect of the best practices or the underlying protocol. Rather than relying on four sets of comprehensive docs—possibly written by different people, possibly of varying levels of completeness, and possibly without links to each other—sprinkling in bits of all quadrants within the docs takes little space in practice and gives you all the information you need in one place.
Some of my favorite docs are those of FastAPI, which mix all of the above in its documentation.
Unity
Unity (a game engine) looks a lot like Blender on the surface, and I think the 80/20 rule absolutely applies to its complex interface as well. But with Unity, I want to try to articulate something different about primary choices not being on the surface.
There are a small handful of things I want to do at the beginning of a project, related to configuring the overall environment, the look, or basic properties of objects.07 And these choices are so deep within menus, that you would never discover them, and as you’re choosing them, you can’t help but think, this has to be wrong, something so primary couldn’t be buried.
The confusion here is that the interface of all these items in a big sub-menu is a completely flat, uniform list. Everything has equal importance visually. However, some items are much more important for getting started than others.
I think this anti-pattern of design comes from a large program being developed over time. The developers add a core feature but there isn’t the design support to change the interface, so it gets slotted into some menu. If you know how important the feature is, you’ll learn how to find it, but if you don’t, there’s no way you will without being taught.
In other words, this issue is about primary configuration being discoverable.
Discoverability
Once you already know the core, most important controls, how do you figure out how to use the next layer deeper? Intermediate, lesser-used controls? How do you find out whether these features exist?
This is closely related to knowing what’s important, but not exactly the same.
Here’s a quote about Microsoft Office’s issue with this:
In 2006 Microsoft conducted a customer survey to find what new features users want in new versions of Microsoft Office. To their surprise, more than 90% of what users asked for already existed, they just didn’t know about it. To address the “discoverability” issue, they came up with the “Ribbon UI” that we know from Microsoft Office products today.
— Haki Benita in Lesser Known PostgreSQL Features (source)
I haven’t used Office in years (maybe a decade?), but I would guess that the Ribbon surfaces context-sensitive UI options. I’d also guess that it labels these icons with text so users are more likely to try them.
Games also surface contextually relevant controls. I distinctly remember seeing this in Nintendo games starting around 2017, with Super Mario Odyssey (2017)08 and Zelda: Breath of the Wild (2017), but it could have started earlier.
Super Mario Odyssey explicitly shows controls all the time, including the very basics of movement at the start of the game. This tactic was likely at least partially chosen because the control scheme changes frequently with the game mechanic of possessing other creatures. (source)
In Breath of the Wild, Link’s various activities also trigger context-sensitive controls to appear at the bottom of the screen. This is in addition to the permanent control reminder at the top left of the screen (not pictured). Like with Odyssey, BOTW’s control scheme changes frequently, and is remarkably complex by default. (source)
Teaching
How do you teach what’s important? As with much in software, games offer a masterclass here.
Good (modern) games don’t even show you all the controls first. They introduce interface elements over time. They only give you access to a subset of the menus and abilities that you use at the start. Often you even start with no controls, and they’ll introduce movement, followed by one button at a time.09
As with much in games, this is a convenience that older games lacked, and which we now expect. Starting an even slightly older complicated game—like The Witcher III (2015)—often feels overwhelming, as you are immediately shown a gazillion things to learn all at once.
Several different menu screens from the start of The Witcher III. It’s a bit hard to convey the amount of stuff there is to familiarize yourself with unless you’re playing. (source)
I think a critical aspect is not even seeing the stuff you haven’t learned yet.
Physical Controls
What’s important when controlling a device?
For me, washing machines and microwaves classically have way too many buttons. I always want to know what the setting I’m choosing is mapping onto—is this just a particular preset (e.g., power & time, or temp & time & spin), or is there some more complicated feature I’m accessing (e.g., pulsing power on and off, some sensor)?
Here’s one such washing machine’s controls:
Like, what is the pizza vs mechanic grease setting? (Also, I was convinced that was a surfboard sticking out of the grass until now, like the eighth time I’ve looked at it. It’s just grass.)
I have encountered one microwave ever with nearly perfect controls: two knobs, one for power and one for time.
The only problems are (a) the power should show effective Watts because everything with microwave instructions specifies it, and (b) the time selection has got to be nonlinear. I care about 15s vs 30s all the time, and I have never microwaved anything for 35 minutes and would be scared to try.
Controls Changing
Related to important controls, is how frustrating it is for users when they change.
My dad complains to me about Lightroom (photo editing software) constantly changing their controls, whereafter he has to re-learn where a feature is, how to use it, or both.
Here’s a clip of 3D digital fashion artist and Twitch streamer Stephy Fung getting frustrated by a changed feature in a 3D modeling software (Cinema 4D?).
Self-Teaching
I think as a kid, you’re just content to “waste” a lot more time fooling around with new programs and interfaces, which can organically lead you to mastery. I used to think that as an adult, my brain is now dramatically less plastic. Now, I think the bigger factor is just that I’m less patient to play in the unknown.
YouTube videos are of varying helpfulness for this. For example, I’ve found videos for Unity (game development) are generally great, and videos for Logic (music production) are generally bad.10
Shadowing
I think figuring out what’s important is a major reason why shadowing—i.e., following an experienced person around on their job—should be way more common in more industries.
I have learned so much the few times I’ve watched a coworker or fellow grad student work for even a couple minutes. So much is interesting, but relevant here: how they distribute their time, and which details they fuss over.
Cooking
The point-of-view YouTube cooking streams of J. Kenji López-Alt are my absolute favorite video material for learning how to cook. It’s like shadowing a chef at home. He talks through alternatives, substitutions, omissions, and timings—in other words, he explicitly tells you what is important for preparing a kind of food. This kind of information is so helpful beyond a recipe, where if you don’t know better, you assume everything, including that 3/4 tsp of whatever spice, is of equal importance.11
Late night Carbonara and Whiskey Sour is my favorite of this video series, though others have significantly more talking and educational value.
Other great food writers (Claire Saffitz’s Dessert Person (2020) comes to mind) include what’s important-style material in their recipes, too.
Footnotes
Actually, I bet it’s more like 97% / 3%. ↩︎
This reminds me of quality, cf. Zen and the Art of Motorcycle Maintenance (Pirsig, 1974). You have to figure out what quality looks like, and then spend a bunch of time figuring out how to achieve it. At that point, you can produce work with quality, but there’s probably still a lot of superfluous stuff too. But then if you work at something long enough, I think you start to understand the true essence of quality. Distilled potent essence of the craft. ↩︎
Let me wax slightly more on this one because I still don’t feel like I’ve captured it. In Logic you have thousands of controls to press, and millions of workflows. The extent of “I don’t know what to do here” is hard to fully capture. It feels like you’ve entered a room with piles of sharp tools and beeping equipment, no medical training, a patient with an unknown problem lies on the table, and you’ve got to do surgery. ↩︎
This “understanding the common case” is one thing LLMs are extraordinarily helpful with, because it’s what they have seen and want to output the most. ↩︎
In case you know pandas: I’d just iterate over the rows of a
DataFrameand write vanilla Python on vanilla objects. ↩︎The current API does show an example with
DataFrames, but I still find it somehow too vague and too specific to quickly understand how to use it. ↩︎My writing is unfortunately vague here because my memory of the details is fuzzy ↩︎
Notice the significant departure from the much lauded start of Super Mario Bros (1985), where Mario is simply placed slightly to the left of the screen, implying that you should run right. I think this is good and necessary. ↩︎
Games also benefit from conventions (e.g., confirm is X, crouch is L3, shoot is R1), which helps experienced players rapidly learn new games. ↩︎
Most beginner YouTube videos for Logic feel like scams: check out this very specific thing I clickbaited with a vague and grandiose video title. Oh, want to know more? Sign up for my paid class. ↩︎
Many gleanable nuggets live in Kenji’s videos, even, e.g., where he keeps stuff in the kitchen. ↩︎
