Year of the wiki

Essay Architecture undergoes a phase change

Thanks for supporting Essay Architecture. After 4 years of writing online, 2024 was the year I found some exciting momentum with it. I won a fellowship grant from O’Shaughnessy Ventures, and it let me make this project my main focus. I changed the name of this publication, turned on paid subscriptions, wrote some foundational essays (ie: A Pattern Language and Essay Writing as Personal Sovereignty), shared Dean’s List, launched Essay Club, and made good progress on software that gives you feedback on your essays.

Today’s post is about how I see the project evolving in 2025, and what you can expect from this Substack.

The obligatory galaxy brain IMG from anyone 30 days into an Obsidian fever.

January is a natural time to celebrate all that’s working and to let die whatever isn’t. To me, it’s less about looking backwards or forwards and more about ruthlessly pruning. After all, it’s a new year, and the origin of the word resolution¹ means “to release.” It’s a time to audit and purge the physical and mental, the small and big. And so in the last week I donated books and clothes, nuked the fridge, emptied whole drawers of forgotten sundries,² unpinned uninspiring posters, and—after much needles toil—decided:

“Essay Architecture doesn’t need to exist as a book, actually.”

I conceived this project in November of 2023, and the idea was to release one chapter at a time on Substack until I had the material for a physical book (which at one point I joked would be so fat that it would be a cube). I’m proud of the 14 Essay Architecture posts I shared in 2024, but this process isn’t suited for what’s ahead. I need to molt³ (again).

Ultimately, my problem: I had a linear process for a non-linear project.⁴ I went through my 27 patterns, one-by-one, and turned each one into a polished, illustrated, longform essay. This should have been suspicious to me: I wouldn’t want readers to read the finished book chapter-by-chapter, so why should my process be so linear? The omega point of Essay Architecture is to create a history-infused draft analyzer that finds your compositional blindspots and points you to specific chapters (“Nice essay, you’re doing great with X, Y, Z, but you should check out Chapters 3, 5, and 17 to improve your…”).

I still recommend my chapters on thesis, microcosm, response, catalyst, and material, but the mode of turning research into prose is very different from what’s required to make an actually good AI feedback app (one that doesn’t bullshit you). No matter how hard I work, I can’t make a great app and a great book. I need to make an app that is a book. (Not a book that is an app—big difference.⁵)

So the book isn’t dead; I now get to re-imagine what a digital (text)book can be. How would Christopher Alexander release A Pattern Language in 2025? What format can a non-linear always-evolving corpus of ideas take? What would let me publish a v1 of all 40 chapters by February 1st?

The answer to these three questions is the same: an essay wiki. First I’ll share why I think this format best captures the spirit of the project, and then I’ll get into what you might expect from this Substack moving forward.

---

The Essay Architecture framework is already wikiesque_

A pattern language has two modes of organization: hierarchy and associations. A wiki captures both. The hierarchy locates a concept in the framework, the associations link it to related concepts. For example, the Image pattern is located under Voice/Sight(9)/Image(9.1), and if you want to know the other patterns that image-making is linked to, you might want to check out Microcosm (2.1), Title/Mystery (3.1), Hook (4.3), Paragraph/Finale (6.3), Motif (9.3), etc.

This image from August was my first attempt to visualize how the patterns are related. It so obviously screams “make a wiki.” I underestimated the blinding effects of book mind.

I could write 5,000 words of prose about imagery, but the inherent linearity of an essay clashes with the curiosity of a self-directed learner. While prose is best at conveying a single arc, it’s not great at conveying complex maps. The only way to learn a hyperobject⁶ is to burrow through it. A wiki lets you locate, skim, zoom into, and link out to related ideas. The ultimate affirmation here is that Christopher Alexander—the architect who wrote A Pattern Language in 1978, and a big influence on this project—was a direct inspiration to the inventor of the wiki in 1994.⁷

So the plan now is to build a wiki directly into this app.⁸ You upload your draft, wait ~3 minutes, and get a feedback report—it gives you insights into your strengths and weaknesses, and each hyperlinked term is a portal into the wikiverse. This switches learning from a push to a pull model: you don’t have to trek through a textbook and cram concepts; instead, the right theory wraps itself around you at the right time.

AI isn’t writing your sentences, AI is the bridge between your shitty first draft and a human-created network of composition theory.

The wiki format also makes the project faster, flexible, collaborative, and sustainable. Unlike essays (which are like tightly-wound puzzles of prose), a wiki page is made of modules that exist in an outline. There doesn’t need to be a final/ultimate order. Over months and years, I can add/edit/arrange/delete. Now the project feels like an infinite game. It’s rejuvenating. I’m on track to release a v1 of all the core pages in February. From there, I hope it evolves into an online home for essayists to learn from and contribute to.

---

What to expect from this Substack in 2025_

• Monthly updates: I’ll send updates like this near the beginning of each month. As the project evolves I imagine I’ll share links, forms, videos, the challenges of building something, etc. I’ll also curate any new writings or wikis.

• Essay Architecture posts: The goal here is to send one post a month for paid subscribers. These might take the form of essay reviews, pattern deep-dives, visual breakdowns, or something unpredictable (I want to break from linear chapters, and instead share something I’m actively studying).

• Longform essays: I have three of these that are nearly finished, and hope to ship them monthly. Even though the title of this thing is Essay Architecture, it’s important to me that I ship essays on completely random topics (Apple Vision Pro, human evolution, NPC streamers on TikTok, etc.). By escaping the meta-project and sharing my own essays on anything I feel like, I stay grounded in the process and can speak to it better.

• Essay Club: Last October I started a publishing accountability group on my Founding Members tier. Every month we set a goal, have 6-7 feedback exchanges on Zoom, and then publish by the 1st. I track and share streaks, so if you miss, you restart at 0. If you want to build a longterm writing habit, check it out. I felt compelled to start this after the closing of Write of Passage (the writing course I took in 2020 and have worked closely with since). It’s worth noting, I’m a member of Essay Club too. I have ambitious publishing goals⁹ and I see this group as a forcing function to make it possible.

• Substack Chat: I just launched this yesterday and want to experiment with it this year. The settings are somewhat cryptic, so I apologize if you get bombed with notifications (you can change these in your app settings). Two goals here: 1) to share emerging ideas, research, and questions around Essay Architecture, and 2) to create a place for Essay Club to chat & share drafts.

---

---

Thanks for reading! I’m aiming to send two more posts this month: one is a review of the 2024 edition of the Best American Essays series, the other is a long-overdue account of my 10 days with the Apple Visio Pro. Until then, say hey in the comments or in the chat (let’s talk wikis).

---

Footnotes:

These are ideas that were either (a) originally in the essay but edited out, or (b) triggered in the flow of writing—they seemed too tangential to include, but interesting enough to want to write out after. This is something like a graveyard of thoughts that didn’t make the cut.

1. Around the holidays, I noticed some family members using “New Year’s resolutions,” as synonymous with “goals.” I started a debate over the meaning of the word resolution and nobody cared. It comes from resolver which means “to loosen, to release.” I asked AI if there was a term for when a combination of words gets popular and causes us to forget the meaning of the individual words, and it came back with semantic bleaching.

2. I spent too long deciding if I should use the word “sundries” (suhn-dreez), but I’m happy I did, because it helped clarify to me when and how you can use rare words. I looked it up in COCA (Corpus of Contemporary American English) and it’s only used 103 times in a billion. Compare this to the simpler word, “objects,” which is used 39,000 in a billion (around 400x more familiar). I couldn’t find a good word to describe the random crap in my drawer, and the definition of sundries fit perfectly: “various items not important enough to be mentioned individually.” I could have been lazy and used a curse-word, “crap,” but sundries is an even worse insult: these things are so irrelevant they don’t even deserve a name. The dilemma of vocabulary is that the best definitions hide in pretentious words. To solve this, a rare word needs to be nestled inside of context clues that makes its meaning so obvious that it almost doesn’t matter what the word is. So when I say, “I emptied whole drawers of forgotten [word],” the lead-up serves as the definition: sundries are the miscellaneous things in life you save but eventually throw out. I still think that no matter how good a context clue is, some people will still get pissed at a rare word, so you have to consider precision vs. readerly ease. On X, they algorithmically punish rare words.

3. Molting is when an insect sheds its skin. It’s a multi-day process where they basically freeze in their shell before they burst out of their head in newly evolved form (like Charmeleon > Charizard). This is a good metaphor for psychological change too, and I wrote an essay two years ago called The 1,000 Day Molt.

4. I think both linear and non-linear thinking have a place, and it’s important to know your mode. Sometimes “linear thinking” is dished as an insult (ie: you’re trapped in a fixed order of thinking and can’t approach topics from different angles). But the essay, by definition (according to me), is an act of linear will—it a singular, unchanging path between start and end. Because of this constraint, the creator has to make hard decisions on what to include and what not to include. Linear art requires editing. The problem with non-linear thinking is that you can evade editing and just connect everything to everything. This is why I tell new writers to avoid note-connecting software and just focus on writing good paragraphs. I think there’s a time and place for non-linear knowledge graphs, and it’s probably best for specialized research projects. Essay Architecture happens to be that. There’s a funny irony that essays might not be the best way to teach essay writing.

5. You’re already starting to see traditional books turn into “apps.” Generally, this means you feed the entire text into a custom GPT and let readers ask it questions. That is book-turned-chatbot. The inverse of this—an app that is a book—is software- first. It has a range of interfaces and functions, but instead of filling it with user-generated content or scraped data, it’s filled with paragraphs of prose, written by the software’s creators. One wraps technology around a book, the other melts the book into technology.

6. Hyperobject: a set of relationships that are too complex for the mind to hold all at once. This causes doubt over its existence, but 1) it can be visualized through maps or data, and 2) it can be understood instance-by-instance. A hyperobject could apply to something massive (like climate change) or something more theoretical (in this case, a framework for essay composition). In our case, you can’t “download” a hyperobject by learning it rationally. You need to practice patterns, one at a time, until they become automatic. Maybe this explains why masters in any field can’t logically explain how they’re doing what they’re doing.

7. Alexander’s Pattern Language (1978) contains 253 chapters and each one is structured like a wiki: they follow the same template, and they’re filled with (non-clickable) hyperlinks. In his introduction, he encourages others to build their own pattern languages outside the field of architecture. In 1994, Ward Cunningham invented the first wiki—called WikiWikiWeb—and cited Alexander as a direct inspiration. It let users upload and link together software patterns. This inspired more specialized wikis in the following years, and finally in 2001, the wiki for everything, Wikipedia.

8. Here’s a quick secret update on my progress with the app. I’ve built, (1) a scalable system to test, measure, and refine the accuracy of my pattern scoring, (2) a very basic backend which lets you sign-in, upload and save drafts, pay, etc., and (3) an interface that breaks your analysis down into a navigable feedback report. There are still endless details and features to build within those, but the next big thing to setup is the wiki itself. I plan to personally and slowly onboard writers over the next few months, before I release a wider beta. More on that soon.

9. Last year I published 83 times (~7x a month), but only sent 18 posts to my main list of subscribers (1.5x a month). I ghost-post 78% of my writing, intentionally. I think there’s an unrecognized value in low-visibility experiments, which I’ll have to write about someday. If you want to follow my Experiments in real-time, you can opt-in via your account settings (note: much of it is locked for paid subscribers). But, my goal for 2025 is to publish 3x a month to the main list, which is twice the volume of last year.