Art is Struggle (or Why I Recreated a Flourish Chart in D3.JS)

This “art is struggle” post has work-in-progress D3 charts because it’s raw, unfiltered, and showing parts of the process…so it’s best viewed on the web. Don’t complain to me about things looking wonky on mobile. You’ve been warned.

Also – this was finished in January 2026. I didn’t feel ready to publish it until now. Some of the news references are therefore out of date – but if anything, it just highlights how much worse things have gotten.

I’ve been angry for a while.

And to be honest, I didn’t know why.

I thought it was because a bunch of neophytes1I hate using big words. But I dare you to google the definition of “neophyte” and try to find a better word to capture what I’m talking about. suddenly were talking about how AI meant you didn’t need to learn to write, compose a song, or program a script. (Or learn to think).

Or because there’s a bunch of articles2A16z’s content is contrarian for a reason, it takes outlandish stances to force people to think. Still, this article, based on a single premise, that the war for tech talent will totally transform because of the rise of one-man startups, because of AI, really takes the cake. on how, with AI now, you can start a one-man empire.

But then I narrowed it down.

A clue lies in something I heard in a presentation recently (paraphrased and exaggerated because my memory is poor, and verisimilitude):

“I’ve always wanted to create children’s books, but I didn’t have the time. So, I put one of my ideas into Google Gemini, and it created a beautiful children’s book. That’s a way to use these tools, put in some of your ideas that you don’t have time to work on, and marvel at the results.”3Let’s pretend and ignore the fact that the tool just delivered an output trained on a bunch of illustrators’ efforts who’ve received no compensation for it…

Or in CEOs and Founders saying stuff like this to make a buck:

“It’s not really enjoyable to make music now. People enjoy it, sure, but it takes a lot of time. It takes a lot of practice. You need to get really good at an instrument, or really good at a piece of production software.

I think the majority of people don’t enjoy the majority of the time they spend making music.“4Shout out to the amazing John Oliver for surfacing this in his excellent episode on AI slop.

And now I know why I’m furious.

These guys don’t care about creativity. They don’t care about the joy in inventing or creating something. They only care about riding this trend and getting rewards out of it.

Because you see, to create beautiful stuff. You’ve got to struggle. Art is struggle. I thought this was obvious. But then turns out I’m the dumbass for thinking so.

The Data Visualisation Struggle Journey

In 2019, I started out as a Data Journalist for LandTech, the fastest-growing UK PropTech company. This was a new career, and my introduction to data visualisation.

After a few stories (and wins) published with Datawrapper and Flourish, I realised that to really create the visualisations I wanted, I would need to learn D3.JS.

But I didn’t have time (you try to upskill while working in a marketing dept in a scale-up). This was going to be a mountain of a journey. My own personal Everest.

Fast forward some years, and I finally had the time I needed to begin learning.

It’s been nearly 18 months now. Two textbooks, uncountable text files and git commits. Several handwritten pages to clarify confusing concepts5The missing knowledge in my tech stack has been JavaScript (knowledge of HTML, CSS, and programming concepts via Python (and SQL, though that’s a declarative language). Because I don’t want to become a web developer, just someone who can make gorgeous data-viz, I’ve skipped learning JavaScript and gone straight to learning D3.JS. Filling in the blanks where needed. Good is better than perfect.. And I finally felt ready.

No, not ready to create my own project. Are you mad? No, no, no. Not at all.

No, I felt ready to reverse-engineer one of my own existing data visualisations.

Sign-up form not appearing above? Pesky JavaScript blocker… click here and scroll to the bottom for a vanilla boring Mailchimp form that works on all devices.

D3 and Deliberate Practice

Why this gigantic waste of time?

Because of deliberate practice. Because art is struggle.

I don’t want to just know how to use D3. I want to be able to create the viz’s I want intuitively.

Instead of defining deliberate practice, I’ll share a few examples.

  1. Deliberate practice is Air Force pilot Cesar Rodriguez making his instructor force him to do the flying drills he was weakest at, ten times more than the other students. He would earn the nickname The Last American Ace.6In Desert Storm, Cesar Rodriguez and his wingman were lured into a trap by Iraqi MiG fighters. To escape, he performed a series of bold, perfectly timed, and lightning-fast manoeuvres that saved him and his wingman. The encounter lasted barely four minutes, with one of the dogfights lasting seconds. Upon watching the videos during the debrief, he had almost no recollection of any of it or where his ideas had come from. His intuition had come into play; he couldn’t explain how he had performed tighter turning circles on his opponent while keeping the desert floor in mind. Or why he’d chosen to jettison his fuel tanks at the beginning of the dogfight, which gave him greater speed and manoeuverability. That’s the power of deliberate practice. Source: Mastery by Robert Greene
  2. Deliberate practice is Bill Bradley practising basketball dribbling with eyeglass frames with bits of cardboard taped to the bottom so he couldn’t see the ball.
  3. Deliberate practice is Tom Cruise training for a year – practising 30 motorcycle jumps a day, clocking over 500 skydives, and 13,000 motocross jumps for that epic motorcycle stunt scene in Mission Impossible Dead Reckoning Part One.

Art is struggle.

Deliberate practice isn’t getting your prompt engineering certificate, and writing a good prompt into ChatGPT, Claude, Codex, or whatever-swanky-overhyped-tool an AI founder is selling us on using just to plug the gap in their earning projections and data centre costs.

Deliberate practice is Van Gogh writing,

“Art demands dogged work, work in spite of everything and continuous observation. By dogged, I mean in the first place incessant labour…”

To Theo van Gogh. The Hague, Sunday, 23 July 1882 – Van Gogh Letters

The fruits of that “incessant labour”? Look at how his sense of proportion and placement of forms evolved in less than two years7Source: Drawing on the Right Side of the Brain by Betty Edwards (both artworks presented here courtesy of Rijksmuseum Kröller-Müller Museum, Otterlo):

Deliberate practice is working with the boring tedious sh*t at first, so you appreciate the boring tedious sh*t when you’re doing even more boring tedious sh*t8Another benefit of deliberate practice (I can’t believe I need to write about things that are common sense, but AI and social media has made dumbasses out of some smart people), is it becomes far easier to troubleshoot problems. You can skip to the 2nd or 3rd level of the “five whys” because of intuition. Watch an experienced programmer debug their Copilot-assisted code vs a junior programmer.. Except it doesn’t look boring to other people.

It looks beautiful. 

Deliberate Practice and Invisible Wins

Before I chose to work in D3, creating a multi-series chart in Flourish was a doozy.

You transform your data to have multiple series, like in Excel. So column A is the date, for example, and Column B is Series 1, Series 2, etc. Like so:

And it produces a multi-line chart as so:9I mean, there was a lot of work in wrangling and filtering the data to that point, and then in creating the design brand for the visuals, and then going through Flourish’s millions of options to create it. But. Y’know. You don’t have to think about how the data gets bound to the DOM.

(Click here to view in full screen)

I had the template and understanding of how to draw a single-line chart from Interactive Visualization for the Web by Scott Murray, and D3.JS in Action by Elijah Meeks. Heck, I’d tweaked the exercise in D3.JS in Action to have my brand as a bit of fun and to get some practice.

Making it into multi-line though? That took quite a bit of problem-solving.

Before I could write my line generator code, I had to prep the data, adding an extra key for the region in the JSON:

/**********************/
/*  Prepare the data  */
/**********************/
//Want to visualise several lines for each region. Data needs to be formatted appropriately

//First get an array of regions using .map() to get region names into array (will be non deduped)
const regions1 = data.map(d => d.Region);
//console.log("regions", regions1);

//Use set to dedupe 
const regions2 = new Set(regions1);
//console.log("de-duped regions", regions2);

// Convert set into array
const uniqueRegions = [...regions2];
console.log("regions array", uniqueRegions);

//Loop over this array, going through the data, and returning the regions in each object

const regionsData = uniqueRegions.map(currentRegion => {

  const regionData = data.filter(dataRow => {
    return dataRow.Region === currentRegion;
  });

  return {
    region: currentRegion,
    values: regionData
  };
});

console.log("New Regions Data Array", regionsData);

What did I get for all that work? Oh nothing. The chart still looked like how I’d started it. Just some unstyled axes:

But the console.log? Oh. That’s where the joy is.

That might look like a lot of work, but the craft is working on me.

By understanding how to reformat my data to bind to the DOM10DOM = Document Object Model. the way I want, I’m adding to my problem-solving skills toolset. I’m understanding on an intuitive level how D3 works with JSONs.11I’m a data guy. Which means my “base level” thinking is in terms of spreadsheets and CSVs. One column for one variable. Got a category? Make another column which holds that larger category. E.g. Column C is rent. Column B is the city, e.g. London, Column A is the region, e.g. Greater London. JSONs aren’t too different. You just have keys and values instead of columns and values. Represent the same thing. Except when you’re trying to bind pixels in the DOM to these data values. Then you have to wrap your head around what D3 is trying to do. An analogy, if you remember maths & physics class – going from thinking in terms of degrees with angles, to thinking in terms of radians. An analogy of maths makes you snore – going from thinking in terms of traditional left vs right political values in terms of wealth creation and distribution policies, to thinking in terms of individual priority of values or order/preserve the status quo and maintain security vs harm reduction. A shift in thinking to represent the same underlying state of play.

This adds to my problem-solving toolkit. This allows me to feel confident with more complex visualisations (node trees, I’m coming for you). This helps me change the way I think in other domains.

The work works on you, because art is struggle.

Sign-up form not appearing above? Pesky JavaScript blocker… click here and scroll to the bottom for a vanilla boring Mailchimpform that works on all devices.

Why Reverse Engineer an Existing Project?

As I’ve alluded to, I reverse-engineered one of my own existing projects, instead of starting something new from scratch or reverse-engineering another person’s project.

Specifically, this chart:

(Click here to view in full screen)

Why?

Well, it means I don’t have to focus on any of the design choices. Or even if the story is relevant. I’m not faced with the overwhelming responsibility of choice I face when choosing an idea. Oh, you don’t think that’s agonising? Sylvia Plath captured exactly how I feel each time when choosing which project to start:

“I saw my life branching out before me like the green fig tree in the story. From the tip of every branch, like a fat purple fig, a wonderful future beckoned and winked. One fig was a husband and a happy home and children, and another fig was a famous poet and another fig was a brilliant professor, and another fig was Ee Gee, the amazing editor, and another fig was Europe and Africa and South America, and another fig was Constantin and Socrates and Attila and a pack of other lovers with queer names and offbeat professions, and another fig was an Olympic lady crew champion, and beyond and above these figs were many more figs I couldn’t quite make out. I saw myself sitting in the crotch of this fig tree, starving to death, just because I couldn’t make up my mind which of the figs I would choose. I wanted each and every one of them, but choosing one meant losing all the rest, and, as I sat there, unable to decide, the figs began to wrinkle and go black, and, one by one, they plopped to the ground at my feet.”

The Bell Jar, Sylvia Plath12Thanks Craig Beadle for the h/t on this quote ages ago.

Replace life with time. Replace life paths with projects. Keep the existential crisis.

I don’t need to focus on any of these things. I can focus purely on the craft. On building. On mastery.

It means what I’m setting out to do is achievable, yet ambitious. Remake something I’ve done, which did well (featured in my exclusive newsletter on London Affordability). Yet do it in a demanding imperative-like language that asks you to write code for the axes. Verbose so you can create art.

40 lines of code to read your data, declare scales, and draw axes is not unheard of. You wouldn’t use D3 to create a simple multi-line chart. But by using it in this familiar and unfamiliar environment, I’m building the foundation for a more ambitious project.

What I’m trying to say is I’m learning damnit. And I’m learning in a way that I won’t get overwhelmed and give up. I’m learning so I can do more ambitious things.

That’s why art is struggle. That’s why it involves mastery. And that’s why you’re foolish to rely on AI to think for you.13Use it for the repetitive parts. The mind-numbing tedious parts that are already part of your muscle memory. Don’t use it for the abstract thinking parts.

Let me give an example that illustrates some of what I’m talking about.

The Struggle of Adding Animations

Getting to Static

My first visual milestone was to recreate a static, no animations or transitions, version of my original Flourish chart. This took approximately 14 days of half hour to 1 hour morning coding sessions. Most of this is because this is my first work-from-scratch project in D3.

Original flourish chart for reference.

Time taken: (Once I had the data) – Excluding the 500+ hours of experience I have working with Flourish, or the 2-5 hours I spent making my brand templates (excluding the brand research)… 15 minutes.

The first step – building the html and javascipt structure, and reformatting the data. Visual milestone: hey some axes!

Time taken: 2-4 hours.

Something that’s starting to look like the original. Paths are drawn, and have the same colour scale as the original.

Time taken: 2 hours

I don’t want to admit the rabbit holes I nearly went down on adding labels to this chart (I wanted to space them out like the original, but the technicalities of using the d3-force module were beyond me…for now). But we’ve got labels, and they’re placed in more strategic places than the original. Something I couldn’t do with Flourish.

Time taken: 4 hours

I had to tweak my wrangle code so that London and England had an extra key, allowing me to style them as dashed paths. But this chart is starting to look similar to the original, with some slight tweaks (label placement, font style). Tweaks that were impossible with vanilla Flourish.

Time taken: 2 hours

Total time taken to get here: 12-14 hours

Adding Transitions and Animations

After 14 days of toil and trouble, I had a static recreation of my original Flourish chart. I felt proud. Sure, there were easier ways to make this (like in Flourish), but I’d done it with D3. The tool I’d always wanted to produce visuals in.

And it was time to unlock D3’s secret sauce: animated transitions.

In case you were born in the data visualisation landscape yesterday, one of the reasons D3 took off14I believe that one of the problems of our modern age is the fact that a near double-digit percentage of people in education and working age years have way, way too much time on their hands. Otherwise, how can a simple post where I write why I think D3 is verbose lead to a Hacker News comment thread where a dozen people shout me down on using D3, instead of a more modern framework, and two friends text me saying “don’t use D3. Use React”. Bloody hell, guys. Tools have pros and cons. And you have to find the pros and cons that work for you. I’m a Data Journalist, not a Web Developer. D3 has its cons. But animated transitions, and automagic scaling of data out of the box are incredible pros. I’ll stick with it for now.  , is because it supports animating smoothly between different transition states. Something that’s pretty handy (and cool!) in data visualisation.

Adding animations is easy for most elements. Later on, I’d animate the axis line and the data points by chaining .transition() in the element where the SVG is drawn. E.g. asking D3 to draw the y-axis goes from:

innerChart
	.append(“g”)				     //Append SVG group
	.attr(“class”, “axis-y”)	//Give SVG group a css class		
	.call(leftAxis);			       //Call leftAxis which I’ve defined earlier.

To:

innerChart
	.append(“g”)
		.attr(“opacity”, 0)		                //Set opacity to 0 for fade transition effect
	.attr(“class”, “axis-y”)
	.call(leftAxis)
	.transition(transition_const)		//Call transition where transition_const is defined elsewhere
		.attr(“opacity”, 1);		                //Transition to fade in.

Transition slowed to two seconds on the second chart to illustrate the difference. Make sure to refresh your page to see the loading animation.

Nothing too different here.

But the real thing I want to animate is the data in the chart. Just like the original Flourish chart (click here to see it in full screen and watch how it animates on load), I want the data – the curved lines – to animate on load.

And that’s when I ran into a right of passage for every D3 data visualizer. Animating data paths is hard.

Using Illusions

In D3 and SVG, the <circle> attribute creates a circle. The <line> attribute creates a line. The <rect> attribute…you get the idea.

But what happens if you want to create a custom drawing? Do you create a bunch of lines? No…but you do something a bit similar.

The <path> attribute takes multiple arguments to create multiple straight or curved lines to create your desired shape.

What I’m writing is either going to make sense if you’ve worked with vectors before, or is going to sound like gobbledygook. And if I take some time to explain how paths work in SVG, this article might double in size. So here’s the best explainer Google found for me. It’s an interactive tool that shows you how different SVG paths are made. Try it out. Also, the MDN docs on SVG paths are here.

What this all means is you can’t just chain .transition() to animate your path.

Transition requires the enter, update, and exit states of the DOM. This is straightforward for most SVG objects. In the beginning, there is no text. Then there is text. And sometimes, maybe, the text dies and disappears. So says the gospel of D3.

But for a path? Which is a fancy, calculated object combining various shapes? No dice.

After some trial and error and some googling, I stumbled on Visual Cinnamon’s tutorial on this specific problem.  

The trick isn’t to animate the paths. The trick is to create an illusion of the paths being animated. You add a mask on the path, and you pull it off, making it look like it’s animated.15I’m telling lies. It’s more complicated than this. You actually create an invisible path, that’s the same path and length as your path, and then shove it onto the right of your path. You then offset things so that the starting state is this invisible path. Then you decrease this offset to 0, and in doing this, the invisible path decreases to 0, bringing in your actual path from the left, making it look like it’s being animated in. If you followed along with that, congrats, you’re far smarter than me. If you didn’t, here’s a visual explainer by Visual Cinnamon which does a far better job of explaining this than I ever could, and it’s still complicated. The point is. You’re creating the illusion of the path being animated.

It took a bit of time to figure out how to apply Visual Cinnamon’s tactic to my code, but I finally did it. If you can understand what’s going on, then you deserve a gold star. No. I won’t give you one. Go to the shops and give yourself one:

//Animate solid lines - stroke array to line technique
paths.filter(".solid-line")
  .each(function(d)
    d.totalLength = this.getTotalLength();
  })
  .attr("stroke-dasharray", d => `${d.totalLength} ${d.totalLength}`)
  .attr("stroke-dashoffset", d => d.totalLength)
  .transition(transition_const)
    .attr("stroke-dashoffset", 0);

The result?

(First chart is the static chart for comparison. Second is the animated one. Refresh the page to see the animation. Slowed duration to two seconds to illustrate. Notice how the solid lines smoothly animate, and the dashed ones, uh, don’t.)

Look how smoothly those solid lines animate. Isn’t that cool? Oh. You don’t think so? Okay. Well, I do. Because it took a lot of trial and error and metaphorically smashing my head on the keyboard to get here.

It’s god-damn beautiful, even though the dashed lines, uh, aren’t working.

In fact, they look like they’re ants marching along the trail, instead of smoothly drawing out. Bloody hell. Why can’t things be simple? Oh. Wait. Because art is struggle.

Sign-up form not appearing above? Pesky JavaScript blocker… click here and scroll to the bottom for a vanilla boring Mailchimpform that works on all devices.

Animating the Dashed Lines

While the trick to animating the solid lines was difficult, a part of me intuitively got what was going on. But obviously, I didn’t understand it enough. Because when I tried it with the dashed lines, I was just getting that ants marching effect.

It was time to put the keyboard away. I had to understand what was going on.

I pulled up the tutorial and read it slowly from beginning to end, instead of just scanning to the fix I needed. On the side, I had an A4 pad of paper. I sat there, and tried to explain to a dumber version of me what was going on16The Feynman Technique. I included some diagrams. Slowly, the nuances of the solution became apparent to me.

I looked back through my code and thought things through with what I just learned. I hacked away. I attempted a few fixes. Each failed trial taught me something new about this concept. And then. I got something that worked.

(Previous chart there for comparison. Refresh the page to see the animation. Notice how the dashed lines draw smoothly with the rest of the chart.)

This animation exercise was struggle. It helped me learn a new way of thinking. How to use maths to create visual illusions in D3. It expanded my mind.

Next time I hit a similar D3 problem, I’ll think “what about if I create a visual mask and slide it off?”.

Art is struggle.

Bisecting and Creating Tooltips

Or let’s look at creating tooltips. Hover over both of these (or tap on your phone…but remember this post is optimised for the web) to bring up the tooltips. Compare the two:

(First chart is a simpler tooltip method. Second is a more intuitive, but harder to build, tooltip method.)

The second one is nicer and easier to use, right? In fact, it’s incredibly hard to get a tooltip with the first right?

Most textbooks and tutorials provide guidance on how to build the first. You bind an event listener to each data point in the DOM, and when your mouse hovers over, you have a bit of code to pop out text that references the datum bound to that element in the DOM.

The second, though? The more intuitive approach? I had to really dig to find it. Luckily, I had my good old original Flourish chart as a reference. Visually, I knew what I wanted to create.

It took a bit of digging, but I discovered the d3.bisect method. There were tutorials out there by data visualisers who’d run into the same problem – how to create tooltips which just worked with line charts.

I won’t share the code, but the concept is simple. You create an invisible grid, like putting tracing paper on top of the graph paper you’d use in maths class, and then each point in the grid is mapped to the closest data point. D3.bisect() allows your mouse to “snap” to the closest data point, providing a much nicer tooltip.

This is, again, a new way of thinking. In my day job as a Data Analyst, we use bandings and cut-offs all the time as a technique. But what about if I were to go two-dimensional? Could a grid system achieve a better effect for some gnarly problems?

Deliberate Practice and the Dunning-Kreuger effect

You have probably heard of the Dunning-Kreuger effect, but in case you haven’t, I’ll leave the All-Knowing Wikipedia definition below:

“The Dunning–Kruger effect is a cognitive bias that describes the systematic tendency of people with low ability in a specific area to give overly positive assessments of this ability. The term may also describe the tendency of high performers to underestimate their skills.”17It’s worth adding the last sentence in the definition: “In popular culture, the Dunning–Kruger effect is sometimes misunderstood as claiming that people with low intelligence are generally overconfident, instead of describing the specific overconfidence of people unskilled at particular areas”. It’s not that dumb people overestimate their ability to do something…but that smart people are also likely to overestimate their ability in an area where they have no expertise. Doctors believing they understand economics for instance, or successful CEOs believing they’ll be good at politics…

Dunning-Kruger Effect – Wikipedia

I fell for this trap.

After adding animations and a bisected tooltip and a few other tweaks, I wanted to replicate the shadow effect the tooltip has in the Flourish chart.

The shadow effect around the tooltip is very slight, but it helps add depth to the design.

Little naïve me thought this was as simple as adding a box-shadow effect to the CSS.

Oh, how wrong I was.

Added the CSS rule… nothing.

Debugged my selectors… nothing.

I thought I was losing my mind. Reality was messing with me. Soon, a pig would fly out of the window. This was a five-minute job. How had I spent a half-hour already?

I went to Dr Google, and it clicked.

I’d forgotten that that textbox wasn’t an HTML element, but an SVG element. In fact, this is a basic mistake when working with D3. From the first book to teach me D3:

“If you ever find yourself trying to style SVG elements, but for some reason the stupid CSS code just isn’t working (Grrr!), I suggest you take a deep breath, pause, and then review your property names very closely to ensure you’re using SVG names, not CSS ones.“

Interactive Data Visualization for the Web by Scott Murray

I figured out that I needed to use the `<feDropShadow>` SVG filter. I played around with the variables to achieve my desired effect.

(Hover over to see the difference in the tooltips. It’s slight, but the slight visual depth is worth it.)

I thought this was a five-minute job. It took half an hour. Four times larger than my estimate.

Why? Because of the Dunning-Kreuger effect. I’ve been styling CSS boxes for years. I thought it was that simple with a D3 tooltip.

But I wouldn’t have found out this gap in my knowledge without deliberate practice.

This is humbling. Mastery is humbling. It teaches you what you don’t know. It inspires you to learn and improve.

You get better. And you get humble as well.

The Beauty of Struggle

This is the original chart I made in Flourish.

My first prototype in D3 looked like this ugly thing.

Then I got it to look pretty stylish.

Then after much trial and error I added animations.

This is the last version before I threw in the towel (I was trying to get it responsive). Not shabby when you compare it to the original… and I’ve been able to add a few extra personal twists.

I’m pretty proud of what I’ve done. Is it perfect? Oh hell no. But a year ago, recreating one of my Flourish charts in D3 felt far off…nearly impossible.

Seeing the progress that’s been made. Understanding the pain of the lessons I learned along the way. That’s been meaningful. It’s been a struggle. And it’s been beautiful.

Art is Struggle (or Deliberate Practice is Beautiful)

“In the early 1990s, Anders Ericsson, a colleague of Neil Charness at Florida State University, coined the term ‘deliberate practice’ to describe this style of serious study, defining it formally as an ‘activity designed, typically by a teacher, for the sole purpose of effectively improving specific aspects of an individual’s performance. As hundreds of follow-up studies have since shown, deliberate practice provides the key to excellence in a diverse array of fields, among which are chess, medicine, auditing, computer programming, bridge, physics…”

So Good They Can’t Ignore You by Cal Newport

“If we are learning a complex skill… we must master a series of simple skills, one on top of the other. Each time one skill becomes automatic, the mind is freed up to focus on the higher one. At the very end of this process, when there are no more simple skills to learn, the brain has assimilated an incredible amount of information, all of which has become internalized, part of our nervous system. The whole complex skill is now inside us and at our fingertips. We are thinking, but in a different way—with the body and mind completely fused. We are transformed. We possess a form of intelligence that allows us to approximate the instinctual power of animals, but only through a conscious deliberate, and extended practice.”

Mastery by Robert Greene

I’ve been angry about AI.

Not because they’ve been training AI on works protected by copyright.

Not because Senior Management at large companies, or AI companies, have been pushing us to use AI, but they don’t have a clue how.

Not even because AI firm valuations are making the threat of a deep global recession likelier by the day.

No, I’ve been angry because in all the discussions of AI, the value of craftsmanship is being ignored.

We’ve been thinking we’re the blacksmiths or carpenters about to go out of a job because of the factory. Because there are now tools that can play a very, very good game of Guess the Next Word.

… and we’ve been forgetting that craftsmanship isn’t just there for its capitalist value. Craftsmanship is because of how the work works on us.

Craftsmanship rewires our brains. It makes us more grounded. More humble. More, dare I say it, wise.

There are entrepreneurs and investors on the hook to plug a financial black hole the size of Greece’s GDP18As of writing (December 2025) Analysis by HSBC found that OpenAI has a $207bn funding hole it needs to plug for its $792bn projected data centre rental costs by 2030 [FT, paywall]. The incestuous financial AI ecosystem [Bloomberg, paywall] means lots of players are carrying various debt commitments [FT paywall] that amount to $100bn, and OpenAI signed $1.4tn worth of deals to meet its spending commitments (data centre rental costs). The tech industry has already invested over $605 billion in capex investments in AI, and one estimate says AI needs to add $2 trillion in AI revenue over the next four years for this to be worthwhile. That’s about $500 billion a year. The nominal GDP of Greece is $282 billion. because of AI investment. And I wish them luck. I’m a techno-optimist at heart. Creative destruction has consistently brought us wealth and progress as a society.

But god-damn it, stop selling us the idea that AI now means anyone can be an artist.

Anyone could always be an artist. They just had to sacrifice time. They had to embrace pain. They had to embrace boredom. They had to struggle.

Art is beautiful, god damn it. So, stop shilling how to make it ugly.

Data Essays Inspired by Struggle.

I can’t promise you the answers to life’s deep questions, but I can promise you a beautiful newsletter of data essays that tackle some questions you’re thinking about.

For three to four years, I wrestled with the question: Is buying a home in London a) achievable? b) worth it? The result was a three-part interactive data essay diving deep into London affordability. It’s available only in my email newsletter.

Sign up below to get instant access

Sign-up form not appearing above? Pesky JavaScript blocker… click here and scroll to the bottom for a vanilla boring Mailchimpform that works on all devices.

Footnotes

  • 1
    I hate using big words. But I dare you to google the definition of “neophyte” and try to find a better word to capture what I’m talking about. ↩︎
  • 2
    A16z’s content is contrarian for a reason, it takes outlandish stances to force people to think. Still, this article, based on a single premise, that the war for tech talent will totally transform because of the rise of one-man startups, because of AI, really takes the cake. ↩︎
  • 3
    Let’s pretend and ignore the fact that the tool just delivered an output trained on a bunch of illustrators’ efforts who’ve received no compensation for it… ↩︎
  • 4
    Shout out to the amazing John Oliver for surfacing this in his excellent episode on AI slop. ↩︎
  • 5
    The missing knowledge in my tech stack has been JavaScript (knowledge of HTML, CSS, and programming concepts via Python (and SQL, though that’s a declarative language). Because I don’t want to become a web developer, just someone who can make gorgeous data-viz, I’ve skipped learning JavaScript and gone straight to learning D3.JS. Filling in the blanks where needed. Good is better than perfect. ↩︎
  • 6
    In Desert Storm, Cesar Rodriguez and his wingman were lured into a trap by Iraqi MiG fighters. To escape, he performed a series of bold, perfectly timed, and lightning-fast manoeuvres that saved him and his wingman. The encounter lasted barely four minutes, with one of the dogfights lasting seconds. Upon watching the videos during the debrief, he had almost no recollection of any of it or where his ideas had come from. His intuition had come into play; he couldn’t explain how he had performed tighter turning circles on his opponent while keeping the desert floor in mind. Or why he’d chosen to jettison his fuel tanks at the beginning of the dogfight, which gave him greater speed and manoeuverability. That’s the power of deliberate practice. Source: Mastery by Robert Greene ↩︎
  • 7
  • 8
    Another benefit of deliberate practice (I can’t believe I need to write about things that are common sense, but AI and social media has made dumbasses out of some smart people), is it becomes far easier to troubleshoot problems. You can skip to the 2nd or 3rd level of the “five whys” because of intuition. Watch an experienced programmer debug their Copilot-assisted code vs a junior programmer. ↩︎
  • 9
    I mean, there was a lot of work in wrangling and filtering the data to that point, and then in creating the design brand for the visuals, and then going through Flourish’s millions of options to create it. But. Y’know. You don’t have to think about how the data gets bound to the DOM. ↩︎
  • 10
    DOM = Document Object Model. ↩︎
  • 11
    I’m a data guy. Which means my “base level” thinking is in terms of spreadsheets and CSVs. One column for one variable. Got a category? Make another column which holds that larger category. E.g. Column C is rent. Column B is the city, e.g. London, Column A is the region, e.g. Greater London. JSONs aren’t too different. You just have keys and values instead of columns and values. Represent the same thing. Except when you’re trying to bind pixels in the DOM to these data values. Then you have to wrap your head around what D3 is trying to do. An analogy, if you remember maths & physics class – going from thinking in terms of degrees with angles, to thinking in terms of radians. An analogy of maths makes you snore – going from thinking in terms of traditional left vs right political values in terms of wealth creation and distribution policies, to thinking in terms of individual priority of values or order/preserve the status quo and maintain security vs harm reduction. A shift in thinking to represent the same underlying state of play. ↩︎
  • 12
    Thanks Craig Beadle for the h/t on this quote ages ago. ↩︎
  • 13
    Use it for the repetitive parts. The mind-numbing tedious parts that are already part of your muscle memory. Don’t use it for the abstract thinking parts. ↩︎
  • 14
    I believe that one of the problems of our modern age is the fact that a near double-digit percentage of people in education and working age years have way, way too much time on their hands. Otherwise, how can a simple post where I write why I think D3 is verbose lead to a Hacker News comment thread where a dozen people shout me down on using D3, instead of a more modern framework, and two friends text me saying “don’t use D3. Use React”. Bloody hell, guys. Tools have pros and cons. And you have to find the pros and cons that work for you. I’m a Data Journalist, not a Web Developer. D3 has its cons. But animated transitions, and automagic scaling of data out of the box are incredible pros. I’ll stick with it for now.   ↩︎
  • 15
    I’m telling lies. It’s more complicated than this. You actually create an invisible path, that’s the same path and length as your path, and then shove it onto the right of your path. You then offset things so that the starting state is this invisible path. Then you decrease this offset to 0, and in doing this, the invisible path decreases to 0, bringing in your actual path from the left, making it look like it’s being animated in. If you followed along with that, congrats, you’re far smarter than me. If you didn’t, here’s a visual explainer by Visual Cinnamon which does a far better job of explaining this than I ever could, and it’s still complicated. The point is. You’re creating the illusion of the path being animated. ↩︎
  • 16
  • 17
    It’s worth adding the last sentence in the definition: “In popular culture, the Dunning–Kruger effect is sometimes misunderstood as claiming that people with low intelligence are generally overconfident, instead of describing the specific overconfidence of people unskilled at particular areas”. It’s not that dumb people overestimate their ability to do something…but that smart people are also likely to overestimate their ability in an area where they have no expertise. Doctors believing they understand economics for instance, or successful CEOs believing they’ll be good at politics… ↩︎
  • 18
    As of writing (December 2025) Analysis by HSBC found that OpenAI has a $207bn funding hole it needs to plug for its $792bn projected data centre rental costs by 2030 [FT, paywall]. The incestuous financial AI ecosystem [Bloomberg, paywall] means lots of players are carrying various debt commitments [FT paywall] that amount to $100bn, and OpenAI signed $1.4tn worth of deals to meet its spending commitments (data centre rental costs). The tech industry has already invested over $605 billion in capex investments in AI, and one estimate says AI needs to add $2 trillion in AI revenue over the next four years for this to be worthwhile. That’s about $500 billion a year. The nominal GDP of Greece is $282 billion. ↩︎