michael-dean-k/

The neat thing about being a "writer-coder" is that you can shape your own writing environment as you write. This is being written inside of Margin Muse. Just after cobbling together that opening sentence, I've decided something: the text is too small! (PROMPT: "can we add another slider for text size?") An easy way to get over the blank-page effect is to see lines pile downward as fast as possible. This happens faster if the text is bigger. By the time I've gotten to the end of this dreadful, placeholderly paragraph, the feature is implemented.

Alright, that increased this from 5 lines to 7 lines, but I want the font to be untastefully big. (PROMPT: "why stop at size 22? what if we went up to 48?"). [REFRESH]. Now that's big! Size 36 feels like a nice starter font; it gives me the illusion that I'm writing into a vertical mobile screen. I am a sucker for typeface and design. It feels great writing in here, even if the content is drivel. Sometimes you just need to get started before you try to make it great.

There are hundreds or thousands of little decisions that go into a writing environment. Something like Google Docs was not designed for a creative process—it's an extension of being a secretary, the ultimate tool for collaborative, real-time meeting minutes. If you make 1,000 decisions geared towards the creative process, it might have a meaningful impact on it.

Is this not the best kind of software to build? It's not productivity tooling, not a scheme, not intended for anyone else. I'm building technology for the thing I love most: writing. By taking this seriously, it will—hopefully—be inspiring enough for others use as their default. Even if I'm the only user, all that matters is that my process compounds.

As I write all this, I'm mostly focused on making sentences, but there's a subtle awareness that the container I'm writing in is gooey and bendable. And so if an idea emerges, I shoot it into Cursor and continue prosing as the feature brews. Features pour in at the same rate it takes me to finish paragraphs.

Would other writers want to write within my own idiosyncratic editor? Of course not! But every time I add a feature, I need to remember that the thing I need is just one expression of feature set. Essay-to-essay, what I want might change. Within an essay, it might change too. Already, I find myself wanting smaller font, now that I'm six paragraphs in. And yet, I'm too lazy to change the font! I'm reclined in my chair, with my legs on my desk (terrible posture), and I don't want to reach for my mouse. It would break my flow of prose...

Imagine if there were key commands in here? All I'd have to do is press the slash key twice ("//") to open up a little pill that I can type commands into: "//font-size:16//". Maybe the entire interface, and all of the margin muses, are controllable through key commands! I like the idea of calling a prompt within here without breaking flow. I could write "//prompt: Where am I going with this?//" and then AI would think for a bit, process that as I continue to write, and then leave a comment within the document.

Imagine if I develop a whole new syntax for prose writing? I mean, in the end, none of it would likely be visible to the end reader. It's more like a process language for "draft zeroes"—a meta document where you write to conceive what it is you actually want to write. There could be unique punctuation and syntax to help call upon different AIs without breaking your stream of prose. In the end, I suppose this fuses prose & prompting together into a single field, and through the appropriate syntax, you can control how much of that questioning you expose to the reader.

I've arrived at what I want to say: is prompting not the ultimate skill of the essayist? If an article is centered around an event, and a profile around a person, is an essay not centered around a question? An inquiry? Are prompts not questions and inquiries? This taps into what I think AI is most useful for: asking questions. This sounds trite, because you cannot interact with an AI without asking questions. I'm sure everyone claims to be a great prompter. But the nature of writing, of design, of art, is basically about asking a chain of great questions. Of course, Montaigne did not have a mechanical muse answer questions for him; but a striking question prefigures the answer. It takes many drafts to get to the real question we want to answer, and once we know it, it's as if the essay writes itself.

The ultimate essay composition tool is one that turns you into a lightning rod for inquiry.

Moment of truth: do my slash commands work? ("//size:13/") They do! Unreal. Signing off with legs up and the full essay in sight.

The Internet flattened the world into a digital puzzle—a kind of intelligence test to see who can combine the correct symbols in the correct corner of a social swarm; plucking a nerve to some various end, bringing fortune, fame, or cringe. Your conception of the game shapes how you play it. That’s the real limiting framework: your mental model. You can’t understand emerging social systems without flailing around to update those models. It is very easy to stay sidelined with some excuse that prevents you from getting embarrassed.

This same paradigm applies to AI. It’s limiting to anthropomorphize it; instead, think of it as a linguistic puzzle. It has ingested trillions of words and arranged meaning into a multi-dimensional geometric map, a massive mystery that no one yet has fully uncovered. It’s your job to craft and aim language into this hall of mirrors. We can only do this blindly. But nonetheless, it brings back a result, and how we respond to it is revealing. There all sorts of distortions—cliche prose, lying, hallucinations—and if you’re metaphor for AI is an entity, then you’ll naturally get annoyed or offended and stop playing the game. It’s better to think of your result not as a failure of the LLM, but a failure in your probe. You’ve asked a wrong or vague question. By sending multiple probes, you start to build a fuzzy map of the hyperobject you’re prompting into. Since it gives back feasible answers back so instantly, we expect it to be right for immediately, with little exploration, but that lack of patience is where the frustration comes. It’s a more a matter of rotation, to approach something that matters to you from many angles until you crack it.

This essay tapped into a striking definition of timelessness. He doesn’t get there until halfway through though, and I found myself disagreeing with—or at least, questioning—a lot of his earlier points (I’ll come back to this). The main point is distilled into this: the best essays are “ineffective” because they reveal the timeless problems that each generation fails to synthesize. Timeless essays speak to the common foils in the human operating system: the blindspots of parents, the lies of institutions, the avoidance of mortality, the ineffability of relationships, the mundanities that are never captured in enough detail. These are different than “discovery” essays, like Darwin’s Origin of Species. The holy grail of an essay is surprise, and a timeless essay is not just sueprising for one generation, it’s surprising for every generation. And so timelessness, then, is a type of “breadth of applicability.”

PG also ventures into a familiar territory of “essay as a mode of thinking.” Where as in the past he used “the river” as his metaphor (2004), this time it’s a tree. You start from an origin, and then you explore many different branches in search of generality x novelty. What is a good starting question? He says a good one is “outrageous, counterintuitive, overambitious, and heterodox.” It doesn’t have to be a complete thesis, but some puzzling gap, and importantly something you care about. You won’t be able to stretch an origin question into cascading insight unless you have a unique angle into it. The origin doesn’t matter too much though, because it’s a recursive process, and you can eventually get to the best question in “a few hops.” I love how he emphasizes that you need to write to explore branches of a tree, and there are many dead ends; you realize how you are mistaken, incomplete, and inelegant (you go from vague to bad). Don’t get discouraged by these; finding your false assumptions is possibly the only way to really begin.

Despite loving his whole exploration of “mode,” I don’t think that means you have to neglect essay as “genre”; he says form/style don’t matter in “the best essay,” and I disagree, obviously. He has Darwin as the pinnacle example of an essay, and I’m really challenged by that (I definitely have to read it now). Is that an essay or a scientific paper, just captured in shortform non-fiction? He seems to imply that the essay is at its best a vehicle for discovery, as a mechanism to bring forth surprising, important, and useful ideas. From the creator of “make things people want,” this isn’t surprising. Even though a new theory of evolution had broad implications for society, I assume the paper itself is technical, intended for a scientific niche audience, which in my mind, makes it more like a scientific paper than an essay. An essay is something that is universal/general enough for the average person to read. An essay, I think, functions like an information transfer system between specialized facets of society; it’s about making your specific niche legible to all the other niches, and I don’t think that was the specific goal of Darwin's writing (even though it was inevitably understood by everyone, it wasn’t through the writing, but from the effects of the writing).

(Added: Another note on Graham’s notion of best as timelessness: he says that timeless esasys are the perennial insights that each generation can’t absob. This implies that the insight is never enough: even if you know something, there is often a lack of wisdom in applying it to your own circumstance. And so really, these unteachable lessons are ones that can only be obtained through personal experience. Does this point to the fact that all essays need to be personal? Maybe bland insights can’t be digested by a reader, but if they are integrated to vivid personal experience, experience vicariously, then might this actually be the best medium to transfer wisdom?)

New idea for an "Architecture of Everything" newsletter series, where I present my favorite problems (Feynman style) as a daily email, and prompt you to share your thoughts. It’s a way to crowd source answers to the questions I care about most:

The architecture of death, consciousness, space, history, writing, government, video games, technology, music, religion, news, social media, mind, software.

Some of the latest:

• Is the act of dying a time-dilation odyssey, where 3 minutes feels like a 30,000 years afterlife?
• If I capture my consciousness in 10 million words of logs and essays, could that enable an AI textual replica to evolve and engage with the world 500 years beyond my death?
• Can we resurrect Christianity by putting psychedelics back in the holy wine?
• What will be the psychic and cultural effects when our species understands "spatial relativity," that the Big Bang emerged from a black hole in a parent universe?
• If cycles emerge form order, can we predict the future based on historical patterns?
• If there is a universal language of patterns beneath all essays, can we build an AI to give world-class feedback and make it more approachable to become a master at writing?
• Were psilocybin mushrooms a linguistic mutagen that accelerated the evolution of human consciousness?
• Could we restructure the thesaurus to 3x the vocabulary of the average person?
• Might blockchain-based governance be the civic breakthrough required for a species not to exterminate itself?
• What text-based video game formats are undiscovered?
• What are the societal effects when AR/VR is mature enough to simulate teleportation, and how can we design the metaverse to promote human flourishing?
• How can popular music change the values system of a culture?
• What systems of attention, language, and action lead to a transcendent consciousness?
• Can we use AI to filter through millions of comments on breaking news, structuring each event as a range of unique interpretations?
• How might a new social media algorithm trigger a Renaissance in connection, self-expression, and agency?

(Image getting one of these per day with 300-500 words of context).