How I Write
And some questions on rationality, destruction, and disobedience
Updates:
I was on the How I Write podcast with David Perell. We deconstructed two essays (Consider the Lobster by David Foster Wallace and Shooting an Elephant by George Orwell) and got into process/technology. Link and timestamps below.
essayarchitecture.com is live! Enter your email there if you’re interested in more frequent product updates (starting soon). There’s also a form to join the beta and try it early. Thanks for your patience if you’re on the waitlist. I’m onboarding writers one at a time, and attempting to make meaningful updates between sessions.
Upcoming essays:
“The Babbling Idiot and the Tribe.” This one is a Response to Food of the Gods, the book where Terence McKenna popularized the Stone Ape Theory. Did psychedelic mushrooms played a role in human evolution?
“45 predictions on writing in 2045.” Pretty self-explanatory. The goal of this thought experiment is to map out how weird the future of writing might get and to find my place in it. The good news is that I think the timeless reasons to write hold up, and even evolve in some positive ways.
Logs: March and April were intense for me, and I feel guilty for not sharing as much as I planned on, but I still managed to publish 25k words without realizing. Here are my logs for March (locked), April (open), and May (in progress). This is a reminder that, if you’re low on time, you can lower the scope to a single paragraph, and put any mild epiphanies you have into words.
How I Write:
Introduction
Opening (0:00)
Deconstructing two classic essays
The Pick One Volcano principle (to limit your scope) (10:16)
Deconstructing the maximalism of David Foster Wallace (18:24)
Personal experience = biography + interiority + outlook (22:35)
On sound and rhyme in prose (41:01)
On process (reading, writing, editing)
Balancing intuition and analysis (44:02)
Why is voice important? (53:12)
Privacy can unlock writing voice (57:59)
How to ask for feedback (59:50)
How to read analytically (1:01:38)
A Pattern Language
A pattern language is a hierarchical network of constraints (1:02:25)
Storytelling is stretching the unknown across time (1:06:38)
On titles and the way they sound (1:07:42)
Repetition, rhythm, rhyme (1:10:04)
Word choice is negotiation between concision, precision, invention (1:11:40)
What causes reader’s trance? (1:15:05)
AI, Education, & the Future
On banning ChatGPT in schools (1:24:54)
On adapting to the rate of AI progress (1:27:38)
Will you still write if AI gets phenomenal at writing? (1:41:46)
Questions:
If essay means “to try,” then a draft is the trial of a fuzzy question in your head. A draft isn’t your final answer, but your attempt to formulate a more specific question. If you have clarity on the question you're asking, the essay usually comes out a lot easier. In that spirit, I’m using this space to frame some recent thoughts as questions. This is an invitation for you to share your own answers (either leave a comment, or use these as prompts for essays of your own).
q1. (rationality):
Why does popular creativity advice downplay the value of technique?
The spirit of Essay Architecture runs counter to the spirit of our times. It feels like everything runs on vibes. Maybe it’s dramatic to call our moment a “revolt against reason,” but it seems many writers believe the mind is detrimental to making things. Is all thinking overthinking? It’s increasingly popular to justify anything you do with your taste or intuition—to celebrate laziness, cheating, slop, and sloppiness. Typos are in. Effort is mocked.
It’s probably fair to call Rick Ruben our leading figurehead in non-rational creativity. There’s much I respect about A Creative Act (2023)—I think psychological attunement is real—I just think it’s a bad starting point for beginners. Ruben is famous for unlocking artists who were a decade or more into their journey. Surrendering your ego makes a lot of sense if you are post-peak Johnny Cash. But if you’re not fluent in your medium yet, it might be better to focus on discipline and craft instead of “being open to the world to make something you love.” Learn the scales!
I’ve been wondering when, how, and why non-rational creativity became so popular. The best selling book in this genre is The Artists Way (1992) by Julia Cameron. She de-emphasizes the role of technique; instead she wants you to dissolve resistance, experiment, and heal emotionally. She’s most known for “morning pages,” where you dump three pages into a journal each morning and never read it again. She sold five million copies. How pivotal was this in framing creativity as a type of therapy?
I think we should fuse the psychological and the technical—the problem is that the technical realities of an art are hard enough to dissuade beginners. Consider the accessibility of Ruben’s “I just know what I like” compared to Christopher Alexander’s language of 253 patterns. A consequence of democratizing creativity is that we may have inadvertently sterilized our technical creative knowledge. There will always be 100x more beginners than practicing artists, and so it’s far more profitable to share “how to start” than “how to actually do the thing.”
If you have any thoughts on these dynamics, let me know. I wonder if it’s as simple as this: writers justify an avoidance of editing because it’s hard. In addition to building a tool that makes editing approachable, I probably also need to recast a new myth to make editing emotionally and psychologically approachable (a v2 of alchemy of the rewrite). There’s joy in doing hard things.
q2. (destruction):
When is it worth starting over from scratch?
To make something great—an essay, a song, a building, software, anything—it often helps to destroy the design concept and start over, possibly multiple times. Rewriting is underrated. I won’t expand on all the reasons why we avoid it (ego, impatience, the sunk-cost fallacy, the secret hope that this draft is the draft); instead, I’m wondering when it’s actually worth it. You can’t rewrite everything. If you did, every idea would be in permanent limbo. Most ideas are probably fine sharing at 80% of their potential—my logs are maybe at 5%—but there are a few occasions where perfectionism is worth it.
This is relevant to me now because I’m doing a full rebuild on significant parts of my app (on v3 now). After evolving this over 25 interviews, the UX is getting much smoother, but I’m haunted by what might be a limit in suggestion precision. My problem is that I’ve been training it to be an expert analyst, not an expert editor. So even though I’m close to my goal of +90% accuracy in how it evaluates each pattern, and even though it organizes all your scores into a sleek interactive spider graph, its advice on how to improve is only slightly better than a chatbot. Right now, it can’t match the granularity of how I personally give feedback (“you’re doing something similar in p(6) and p(13), what if you combine those, move it up to p(2) and expand on X to achieve Y?”). It makes me wonder—could I rebuild the analysis engine with the main goal of giving great feedback?
For this to work, I need to get absurdly granular. I already have a detailed scoring system (with 81 criteria), but it currently only understands your draft as a single monolithic object. Essays have sub-objects right? They have sections, paragraphs, sentences, and words. If I want paragraph-level suggestions, then I need paragraph-level analysis. If I want to match the intuition of a great human editor, maybe each draft you upload needs to run 5,000 micro-prompts. This feels right, but sounds technically impossible.
But the beauty of starting over is that 1) you understand the problem 10x better, and 2) the context around you is different. I’ve been playing with some new models that are less than a month old, and I’m realizing that I can feed in an array of paragraphs and run hundreds of precise mini-evaluations, all through a single API call. Not only will the suggestions be better, but the whole thing might be cheaper and faster too.
So when is it worth starting over? Maybe: 1) if it has high relative importance compared to your other projects, 2) if the sphere of improvement matters (ie: are you fixing the main thing or a frilly detail?), and 3) if the time it takes to rebuild is small compared to the time the new thing will be useful for (ie: a 2-month delay is okay if the new thing sticks around for 2 years). Any other heuristics?
q3. (disobedience):
Do you want technology to agree with you or challenge you?
When people ask me, “why would I use your app over ChatGPT to get feedback?” I point them to the latest dilemma with 4o. Last week, OpenAI had to rollback their latest 4o model because it was too sycophantic. It’s too obedient. It’s too willing to tell users exactly what they want to hear. There are many screenshots online of ChatGPT calling its user a gifted genius. You can probably guess why this makes it a bad editor: it’s incentivized to not challenge you.
During my 1:1 beta calls, I’ve noticed a few writers score low on a pattern and say, “That makes sense because I wasn’t really going for that.” Maybe a memoirist doesn’t care about the Argument pattern, or an intellectual doesn’t care for Experience or Tone. If I were to follow the best practices of product design, I’d understand what each writer wants and give them exactly that. The more I think about it though, the more I realized that the goal of Essay Architecture is to help you expand your conception of what an essay can be.
This got me to realize that product maxims for an educational tool might be very different from SaaS or content marketing. If you’re building a writing tool and you listen to the average user, you will likely end up with an advanced auto-complete that provides and implements all your edits in a single click. No writing necessary! Earlier this year I published Thank you for the roses, about how TikTok creators devolve into insanity when they optimize for market signals. The Internet enables everyone to get high-volume real-time metrics, and there's a risk in watering down your vision to become a mirror to the market.
If the default is for every app to conform to its user, maybe it’s refreshing to have an editing app that has a single quality standard that applies to everyone? The risk here is obvious. The tool won’t bend to make you immediately happy. It’s disobedient—but good teachers challenge you. Learning involves shedding old models to build new ones, and that initial clash could be enough to scare someone away.
Maybe there’s a way to fuse validation with abrasiveness. For example, good feedback starts by acknowledging the writer’s vision before pushing them further: “You’ve done great with patterns A, B, and C, but you seem to ignore patterns X, Y, and Z, and maybe you think those two sets don’t typically go together, but here’s a link to a classic essay that shows all these patterns working together in a single piece.”
Productive Friction
The shared theme among all 3 questions is the value of “productive friction.” It’s easier to run on vibes than to practice technique. It’s easier to stop at 80% than to start over. It’s easier to build something people want than to build something people need. It makes sense that a technology-obsessed culture strives to eliminate all friction, but a creative practice isn’t forged from inspiration, comfort, or validation—but how you evolve under pressure.
What do you think?
Why does popular creativity advice downplay the value of technique?
When is it worth starting over from scratch?
Do you want technology to agree with you or challenge you?
As I understand it, I think a big part of the promise of your approach is that you are offering a structured way to confront your weaknesses. It’s daunting enough when you know them. Harder when you’re unclear what they are. And harder still if you have no idea what to do about them.
There is so much in this that I agree with - thank you for put it so clearly.
I have worked with writers for many years and it is so much easier to give 'starting' advice than how to 'keep going'. This is where the insight bias come in and the concept of disfluency (subject of my next book). Writing craft, or indeed any form of mastery is hard. Telling people to find their flow can actually do more damage as their expectations get challenged, they doubt their abilities and give up prematurely when it gets hard.