michael-dean-k/

"Vibe coding for wordcels" (s/o to David Kiferbaum). Two thoughts here:

1. In the last 30 years, there’s been low overlap between people who could write great prose and people who could write great code. If code becomes easier, then what might great writers build? How might literature push the frontier of something as simple as a retro arcade game?

2. When I go on X, all I see are like “look, I built an entire CITY in one prompt,” and while yes that’s technically amazing, it’s completely different from full stack development & security & all the things you need to know if you want to build an app that can scale to thousands of users. Feels like we’re entering the “software as meme” era, but software architecture is really where the value is. There’s probably a lot of value in teaching writers how to ship software to their audience.

Building a political party into the root of your identity is a terrible idea that will give you serious tunnel vision.

In my first 100 days post-WOP, I wrote every day and basically published nothing.

What helped me snap out of that rut was to take on a mindset of "single session publishing." This means I sit down for a writing session and don't stop until it goes on my site (sessions were 2-4 hours). They were effectively first drafts, and sometimes I'd come back to revise them. I didn't "blast" them to my email list, but DM'd them to specific people to start a conversation. There was something about having a body of work on my website that took the pressure off (eventually) sending essays to my larger list.

I'm 100% a perfectionist, and think it's worth having a rigorous, increasing quality standard, but it's important to also have places with much lower standards to serve as your release valve.

Also, publishing under my legal/professional identity came with all sorts of psychological blocks that were completely invisible to me until I started publishing under a pseudonym. Every situation is different, but I wonder if shifting to semi-pseudonymous Substack would help take the pressure off everyeone.

I’m sort of inspired now to create a section on my own Substack called “Drafts.” You can customize that so new subscribers by default aren’t on that list. I currently have an “Experiments” section on my site, but “Drafts” feels a lot more self-explanatory. (Some of these drafts will eventually reach “peak” form and get reposted in the Essay section, but that won’t happen with most of them).

11:30 AM—I used Claude to recreate a video game that my dad, brother and I made in 1995. I changed the logic so that every time you get a diamond, a new monster appears. This makes it get progressively harder within the level.

Inside the Monster Human

Prompt:

“Below are rules for a game. Read them and ask me any questions to clarify implementation before you build it: I'd like to create a game called "Inside the Monster Human." * Each level is a 30x30 grid. Cells can either be white or black (if it's black, characters cannot move onto them). Let's make the level something like a star, with a white center with 5 wings you can go down.* There are 10 diamonds randomly placed around the level. * The character (a hand-drawn red smiley face with a white mouth and eyes) uses the arrow keys to move around the grid and collect diamonds. When the character enters the cell of a diamond, the diamond disappears and displays a sound. *Each level contains monsters that are randomly placed. Every time the character moves one cell, the monsters moves in one cell in random direction. If they hit the character, it plays an "Oh no!" sound, displays a game over screen, and restarts the game. The map starts with 3 monsters. But every time a character gets a diamond, it randomly spawns 1 new diamond and 1 new monster. Display the score.”

11:29 PM—I probably got this idea from Ready Player One, but there’s something interesting about baking my consciousness into a video game. There are two interesting mechanics:

1. The highest scorer unlocks a personal piece of writing. You effectively have to beat me at my own game, making it pretty hard. The allure of the essay is more important than the essay itself (but of course, it needs to be awesome / revelatory).
2. As you play the game, each “coin” could unlock one of thousands of memories. Currently have a list of 300+ memories, sorted by year. The idea is the player gets these shards and slowly pieces together a sense of my life.

If I had this game running, it would be incentive to (a) remember all my memories from the past, and (b) to keep uploading them each year. By doing this, I not only have a non-linear game-based memoir, but my cornerstone memories are now machine-legible and can guide some weird AI experiments.

Bought memorydungeon.com. This isn't really a priority, but I imagine if I slowly jot memories all year, I’ll be able to throw this together one weekend in December.

02:33 PM, ONE DAY LATER—I think I get why footnotes don’t make sense in memoir / personal writing. It can feel like dictionary-brain. I also wonder if something is only footnotes (w/o a main narrative) is it even a footnote? I sense there’s something liberating about recording memories as personal paragraphs that don’t have to fit into a story.

So imagine this game is like a turn-based Pacman (but the ghosts are stupid and replicate out of control) — and so the main character is collecting diamonds, and each one has a year on them, so it’s like “Hm, do I go for 1997, or 2009, or 2021?” And then when you get it, it shows a self-contained paragraph from that year.

I think there’s a word for this: “the fragmented essay.” Honestly, they’re my least favorite sub-genre, but I think they weirdly make sense when you put it inside of Pacman.

20th century college/career:

• linear
• singular focus
• credential based
• one income source
• work in private
• up front learning
• declare major up front
• exams is proof of success
• hierarchy of teacher/student
• go into debt to attend

21st century college/career:

• non-linear
• polymathic
• portfolio based
• multiple income streams
• work in public
• continuous learning
• emergent major
• market is proof of success
• everyone teaches and learns
• get paid to contribute to school

There's a challenge in writing for two audiences at once. How can something make sense to someone with no technical knowledge without boring an expert in the field? You have to innovate in how you communicate. Metaphors serve a dual role: they give the novice a bridge, and they give the expert a new lens to the familiar.

Earlier this morning, I was visualizing myself on my drum set 10+ years ago. I could visualize the kit, and had a real-time connection between hitting a drum and hearing the sound. It was real-time enough that I could compose a pretty sophisticated beat on it. I was surprised that I could do this, and it probably helped that I was barely awake. I could never do this on guitar; drums are much simpler, and I have much more experience with them.

Even before I learned drum set, I mastered Snare Drum, which almost seems pointless, but I guess it’s a hyper-isolated way to learn rhythm. I forgot that I did level-6 NYSMA as a kid (for snare drum) and got a perfect score. I often say that “I can’t read music,” but I forget that I was able to sight read advanced snare drum rhythms. Before I started learning drum set (in 10th grade), I had been practicing snare drum for 8 years.

Now I’m wondering what effects that had on me and the way I think/see. How is someone different if they’ve spent that much time on precise rhythms? This is resonating with me because it falls into the category of how slow, pointless, analog things can be formative.

Idea for Timehole (the non-linear novel I'll never write): So there’s a classical pianist in the 2040s who refuses to go into a “time hole” to write music. Basically, there is a headband you can wear that shoots acoustical signals to trigger lucid-dream-like simulations (inspired by Morpheus), where you can experience an hour in a minute (extreme time dilation). So all these musicians go into time holes for hundreds of simulated years and became savant-like, routinely. This classical pianist, the protagonist, is the last analog musician. This is all background context.

Suddenly, there is a UAP phenomenon, something like a first-contact experience, but they send us angelic/bizarre music, and the human race interprets them as wanting a response. The UN General (or whoever is in charge) offers the protagonist the ability to write the species-defining song, if he’s willing to go into a timehole. The idea is, even though he’s not as good as the “timehole composers,” he is the last classically trained maestro, and so his brain is structured differently, and if he were to go into a time hole, he would produce something more profound than the average person who goes in.

So he caves and he does it, and it’s this weird music writing odyssey, and somehow he ascends to a level of music-making where he can be bend reality/time. The ultimate twist is that the UAP phenomenon wasn’t an alien species, it was the musician himself breaking through reality from his timehole from the future.

In the spirit of experimenting, I live-streamed myself drafting on Substack Live yesterday. 50 minutes, 2,700 words, 38 viewers. It was interesting to see how that context shaped the writing process. Since I knew people were reading live, it was like I tapped into this unbreakable flow state in order to weave thought together at a rate of 54 wpm. I thought the whole thing was 10-15 minutes and almost missed my train. It’s similar to typewriter brain, but perhaps the opposite in terms of speed. Maybe it’s a mode that’s actually helpful at certain stages of drafting. TBD.

I'm at a D-tier pizzeria and they’re showing CGI zombies on TV. I think part of entering the public sphere is learning to adapt to whatever is thrown at you. It’s just funny because so many restaurants take the idea of atmosphere so seriously, and now I’m here eating pizza as I watch people eat people.

Saw an app on X where you "set a conversational goal" in a text thread. After someone texts you, it then runs 1,000 possible conversations, and gives you 3-4 options to pick from that are most likely to bring you to your goal. It turns the spontaneity of conversation into multiple choice, optimized over a superficial, narrow aim. In the demo video, the goal was to “Rizze Annie up at Columbia’s hackathon.”

This is telos maxxing, but we should respect the process. The slow act of creation actually changes your conception of the problem in real-time, so you end up in domains you’d never imagined if you only instantly judged near-final outputs from your prompts.

• Socrates = the weight of doubt, the urgency of ethical clarity (2D);
• Plato = analytical (but static) models of wisdom (3D);
• Aristotle = a dynamic system towards wisdom (4D).

An acronym to remember this is "SPA" (they all wore togas to the spa).

When you practice the patterns of essay writing, you don’t just become a better writer, but a better thinker. Rhythm enables you to compress ideas on command. Repetition helps you poetically elaborate. Rhyme helps you search a synonym space based on sound. Imagery enhances your associational and descriptive abilities.

What if the root of an OS was a place to just stream prose without too much concern over where it goes? If you only let yourself write in the context of logs, essay draft, memos, or typewriter one pagers, it’s constraining. I’m experimenting with a daily gDocs page that I use just to think out loud. It should be mostly a stream of consciousness—a way just to shift through ideas in my head. I realized that all morning I was stuck inside of a thought loop. Even though I got some stuff done, I didn’t have my head tuned right. It seems like a simple walk & location change fixed that, and now I’m moving forward

Why do LLMs suck at writing? I’d like to do some research and get semi-technical on why it sucks, but then also explain how some technical advances could make it quite good.

Here’s how I’d summarize the 2 core limitations:

• It doesn’t have the ability to edit or delete. It thinks really hard for a second, and then just rams forward without thinking. It takes every word in your prompt, shoots that into the black-box cube of the LLM, and builds something like a snapshot of your problem at hand, and then just spits out words. New “reasoning models,” extend this thinking process. o1 will think up to a minute, and Deep Research for 10s of minutes. This still isn’t how humans write. Ideally, after each sentence, it’s able to re-read what it wrote, and then redefine the whole problem, and perhaps, start over.

• LLMs don’t have an embedded theory on what makes good writing. It uses probability to figure out the most likely next word, making it something like the Internet’s auto-complete. We’re only bottlenecked here by our ability to convert taste into something that is machine readable.

Creepy AI generated video of humans being enslaved by robots. I imagine anti-technology sentiment will get more visceral in the coming years. AND WORSE... it will give future AIs a visceral sense of our fears. It will see this and know we don't trust it.

More Super Bowl commercial absurdity (from my parent's house): Eugene Levy’s flying eyebrows; UFOs want Doritos, William DeFoe is a pickleball hustler, Seal is a seal, a city of sloths, and a fleshy cowboy head.

Talks about BCIs (brain computer interfaces) naturally skeeve us out because we imagine a doctor drilling a hole in our skull and inserting a silicon chip. What if it weren’t silicon?

I was reading about how AGI would design new kinds of hardware to evolve itself into ASI, and not only does it find the limits of silicon (via. cubic liquid-cooled chips), but it pioneers other forms of computing. Not just quantum computing, but photon-based motherboards and “neuro-morphic chips” (wetware). I need to wrap my head around this more, but imagine biological neurons that are able to store, send, and receive information, reliably and efficiently.

In the future, you might be able to have a BCI without having any metal in your head. It could be something like a small piece of tissue (which interfaces with your own brain matter), and maybe it includes a single qubit chip, which serves as a “quantum wifi” connection to a remote super-intelligence, giving you instant access to alien processing.

I have no expertise here, so this is basically all a hallucination, but I just imagine that maybe BCIs won’t involve that much-feared surgery. Maybe you take a pill that enables a neuro-morphic chip to situate itself in your brain, and the only side effect is you piss silver once (with no side effects).

The point of this thought experiment is to remove the installation process from your calculation on if you’d get a BCI. It changes the question: Would you use technology to permanently alter your consciousness in an extreme way? Would you install an on/off switch that gives you on-demand access to the infinity-brain effect of an LSD trip?

I think we're already nearing a point of diminishing returns with LLMs (ie: we’ve made a huge lizard brain, but what if we made that lizard brain 10,000x bigger?). Hallucinations might be unfixable with pure LLM systems, but neuro-symbolic AI is probably going to fix this. This means putting rules-based logic at the foundation layer (beyond just probability). My guess is that 2025-2028 is when companies realize that neuro-symbolic AI brings more gains than mono-dimensional scaling. 2029-2032 could be when it’s fully deployed in mainstream products. It might also be the thing that leads to AGI/ASI. (Or maybe I’m hallucinating based on the hallucinations in those reports.)

Maybe it makes sense to look at current LLMs like pre-mammalian enslaved synthetic beings. They don’t have all the regions developed yet and their zoomaster has trained them to be sycophantic house pets. I think I agree with everyone to the degree that, even if you make the cage and reptile 100x bigger, a reptile is a reptile. But if you look beyond the products and incentives, and speculate on the engineering possibilities, I think it’s feasible that we make other kinds of digital brains. I don’t think it stops at making synthetic human brains, with all its properties (intelligence, creativity, emotion, agency, etc.); I think our science leads us to beings with spheres and scales of mind that exceed our imagination.

Our society might be going through a kind of cognitive Copernican shift: maybe the human mind isn’t actually the apex of consciousness. It’s easy to point out current features and say, “Look! It is so limited compared to our superior naturally-evolved human minds.” But soon enough, we might realize it’s us that have the limitations to grow, expand, and evolve. This is disturbing, and it’s possible to harbor both awe at the possibility and anxiety over the socio-political form it might take.

My read on the situation is that our engineer-class is leading us towards the grandest possible climax of philosophical, moral, ethical, and religious problems, ones that are out of their depth to handle. And so instead of doubting engineering, there’s an urgent need for meme-poets to reach into the future and past to forge new myths that touch the hearts of the midwits in charge.

I think Trump's assassination attempt might be considered The Fourth Turning in the Strauss-Howe generational theory model. The past turnings have been wars (the Revolutionary War, the Civil War, World War II), but it’s less about a defining conflict, and more about a changing structure of the world. It definitely feels like we’re entering a new world order in more ways than one (America, AI, crypto), but we’ll only know in hindsight. In last cycle it was about American hegemony, in this one I think it will be something like an ASI hegemony. Still points to 2045 as the spiritual crisis.

Some points to consider for the age of AI:

• Have a bold vision for your life and art.
• Don’t compete with commodities or status.
• Make yourself legible to machines.
• Turn your frameworks into code.
• Start a vlog.
• Go off-script.
• Take leisure seriously.
• Meditate multiple times per day.
• Join the Boy Scouts.
• Stop watching porn.
• Aim to be sublimely well-rounded.
• Embrace for the weirdest possible future.
• Be humble as you lose your edge.
• Accept death in this hour.

I'm using Deep Research to generate reports that extrapolate the trends from a specific field 20 year into the future, chunked in 4-year increments. These reports are filled with "die-progress unit" events (a term from Tim Urban). They bring awe and anxiety. My plan is to generate and synthesize 20-50 of these reports and then write some Tim Urban-like piece that shows the future of our culture in nauseatingly high detail. The goal is to experience the DPU in literature so that you’re more prepared for when it actually happens. My understanding of what writing could be like in the 2030s and 2040s is so different now, that I won’t be bothered by an upgrade in 2026.

These reports are more like LSD then cocaine, after you try it, you want less of it. That was a joke, but actually I just had a “bad trip” on my super-intelligence report, and need to put it down and continue tomorrow. It’s not like I run a lot of prompts. 1 per day, 2 max.

( ... Future Cube: idea for an interface that uses LLMs to interpolate details between futuristic predictions. There are 3 axis: time, groups, streams (technological, economic, political, social). You can place 3 events (quantum computing reaches X by 2028, the US federal government dissolves, and unemployment reaches 20%), and then it can make predictions in other fields (crypto, relationships, video games, etc.) based on those events ... )

The only personal brands that survive in the age of AI are the ones that are unoptimized.

An example of how making a sentence more passive gives it power:

• "I found a cockroach on my front door when I showed up to my lease signing."
• "When I showed up to my lease signing, I found a cockroach on my front door."
• "When I showed up to my lease signing, on my front door I found a cockroach."

As the sentence gets more passive, and the surprising term gets shifted to the end of the sentence, it creates more of a reward for finishing the sentence. Passive constructions help you strategically withold information.

Another variation:

• "When I showed up to my lease signing, I found a surprise on my front door: a cockroach."