The blogosphere abounds with free advice, and one of the more frequently-repeated morsels is to “be prolific”, and quality will naturally follow.

The following story, from Art & Fear by David Bayles and Ted Orland, is brought in as supporting evidence:

The ceramics teacher announced on opening day that he was dividing the class into two groups. All those on the left side of the studio, he said, would be graded solely on the quantity of work they produced, all those on the right solely on its quality.

His procedure was simple: on the final day of class he would bring in his bathroom scales and weigh the work of the “quantity” group: fifty pound of pots rated an “A”, forty pounds a “B”, and so on. Those being graded on “quality”, however, needed to produce only one pot – albeit a perfect one – to get an “A”.

Well, came grading time and a curious fact emerged: the works of highest quality were all produced by the group being graded for quantity. It seems that while the “quantity” group was busily churning out piles of work – and learning from their mistakes – the “quality” group had sat theorizing about perfection, and in the end had little more to show for their efforts than grandiose theories and a pile of dead clay.

There are ~1,690 Google results for the exact phrase, “The ceramics teacher announced on opening day”. The story has been picked up by a number of large blogs, including Stratechery, Coding Horror, CSS Tricks, and HuffPost.

I have read many of these posts over the years. They’re inspiring, and seem like great advice — and the scientific-sounding story of the ceramics teacher’s A/B test neatly encapsulates the core idea.

However, the story has been described by some as a “parable”, by others as an “anecdote”. Did it really happen?

Austin Kleon writes that James Clear, who used a different version of the story in his book Atomic Habits, corresponded with Ted Orland (the co-author of Art & Fear) in 2016. It turns out it is, indeed, a parable — a retelling. The “ceramics teacher” was photographer Jerry Uelsmann of the University of Florida.

Fail more

“Be prolific” is a fast car with no steering wheel

That which is asserted without evidence, we are told, can be dismissed without evidence. However, whether myth or fact, the story of the ceramics teacher persists because it encapsulates something that feels true.

However, what’s encapsulated in the ceramics teacher parable, I think, is a small part of the process that’s conjured up in the mind of experts who read it. The unspoken step in the “quantity > quality” formula? Reflection, feedback, and revision. Beginners who hear the story don’t necessarily know this, which can make “create lots of pots” advice unhelpful.

Fail better

Reflection, feedback, and revision

I don’t get better by repeating my mistakes. I have to somehow figure out what went wrong, learn the lesson, and try again, making different smaller mistakes (on and on until the frog tastes treebark).

I don’t need to publish, or sell, my early drafts. I could shape and re-shape one ceramic pot over and over again, or make a hundred and put them in my basement, or make a hundred and try to sell them. I just need to get enough experience to know what works and what doesn’t.

Feedback is fundamental to learning. We hear about “performance, feedback, revision”, learning from failure, and iterating our way to success. But what counts as “performance”? What qualifies as an “iteration”?

Fail faster

What counts as an iteration?

My answer: the unit of effort that results in actionable feedback.

Many people (as in the ceramics teacher parable) conflate iterations with complete, public iterations.

This, as far as I can tell, needn’t be so: here, my opinion diverges from the Doctrine of Prolific Output. I subscribe instead to the Doctrine of Prolific Effort.

There are merits to the Doctrine of Prolific Output:

1. There’s less room for fudging the action-reaction-revision cycle.

2. Actually publishing increases “luck surface area”.

3. Publishing makes me not just a better writer, but a better publisher. Publishing and promoting might need as much practice as the writing.

But.

If I want to become a better writer, I don’t have to publish prolifically. I just need to put in the performance-feedback-revision cycles.

Doing that in private can be more efficient than doing it in public. And rapid feedback cycles are key to learning and growth.

Fail privately

Who are we optimizing for?

The thought-leaders might be right that “putting your work out there”, even if it’s sloppy early work, is helpful for the artist. Maybe somebody will like it, even if I don’t think it’s that good. I think this strategy is more likely to result in professional and financial success.

But the people giving advice are focussed on what’s good for the artist, not what’s good for society as a whole. The world’s well supplied with half-baked blogposts and lumpy ceramic cups, and it’s getting hard to find the masterpieces buried in the rubble.

(It also might give an aura of mysterious authority to appear on the scene already competent, rather than flailing publicly.)

Though I haven’t published much fiction, for a while I made myself outline three stories per day, and write the first one that I felt excited about. I wrote thirty or forty stories, ranging from 1 to 160 pages, and more outlines than I can count. It was after doing this — never self-publishing my fiction — that magazines started buying my stories. It’s all about the cycles.

The first story I sold was the thirty-seventh draft of a story I wrote when I was a teenager. ^[1]

Fail outward

How an idea grows up

Let’s use writing as an example.^[2] The usual advice is to write a lot, put pieces out in the world, and learn from how they’re received.

(This is what I have been doing for ten years, and I have learned a lot from it.)

But there are meaningful feedback loops with lower latency. The simplest? Become ChatGPT:

When writing each word, I try to predict what’s the best word to go there. When I think I’ve got it, I type it. That’s the performance. Then I read it in context, and notice how I feel about that word. This is the feedback. Then, I try to come up with a better one. And our loop is complete. ^[3]

This can be extended to the sentence level. Then, to paragraphs and sections. This is how I write: by the feeling of rightness or wrongness, developed over time.^[4]

Sometimes, reflection is enough, but usually I need an outside perspective. ^[5] Once I think it’s great, the real work begins. Does anyone else like it?

I can start with my best friends, and work my way outward, perfecting as I go. (Sometimes, people will say something is so bad, it’s more efficient to try again from the start than to try to fix what’s broken.)

Why is this a good idea?

Partly, because it’s fast. Magazine and book turnaround times are outrageous. Just getting a form rejection in the literary world can take more than a year. Writing, editing, and polishing a whole piece, and then self-publishing it, is also too slow.

Mostly, it’s because I get more data back. The people around me will give more detailed, and more blunt, feedback. Editors are too busy (also, I want to impress them, so I don’t send them lousy half-baked drafts; but reasoning errors are easier to fix in the rough-cut version than in the polished version.)

Also, it’s because my friends are my target audience. Who I get feedback from matters. If I publish every half-baked draft, I’ll get feedback from a general audience, and it will tend to create an incentive to pander to a general audience, and drive my output toward the mean. Instead, it works better to get feedback from fewer people, who are the people who I would like to be like. The internet may have all kinds of mean (or adulatory!) things to say; but the internet might not even understand what I’m trying to do, let alone whether I succeeded. It’s like scoring a baseball game based on the number of touchdowns.

I show my draft essays to as many friends as I can, as many times as I can — while still having friends to show the next essay to. I write far, far more than I publish.

Once I’ve incorporated feedback from friends, I can start showing pieces of it to pieces of the internet. I buckle in, and read the comments with a box of tissues nearby. I do this with blog posts. Henrik Karlsson beta-tests his ideas piecemeal on Twitter/X, and a secret Substack. In Antimemetics, Nadia Asparouhova talks about people who use group chats for prototyping their ideas. And many have talked about how famous comedians workshop their (not yet famously funny) material at small clubs before turning it into a Netflix special, and touring theatre companies start out in small towns so they’ve smoothed out the rough parts by the time they reach the larger and more difficult-to-impress Broadway audiences.

The moral of the story: don’t send your idea into a war zone when it’s in diapers.

Sometimes a picture is worth thirty-six words:

Bands often have a parallel workflow. First, the songwriter has to like the song. Then they play it for the rest of the band. If the band likes it, they test it out in live shows. If it works live, they might record it. And if the band, producer, and record label executives like the recording, it might make it on the album.

Fail deliberately

On drilling through walls

When people talk about being prolific and increasing “luck surface area”, they often talk about increasing the number of “shots on goal”. The idea is that you can’t succeed without attempting to succeed, and more attempts also gives more opportunity for opportunity, and more feedback on how to do better.

Each shot gives more data. You can’t generate a pattern if there isn’t enough data, so it’s easier to gain insight by taking more shots, than by squeezing your brain. (If you try, it will likely be wrong.)

However (to return to my old saw), there’s a difference between taking more shots during the game, which is what most pundits advocate for, and doing drills to improve your aim.

Drills are more efficient (and the more atomic the drills, the better). It means you can target your weakest point, without wasting time practicing things you are good at, or messing up the game for the other players.

An ideal drill should target a specific skill, have specific criteria for success and failure, and provide actionable feedback. With sports, it’s easy: you try to hit the ball in a certain place, or bounce it on your foot a certain number of times, or whatever, and it’s pretty obvious whether that goal is being reached. With the arts, it’s less clear. Often the drills are about doing the thing. Publish every week. Practice every day. These are good, but they don’t guarantee you get better. These are shots on goal — but the goal is undefined!

External criteria are harder to fudge, but also more susceptible to Goodhart’s law.^[6] This is where that finicky thing called taste comes in; which is, I suppose, what makes the arts arts rather than sciences. If you’ve read this far, I suspect your taste is already ahead of your abilities, and, without trying to pin down what precisely makes something good or bad, like Justice Potter Stewart, we can admit that though we can’t define it, we “know it when we see it.” (This essay isn’t about choosing the destination: it’s about getting there.)

Fail piecemeal

There are many ways to debug a song

There are a couple ways to iterate faster:

1. get the feedback from sources closer by. This often reduces communication latency, increases bandwidth, and results in more targeted feedback.

2. slice the thing into smaller pieces, and get feedback on specific parts.

We’ve talked about #1, but not as much about the slicing.

There are a few ways to do this, depending on the medium. One way is to physically slice the thing up. Get feedback on the first chapter of the book, or the chorus of the song, or the hors d’oeuvre of the dinner. Once that’s dialled in, move on to the rest.

Another approach is to get feedback on a specific “thread”. The rhythm, then the pitch, then the narrative, then the rhyme. The texture, then the taste, then the presentation.

Somehow, you need to break the thing up without breaking it. Sometimes parts stand on their own. Other times, you can’t work on one component in isolation. The chorus might be great! But it also has to lead into the next verse elegantly, and optimizing it on its own might make the song worse. But the same can apply to a “threaded” approach. Improving the rhymes might make the story, or the melody, or the rhythm fall apart in subtle or glaring ways.

I wrote a lot about splitting systems into pieces in my essay on troubleshooting, and a lot of it applies to making things, too. And I also rhapsodized about the importance of fixing the foundations before applying stucco in my revision checklist article. And I wrote about not sacrificing the whole for the parts in holistic perfectionism. (I also translated my best writing advice into caveman-speak, if that’s your thing.)

Fail like Dylan?

Examples and counterexamples

Many of the most popular songwriters of all time are absurdly prolific. Immediately, Bob Dylan, Ed Sheeran, Dolly Parton, and Taylor Swift come to mind. And Bob Dylan, at least, is famous for not revising, and for writing quickly.

Are they counter-examples for my argument? Or, do some people just have so much talent and energy and luck that no matter which approach they take, they succeed?

Then there’s Leonard Cohen, also a highly respected songwriter, who spent years refining his most popular song. And J.D Salinger: famously unproductive, yet famously famous. And Henrik Karlsson, who takes and advocates for a similarly thorough approach. He’s prolific, but his public output is the tip of the iceberg, and he works harder at refining his essays than most people I know. I’ve read his drafts, and his finished pieces, and they’re often like the Ship Of Theseus: almost nothing remains from the first draft, other than the soul of the thing.

There are endless examples from both sides: the hermetic perfectionists, and the prolific experimenters. The examples in the first category seem fewer, but they’re there.

I know many bloggers and songwriters who have oodles of talent and take an excessive number of shots on goal and just keep on missing. It’s heartbreaking to watch someone smart enacting a definition of stupidity: doing the same thing, over and over again, and expecting different results.^[7]

What do Dylan et al do differently than the people who seem stuck in abundant mediocrity? And what do Cohen et al do differently than the Edward Casaubons who obsess and obsess, but never produce a masterpiece, and eventually shrivel up and die with little to show for their efforts?

I suspect that in both cases it has to do with paying attention — reflecting, and being perceptive — and not getting so attached that they can’t bring themselves to make changes, even changes for the better.

It’s clear both approaches can work. Maybe, some people are better suited to refining existing material, and others to trying again from the beginning. Or maybe it’s a question of what the goal is, and the perfectionists are trying to get that particular piece perfect, while the experimenters are trying to make a perfect piece eventually.

People often keep iterating until the glue sets. Writers iterate until they publish. Songwriters iterate until they record (or they have to keep it the same so they don’t confuse the band). Web developers get to keep iterating, but people writing software for embedded devices have fewer shots at getting it right.

In life, it seems people often take an informal approach to the multi armed bandit problem: they iterate until they’re sort of satisfied, and then they settle down (sometimes in a local optima).

Marriage is an example. People flirt and date (fast iteration), and then eventually commit, at which point the finding someone iteration process is complete, unless they get so unhappy with the result that they start over, and the improving the relationship with the person they committed to iteration process begins.

I think the frame of what the “goal” is can make a bit more sense of the types of iterations.

1. Iterations of discovery: experimenter knows they want to make something good, but whether any given attempt is good doesn’t matter as much, they’ll keep trying new things until one of them is great. The goal is to make something perfect.
2. Iterations of refinement: perfectionist knows what they want, and are shaping what they have made until it is what they want to have made. The goal is to make this thing perfect.

How much time to spend on 1 vs 2 depends partly on temperament, and partly on how good the idea you have is. I tend to spend most of my time working on a few big ideas that I’m excited about and want to do justice to, while also dashing off smaller, more experimental projects that are low cost and might have high upside.

Fail to fail

The failure modes of iteration

If effective iteration were easy, people who keep grinding would eventually succeed. It’s pretty clear they don’t.

There are many ways to feel like I’m iterating toward victory, without making forward progress; or to make progress, yet still be sailing slower than the tide.

I will categorize these traps as “broken loops” and “wasteful loops”.

If the loop’s broken, more iterations won’t help! Broken loops — which are also wasteful, but whatever — mean the feedback/reflection is wasted in one of two places:

1. feedback bounces off or passes through the ears. Often, I ask for feedback (or I reflect), but when I receive it, I get defensive and decide it doesn’t apply (feedback bounces off), or it just passes through and doesn’t register as important.

2. “but mum I’m tired”. I hear and understand the feedback, or the voice in my head. I know I should fix the problem: it nags me. But I’m too lazy, and put it off, or ignore it. Either I shelve the project forever; or I ship it in a hurry with the excuse that I didn’t have time to fix it; or I fix it half-heartedly, and am always a bit ashamed that I just slathered fresh paint over the rotten part and hoped nobody would notice.

Technically, failing to reflect or ask for feedback is also a “broken loop”. (Not making anything in the first place is a “no loop” problem.) But we’re grown-ups, and if you’re reading this you’re probably already making stuff, and trying to make it better.

3. shunting the brain. The opposite of being stubborn and defensive about feedback is treating other people’s opinions as gospel, and acting on their feedback without thinking it through. It’s mum I’m tired in yes mum disguise. It acts like a goody two-shoes opposite, but it’s just as lazy. I find myself mindlessly implementing suggestions surprisingly often, and it’s dangerous. I can’t count the number of times I tried to ‘correct’ subtle wordplay that I hadn’t noticed. Oof! And people frequently point out “typos” in my drafts that are not accidental. ^[8]

4. winning the wrong game. Iterating faster is pointless if the goal is no longer meaningful by the time it’s reached, either because it was a dumb goal to start with, or because the design of the iteration process meant it was actually leading towards a subtly different destination, like a compass that points at magnetic north rather than true north: a navigation error that becomes more severe and obvious as the goal gets closer.

5. calling the structural engineer once it’s built. The time to get feedback on fundamentals is while the fundamentals are malleable. Once other things are built on top of them, changing one tiny thing often means upheaval. The time to iterate on a foundation is before it’s poured.

6. l..a..t..e..n..c..y.... Unnecessary distance and delay anywhere in the performance-feedback-revision loop makes iteration inefficient and onerous. Anyone who has been burned by the delay between turning the knob in the shower and when the water temperature changes knows just how frustrating (and dangerous) laggy feedback loops can be.

7. painting lessons from Van Gogh. It feels good to only get input from the best; and sometimes it’s wise, because they notice things others don’t. But often it’s unnecessary, or even counterproductive:

   • they’re usually busy, so it’s slower
   • they’re usually busy, so they have less time once they get around to it
   • it’s disrespectful to demand professional level help for amateur problems
   • their feedback can seem baffling or excessively critical, and lead to daunting gumption traps.

8. missing the tide. Going too slow, or dallying, can have cascading effects down the line that aren’t apparent at the time. Projects have natural cycles, both because of internal and external factors. Missing the window when it’s easy can lead to putting in more effort for less (or zero!) results. ^[9]

Keep that loop running fast and smooth. If you fail to fail, what choice is there but to succeed?

As Zig Ziglar said, what counts is “not how far you fall, but how high you bounce”.

Thanks to Davis for reading a draft of this essay and giving feedback. All errors are due to failing to take my own advice or my advice being wrong. Let’s go with the first one.

What’s your approach to effective iteration? Hit reply or let me know in the comments!