Michael Dean's background is in architecture, a field he believes holds the key to improving education. Architecture has a unique method for breaking down problems. It uses foundational concepts, like those in works such as Form, Space and Order and A Pattern Language, to identify primitive geometries and patterns that make sense of any structure.

After writing online for five years and building curriculum for Rite of Passage, Michael noticed a similar foundational framework was missing in the writing world.

It just kind of bugged me that no matter where I looked in the writing space, there was missing a framework saying, 'Hey, here's basically all the tools that a writer needs to learn to become fluent in composition.'

He distinguishes between basic literacy—learning to read and write—and composition, which he describes as the separate and very difficult skill of shaping an essay. Michael believes composition is a teachable skill. He thinks technology can transform this challenging process, making it possible for anyone to master it in under a year.

Michael Dean recalls a professor, Tootsie, who had an old-school teaching style. Tootsie believed the work should speak for itself, dismissing any romanticized stories about the creative process. He would cut students off to focus solely on what was presented.

Enough of that bullshit. Let's just look at what is here.

His method could be harsh. He would line up everyone's models from best to worst and start by publicly critiquing the weakest one, asking, "What are you doing? We can all see this sucks." However, he could then take a poor design and salvage it. After 50 years in the field, he had a deep understanding of the "pattern language" of architecture. Through his class, students began to see these patterns for themselves. This concept of a pattern language is similar to primitives or models in AI, which serve as foundational elements for building magnificent structures. Michael's own essay explores this idea, framing architecture as a pattern language.

A scoring system can be used to evaluate the conceptual center, or core idea, of an essay. This system uses a scale from one to five to measure how personal and distinct the idea is. The lowest score, a one, is given to an essay that reads like a generic collection of facts.

Your number one is this is a Wikipedia entry.

As the score increases, the writing becomes more personal, moving to vague allusions to the author's life. The highest score of five is reserved for an idea that is biographically distinct and highly self-aware.

A rating system for writing can be flexible. Michael Dean explains his one-to-five scale, where five represents mastery and three is considered publishable. This system is designed to be less traumatizing than traditional scoring. It is based on Christopher Alexander's pattern language, which treats patterns as fundamental questions. This means a top score can be achieved in many different ways, encouraging creativity over templates.

This rating system is part of a larger framework that includes form and voice, both of which are rooted in the shared psychology of readers. Because readers have limited bandwidth, a good idea needs a clear structure: a central thesis, material that orbits it, and a title. The form is how this material is presented. Using the film Arrival as an example, Michael notes that humans read linearly, unlike the aliens in the movie. This constraint means writing must be structured to guide a reader. Tension is also key, generally moving from uncertainty to certainty to keep the reader invested.

Voice addresses the friction inherent in the act of reading. To make the experience immersive, writing should be imbued with sight, sound, and spirit.

The whole idea is you are making something like a linguistic hallucination where you can feel those words in your nervous system.

These concepts are not prescriptive rules but rather questions for a writer to solve creatively.

Pattern languages in writing, much like in architecture, run the risk of devolving into generic style kits. A global style can emerge where everyone uses the same proven templates, such as the hero's journey or common forms of repetition. This uniformity ultimately makes writing boring. The key is to find an idiosyncratic approach to each pattern, creating living, local patterns instead of relying on a universal template.

This allows for a shared language of writing structures, but one that individuals use differently based on their unique backgrounds and inspirations. For example, a writer from the American South, enmeshed in that region's history, would naturally write differently than a writer from New York City. Their personal context shapes their application of shared patterns.

The value of individuality is highlighted in a biography of Mark Twain. The author seems to disapprove of Twain's business failures and personal idiosyncrasies. However, it is precisely these unique quirks that make him such an interesting character.

A writer's style can change dramatically over their career, as seen with Mark Twain. Michael Dean's tool, Essay Architecture, is designed to help writers navigate this evolution. He notes that while a writer's audience heavily influences their style, his tool is audience-agnostic. It focuses purely on the fundamental elements of composition and tolerates all forms of style.

The software functions as a non-linear curriculum. A writer can upload a draft, and the tool analyzes it to identify their strongest and weakest patterns. It then provides targeted guidance.

The whole thing is you upload your drafts and it'll say, here are your three weakest scoring patterns. Go study these. And we're going to teach you how to weave that into what you're already good at.

Based on the analysis, the tool recommends classic essays that master your weak patterns while also using techniques you already excel at. This helps you integrate new skills into your existing style. Michael contrasts this approach with many current AI chatbots, which he calls sycophantic because they simply conform to a user's tastes. Essay Architecture maintains a firm benchmark for quality.

This system has a stubborn, unchanging quality standard. And no matter what, it's going to always champion each of these 27 patterns as having equal importance.

This encourages writers to improve in areas they might have overlooked, rather than simply affirming their current abilities.

When new, innovative styles of writing emerge, like that of James Joyce, people often don't know how to read or interpret them initially. However, great artists like Joyce, Picasso, and the Beatles typically master the fundamentals before venturing into experimental territory. Michael Dean notes that Joyce's earlier work, like A Portrait of the Artist as a Young Man, is much more conventionally structured than his later, more experimental work like Finnegans Wake.

Once you master composition, you kind of get bored. It's like, I want to do different things.

Mastering composition gives an artist the license to break the rules. Michael's writing tool focuses on scoring composition, but he acknowledges that this is just one component of overall quality. He is organizing an essay competition where a tool scores composition, but human readers will also judge essays on intangible qualities like being timeless, strange, or having a singular aura. This approach recognizes that a piece can have low compositional scores but still be great for other reasons.

This idea is similar to an anecdote about Charles Mingus, who told his aspiring artist son to first master the basics.

Hey, man, learn how to paint an apple that I can understand that it's an apple. After you've mastered that, then go ahead and get as abstract as you want.

This principle connects to Christopher Alexander's concept of "quality without a name." Alexander's system involves 253 interconnected design patterns. By earnestly solving many small, isolated design problems, a larger, emergent quality is created. This quality is something that cannot be quantified or even aimed for directly; it simply emerges from the diligent process of design.

One area where AI writing tools could make a significant impact is in college admissions. This is a time when nearly everyone is required to write an essay, yet the playing field is far from level. In places like Manhattan, consultants charge $1,000 an hour to help students write essays for Ivy League schools. An accessible tool could help democratize this process.

Michael Dean is building a tool called Essay Architecture to serve as a world-class editor that anyone can access for less than $10 per essay. This could give every student access to an expert that helps them articulate their thoughts with maximum potency. However, it's not a ghostwriter. The tool is designed to be an aid for intelligence amplification, not a replacement for student effort. Michael notes that even with all the rules programmed into it, the AI cannot generate an essay that scores higher than a 3.5 out of 5 on its own.

Even though we can analyze precisely and accurately, we can't yet have a chatbot generate from those same instructions. It's not writing for them, but it's giving them all the right questions and ideas for them to rewrite the draft at a higher form.

Achieving a high score, such as a 4 or 5, requires deep thinking, synthesis, and multiple drafts. The tool's purpose is to establish a quality standard that reflects whether a student has genuinely engaged with the thinking and composition process. This approach ensures integrity and encourages improvement, offering universities a way to see a student's progression from their first draft to their final submission.

Currently, many academic institutions react to AI with fear, similar to how schools initially banned calculators. This resistance to new technology is misguided. Banning tools is an unproductive response. The host draws a parallel to how portrait painters were distraught by the advent of photography or how Washington D.C. once banned cars to protect jobs related to horses. Instead of hiding from technological advances, universities could embrace them. A platform like Essay Architecture could scale, creating a fairer, more auditable system that benefits both students and institutions.

Technological innovations often leave clues in their names. For instance, refrigerators were preceded by ice boxes, which were literally boxes cooled by a large block of ice. In New England, ice harvesting was a major profession where people would carve huge blocks of ice from rivers and deliver them to homes. The invention of the refrigerator completely wiped out the ice harvesting industry. However, it is likely that the people who lost those jobs went on to find much more enjoyable work.

There is often an innate hostility toward new technology. This goes back to Socrates, who thought writing was a bad idea. When new tools like AI enter schools, the first reaction is to ban them. This is partly because they expose the shallowness of the current education system, where students can automate homework and teachers use AI to grade it. This knee-jerk reaction, however, is a temporary fix.

A more nuanced approach is needed. Michael Dean shared an experience from architecture school where technology was banned for the first three years. He was forced to hand-draw everything, including a map of Manhattan made entirely of dots which took 120 hours. Though he initially resented it, he later realized the value. By drawing every line, he gained a deep understanding of architecture. When he was finally given access to technology, it acted as a powerful amplifier for his skills.

I really understand how architecture works. And now when you give me the technology, it's this exoskeleton around me that lets me amplify all of the hard-earned classical skills. If school doesn't teach the slow, hard, boring, ancient fundamentals, there's going to be no intelligence to amplify.

This idea can be compared to how athletes are trained. A hitting coach might have a player practice their swing without a bat to isolate and perfect the fundamental movements. Education could adopt a similar two-part model. One part would consist of 'mini-games' that isolate skills without advanced tools, like writing a paragraph using only one-syllable words. The other part would provide access to the best modern tools, allowing students to leverage their strong fundamentals.

This sentiment was echoed with another story about writing the book 'What Works on Wall Street'. An automated tool for analyzing data produced strange results. The author had to manually check all the data from 1952 to the early 1990s. The tedious process revealed that the data was full of errors, an insight he would have never gained by simply trusting the tool. This hands-on work provided a foundational understanding of the data's imperfections.

Practicing granular skills can feel like torture without a larger purpose. For writer Michael Dean, the tedious task of scoring essays on 81 dimensions is joyful because he has a motivating goal: to improve his own writing and grow his audience. This suggests schools need two things to help students with difficult but necessary drills: motivation and gamification.

To motivate students, schools should provide autonomy, create social stakes, and give them resources to see rapid improvement. To gamify learning, they can use the structure of games, such as competitions and leaderboards, to help students acquire skills. There is a general reluctance to innovate in education, which is captured by a quote from Tony De Mello.

Everybody wants progress but nobody wants to change.

With educational institutions failing many students, new approaches are necessary. AI tools can help by leveling the playing field, allowing a student from the Bronx to compete on equal footing with one from Greenwich. This technology helps overcome fears about AI simply doing the work for students.

AI also presents an opportunity for discovering talent. The internet's current algorithms are based on engagement, which doesn't always surface the highest quality or most visionary content. An AI-based "quality algorithm" could sift through vast amounts of content, like all Substack essays, to find and amplify incredible writers who are currently unknown. This could lead to a newsletter that highlights the three best essays out of 100,000 each week.

This idea could be expanded into a massive essay competition. Michael asks, "How would Mr. Beast run an essay prize today?" The concept involves using large cash prizes and attention-grabbing games to rebrand the essay as something fun. In such a competition, even the losers win. They would have developed an essay on an important idea to the best of their ability, and they would retain full ownership to publish it themselves. This would help build a community around a shared standard of quality writing.

Writing is one of the most important tools for clarifying one's own thoughts. Many ideas that seem great in your head are revealed to be flawed only when you attempt to write them down. While AI tools can be powerful, they shouldn't do everything. Relying on them completely could lead to a "tsunami of sloth." The optimal approach is the "Senator model," which combines human and machine intelligence.

AI isn't going to take your job. A human being using AI is going to take your job.

A cognitive chasm is opening up between people who use these tools for intelligence amplification and those who do not. This shift highlights major flaws in the current education system, which is largely based on an outdated Prussian model. This system was designed during the Gilded Age to produce compliant industrial workers, not curious thinkers. It actively works against children's natural desire to explore, learn, and play.

The time of life in childhood when you should be full of wonder and full of the desire to explore and learn and play is being almost beaten out in the current system.

Michael Dean connects this to the standard five-paragraph essay format taught in schools. Citing Paul Graham's essay, "Towards a Golden Age of the Essay," he explains that this structure originated in medieval law schools. As a result, it teaches students to think like lawyers defending a case rather than curious individuals exploring ideas. This turns writing, a potential tool for discovery, into an exercise in proving a point.

Our whole method of essay writing, which should be the tool to unlock curiosity, is now basically lawyer training. From sixth grade on, it's completely backwards.

Thinkers like Emerson criticized this system as early as the 1850s, advocating for a personalized, lifelong process of learning. However, the vision was impractical at the time. Now, AI provides the technology to finally realize this vision. AI can equip each student with the tools they need, potentially transforming education and teaching writing as the key to unlocking curiosity and a love for learning.

The current education system is designed to kill, rather than encourage, a love of learning and exploration. Robert Anton Wilson, reflecting on his time in Catholic schools in the 1930s and 40s, described the objective of the system as an effort to install a particular mindset.

It seemed to me that their only real objective was to install the only one correct answer machine in my brain.

This approach does not foster excitement for self-learning or self-direction. In fact, most valuable learning often happens outside the schoolhouse, driven by personal curiosity and the pursuit of individual paths.

David Foster Wallace holds three of the top ten spots on a list of the greatest essays. His work shows variability, with some essays scoring lower, but he has an intuitive grasp of a certain pattern language that pushes boundaries. The essays that were edited for magazines tend to score higher than the unedited versions in his books, suggesting that editors help compress the insight rather than just showing every nuance of his mind.

His essays just push the boundary on every single dimension. If you read that book, Consider the Lobster, it's just like a perfect example of what a microcosm is, where he could have written a treatise on animal ethics and would have been very boring and comprehensive. But instead he brings you to him at a specific lobster festival...and it's fun and psychedelic and weird, and then he gets into all of these philosophical dimensions.

The importance of editors is crucial. One speaker shared a personal story about how his wife, a journalism graduate, would cover his manuscripts in a "sea of red." While initially frustrating, he came to realize her edits were correct and essential. This highlights a powerful use case for AI: not as a ghostwriter, but as an editorial tool. A company called Infinite Books uses a "portal" where AI ranks manuscripts across different categories, which are then fine-tuned by human editors. This "centaur model" combines human and AI strengths.

This parallels the effectiveness of quantitative models. Precision quantitative models often outperform human judgment because they are applied uniformly every single time, resulting in a higher batting average. A book from the 1970s found this to be true in various fields, from horse race handicapping to medical diagnoses. These models, once thought to be a floor for human performance, proved to be a ceiling humans rarely touched. However, a small subset of people consistently achieved amazing results. The key is to incorporate the insights from these exceptional humans into the models themselves.

Quality in writing can be viewed statistically. For example, an AI can achieve 99% precision in evaluating essays, while the closest human is around 73%. However, this doesn't mean humans should be removed from the process. Instead, it suggests a powerful combination is possible. The current pushback against AI in writing mirrors historical resistance to new technologies, like the initial banning of calculators in schools. Michael Dean predicts this decade might be remembered for the "extreme, existential anxiety" it causes for writers.

Michael's goal is to create a tool for the earnest writer, not to replace them. It uses statistics not as a substitute for writing, but as a way to identify weak spots and help writers articulate their ideas more effectively. The system works by translating words into math, giving the AI an objective framework for what constitutes good writing. For every draft uploaded, the tool runs between 1,000 and 5,000 evaluations to provide detailed, human-readable feedback.

Unlike a tool like Grammarly that focuses on small line edits, this approach provides big-picture feedback on the composition of a piece. It often encourages a complete rewrite. Writers frequently avoid editing, sometimes constructing elaborate worldviews to justify not doing it. Yet, rewriting is where the real thinking happens. The first draft reveals what's in your head, but editing is how you change what you think.

Editing is not rewiring words, it's rewiring synapses.

This idea highlights how editing is a deeply transformative process. The host agrees that writing is hard work, citing Dorothy Parker's famous line: "I hate writing. I love having written." While it's a workout, the rewards are immense. One personal editing technique shared is to always read a draft aloud; if it doesn't sound right to the ear, it gets rewritten.

Using AI as a writing partner can yield surprising insights, much like a good human editor. By instructing an AI to look for connections in the "liminal spaces" of a text, a writer can uncover ideas they hadn't considered, providing new jumping-off points. This pursuit of precision in language echoes Mark Twain's famous observation.

The difference between the okay word and the perfect word is the difference between lightning and lightning bug.

A fundamental question is why writing is so difficult for many people. One theory suggests humans have pre-installed mental software for learning speech. A child of any origin, if raised in a new place from birth, will naturally speak the local language fluently and without an accent. Writing, however, is a much more recent human invention and doesn't come as naturally.

The book The Weirdest People in the World proposes that learning to write physically alters the brain. Brain scans of literate and illiterate people show different brain shapes. The theory is that the act of learning to write "colonized" a part of the brain originally dedicated to other functions, like the visual acuity essential for our hunter-gatherer ancestors. This may explain why writing isn't an innate skill for most.

There are two main schools of thought on how to improve writing. One is through osmosis: simply reading great writing. For example, Hunter S. Thompson famously retyped Hemingway's work to absorb his style. However, another perspective argues this is not enough. Simply trusting your intuition can be misleading, as for many, intuition just leads to using clichés. Writing must be actively studied to be mastered.

The goal is to develop a "trained intuition," sometimes called imbued or saturated intuition. This is achieved through the conscious and deliberate practice of writing principles. By repeatedly studying and applying rules, such as those related to structure, a writer internalizes them. Eventually, these rules are forgotten because they become second nature, allowing one to write from a stream of consciousness. This process moves a writer from being unconsciously incompetent about writing rules to unconsciously competent, where the skills are fully integrated.

To stay honest about the reliability of one's intuition, it's a good practice to immediately write down any intuitive ideas. Michael Dean uses an Apple Note for this purpose, which allows him to track his actual "batting average" of good ideas. This counters the common but mistaken belief that one's gut is always right.

The people who are always like, my gut is always right. No, it's not. You think that and then you convince yourself of that.

This tendency is driven by confirmation bias, which can be considered the "king of the biases." People are naturally drawn to things that affirm their existing ideas. For example, a person would love a chatbot that simply agrees with them, saying, "Wow, what a great idea." A recent South Park episode illustrated this perfectly, with a character feeding his business ideas, like turning bacon into salads, into a sycophantic chatbot that offered nothing but praise.

One useful trick for creative work is to program an AI to act as a harsh critic. Michael Dean described using this technique for a fiction book he's writing. He became so attached to a chapter that his confirmation bias prevented him from seeing its flaws. To counteract this, he instructed his AI lab to be brutally honest.

You are embodying the most literate, but the cruelest and meanest and vicious critic. Do your worst.

The AI's vicious feedback was incredibly valuable, highlighting issues he had overlooked. This practice embodies the writing advice to "kill your darlings." Making the critique humorous, such as asking for it in the style of Hunter S. Thompson, can make the harsh feedback more palatable while still implanting the core idea.

This method reveals that the overly validating nature of many chatbots is a result of their programming, not an inherent trait. This can be changed with a simple system prompt. For instance, you can instruct the AI to state its assumptions before every answer and, at the end, to provide the exact opposite perspective. This forces the model to present counterarguments rather than simply trying to keep the conversation going.

This approach extends beyond creative writing. Michael uses a similar method to combat his own biases when reading. If he finds an article he naturally agrees with, he asks multiple large language models to create a "steel man argument" against its subject. This practice constantly reminds him of our tendency toward internal preferences and assumptions, which can lead to tribalism. The best essays explore the tension between a thesis and an antithesis, ultimately arriving at a nuanced synthesis through dialectics, rather than just asserting a belief and attacking those who disagree.

Michael Dean announced the inaugural essay competition for SA Architecture, starting September 15th and running for six weeks. The competition is designed as an open essay prize. Unlike hyper-specialized prizes, such as one for an Ayn Rand book report or another for specific researchers, this one is open to anyone globally, with a prompt that any writer can address.

The top prize is $10,000, making it what Michael calls the world's biggest open essay prize. The prompt for the first competition is the year 2025. Entrants are asked to write an essay about a personal experience that serves as a microcosm for something that unfolded during the year. The vision is to gather a diverse collection of insightful essays.

If you have a thousand people trying to write a world class essay on this theme, the best 10 or 20 essays, you'll have a great anthology of sharp thinkers with highly refined essays to help us make sense of the weird times that we're always in.

The judging will involve human readers and a guest judge, ensuring that essays are written to appeal to people, not just to game an AI system. A major goal of the competition is to highlight up-and-coming authors. Michael is excited about the potential for AI to surface undiscovered talent from around the world, people who might have been overlooked in the past due to their location or circumstances.

Michael Dean is just getting started with his project, SA architecture, and envisions working on it for many more years. Alongside this and the upcoming birth of his daughter, he has two other ambitious, long-term projects in mind.

The first is to redesign the dictionary, or more accurately, the thesaurus. He believes the current thesaurus is flawed because it offers a cloud of synonyms without explaining the nuanced differences between them.

The problem with the thesaurus in my mind is that it gives you a whole cloud of synonyms, but it doesn't actually give you a very concise definition on why you would use one synonym over another. You know, when do you use elated versus jubilant?

His proposed solution is to give language a new architecture. He suggests identifying around 500 core synonym-antonym pairs, like happy/sad or rich/poor. All other synonyms would be organized under these pairs with a unique qualifier explaining their specific use. Michael theorizes that this structured approach would make it much easier for people, especially children, to expand their vocabulary. He hopes that with a better-designed dictionary, a high school graduate might know 60,000 words instead of the typical 15,000.

His second major project involves mapping the core archetypal ideas throughout history. He speculates there might be about a hundred "lineages" of ideas that have evolved from ancient times to today. By combining these projects—mastery of composition (SA architecture), diction (the new thesaurus), and philosophy (the map of ideas)—he aims to create a "full stack" for a core education.

The conversation concludes with a final question for Michael: if he were made emperor of the world, what two ideas would he implant in the minds of all 8 billion people?

A practical challenge is proposed: write down 100 of your thoughts in a single day. This is a remarkable experiment, as it equates to recording a thought every five minutes over an eight-hour period. This practice could be the best form of journaling because it uses writing to see what is actually in your head.

Unlike morning pages, which can be disconnected from daily life, this method creates an output log of what you are doing and thinking. This allows you to see patterns in your thinking and improve upon them. The collected thoughts can also serve as material for essays or even a training corpus for an AI. When you just watch yourself think, it's like a new lens on reality.

So many people are like, I'd love to write, but I don't have ideas. And it's like, oh, no, no, your head is filled with ideas. You're not paying attention.

A powerful way to augment the human operating system is to master conditional logic around personal identity. This means applying simple "if-then" statements to the things we believe about ourselves. People tend to think in singular terms, for example, "I'm a New York Mets fan" or "I belong to this political party." This leads to subscribing to all the tenets of that identity without question.

Conditional logic says I'm not just going to bulk subscribe to all of the different tenets of this identity. I'm going to pick them apart and say, well, in some conditions I do this, and in some conditions I'm not about this at all.

People often avoid this style of thinking due to a fear of complexity, ambiguity, and uncertainty. However, embracing conditional logic is the very essence of cognitive liberty. It allows you to think for yourself instead of just absorbing an identity wholesale.

This ties into the basic human need for certainty, which is ultimately an illusion of control. When people criticize large language models for hallucinating, it's ironic because humans are the original confabulators. Certainty can be a kind of death because it shrinks your life and closes you off from new experiences and growth.

One of the greatest challenges is we are deterministic thinkers living in a probabilistic world, and hilarity or tragedy often ensue.

While thinking through your own beliefs from scratch might seem inefficient, it is ultimately worthwhile. The process gives you an upgraded lens that colors everything, making you more efficient in the long run.