Tinkered Thinking

We had disturbed something in the universe. Humanity had spread to 11 billion galaxies. I use the old language now in these final moments as an atavistic tribute. The current modes of communication rapidly dissolving as the network of humanity sheds.

We found a star of unknowable quality. In intergalactic space, just floating there, dim, almost matte in its luminance. It was a frontier ship that decided to visit it, on the way to another galactic cluster. It's humanity's purview to go out of our way to investigate curiosities. And this one had been glimpsed some years prior. It had been on our list.

Despite our vast and intimate knowledge of matter, manipulations of the quantuminium, this star — if you could call it that — eluded our understanding. It was a bit of a panic, as this sphere became a hole in our unified field theory. A prob was sent, but the moment it touched the surface, it disappeared. And then it began to expand. Very rapidly, engulfing the frontier ship before the systems could even react to the situation. Within minutes it had swallowed several galactic clusters.

All of us could feel the people on that frontier ship vanish, as if you were riding on a crowded subway and closed your eyes and moments later, knew, that somehow, you were now alone. None of their thoughts, their revelations, their subtle slivers of emotions populated throughout the quantuminium. We knew they were gone.

The dim star expanded faster than light could travel. It was only by the slow eating away of the vast fabric of voices on the quantuminium that we even knew something was happening. All that happens on the quantuminium is instant, no matter how vast the eerie distance. And slowly, humanity fell silent.

The giant orb has subsumed so much of the universe that it no longer seems like a growing sphere. So large, it now seems as though it's flipped, and we are inside the last remaining bubble of the universe while it rapidly contracts.

The quantuminum was now quiet, with shock and awe. There was nothing left to think or share for there was no time to react to the closing in of all space and time.

My final thought, as I write it now, seems to be the curious nature of my position in the universe. As the grand spectacle appears to recede into the borders all around me, I realize that I am in the place where all will finally end. The last to exist inside the only bubble of something in an infinity of nothing. Perhaps everyone awaits on the other side. Perhaps I just can't hear them.

The wall of dim matte light appears to slow as it rushes towards me on all sides. I see the the members of the command bridge disappear into the pristine soup. One of them looks at me as their face disappears, the contours of their face retracing until it resolves to the single point where they disappear. It slows faster and faster as it closes in around me now, as I search for the right words, the only words, some way to explain what is occurring. But there is no time to analyze, there is no one left to help me, There is only the contracting void, and my words, these final words. I reach for the last thought, and then it occurs to me.

BENCHMARK THIS! BETTER PROBLEMS & THE WHITE MIRROR PUZZLE

July 14th, 2026

Each time a new AI model comes out, I have my own set of tests. One is a health puzzle. I have a bunch of hormonal data and life style data and i put it into the new model to get it's diagnosis and plan to fix the issue with a mechanistic explanation.

Then I have a novel I've been writing. I've got about about a dozen chapters done. The first chapter has a subtle twist in the last couple sentences. It's subtle but profound for understanding what's going on in the story. I have the new AI model read this incomplete manuscript draft and I judge it on how it judges my manuscript. I see what it picked up on, what it missed, where it has some retarded politically correct "suggestions", what recommendations it has for plot, and if it can understand the larger mission of the book.

These two exercises tell me almost everything I need to know about a model. I don't really care about whether it can solve a math problem or code an app. There is far more information about it's performance when you throw a very noisy, very complex system at it. Biology is such a system - mostly due to our lack of comprehensive knowledge, and a novel, with all its twists and rabbitholes, it's contextual richness. Well, lets just say that the limits of a SOTA model show up quickly and obviously.

There is another test that I like to throw at SOTA models. I give it my latest book WHITE MIRROR,¹ and then I ask it to solve the puzzle.

WHITE MIRROR has within it a puzzle that I designed and embedded into the text. It is not an easy puzzle. It requires thinking across multiple domains and multiple dimensions of each of those domains. It can't say too much about it because, well, naturally, I don't want to take away from anyone who aims to cherish the challenge.

While I'm sure there will come a time and an AI system that will be able to solve my little puzzle in a handful of seconds, I can't deny that a smug and satisfied smile emerges when I see the results of the latest Fable model put to this task. What's the verdict?

Well, when I tried this with Fable, it failed. Spectacularly. The thing had no idea what it was doing, what to look at, what to pay attention to, where to draw connections, what to drill down into. To be completely honest the experience was like that of David Attenborough watching an orangutang bang a saw against a piece of wood, hammer at horizontal nails and get distracted by a banana in the distance.

Don't worry, I'm aware that my puzzle will not remain immune to the possibilities of compute. But that's not the point. It was never the point. I did seek to make something that was fairly robust to AI systems, and given my fairly extensive usage of AI systems I had a bit of an intuition about what would and wouldn't work.

Today, I worked on the code for Tinkered Thinking. It's been on my to-do list for a while and I finally had time this past weekend to lay out some plans. And then of course AI just plowed through the work. Work that I used to do by hand. Tedious work. Mind numbing work. Sure there was a lot in the process of hand-coding that was interesting and challenging. But fiddling with css to get some nice effect to work was never in that category. Now I have better problems. Like, what should I write about? What books should I release? What else can I offer the world through this brand, this. . . experiment.

This is the point of AI, and its under appreciated by the doomers and even the accelerationists. As with all things in the public discourse, any loudly touted position lacks the nuance that would quell such protestations with a more thoughtful contemplation of the topic. Unfortunately most people do a job that requires solving a bunch of mundane and tedious provlems of questionable necessity. And if suddenly relieved of those problems, the concern isn't to get better problems, the concern is where the paycheck will come from.

In Korea for example, the AI-wariness of the states doesn't make any sense. This is because it's apparently quite difficult to terminate positions, and so everyone is welcoming the computer's ability to do a bunch of their work. I'm sure the same would be true in the states. The computer can now do my job, but I still get the paycheck? Sweet.

Unfortunately the USA is a different breed of economics. And this essay about puzzle problems and models isn't the place to solve the disparity between what's going on and what might be needed.

Think about it in terms of your time off. Your weekends are often littered with annoying problems: sign up for car insurance, cancel that subscription, fix the broken drawer in the bureau, clean the bathroom. These are all shitty problems that you don't get paid for. And if you woke up tomorrow and there was suddenly a system capable of taking all these trivial problems off your hands. . . would you object? Or would you instantly think about how you'd be able to spend your time more freely, doing things you actually wanted to do? Do those trivial bullshit problems give meaning to your life? I don't think so. And if you do think so, then I think there's a bowl of psychedelic mushrooms with your name on it, or a tab of acid at the very least. Which is to say: get real. The only reason people attach meaning to mundane jobs is because it's a euphemism for survival by way of a paycheck. The real fear isn't a loss of meaning, it's an insecurity about the possibility of getting another job, a better job, with better problems. This is a very real concern, and for many it's going to be quite bumpy.

The ladder of agency remains: once you've mastered one level, it's a mistake to keep rerunning that level over and over. You need to level up, and dive into something that's beyond your ability. It's the only way your ability expands: with better problems.

And now we have an incredible and dynamic tool that can blast through a lot of those trivial and time consuming problems that used to eat up your hours.

Regardless, you can be sure that each time a new SOTA model drops, I'll be throwing my own puzzle at it:

Benchmark this!

EMBEDDED NOTES

July 13th, 2026

And other updates.

I've always loved footnotes. Endnotes, not so much. They're often a lot . . . dustier. Academic. But footnotes often have a lot of fun in them. In fact, there's been quite a few times when I was puzzled as to why the footnote wasn't in the main text: That was so informative! That was hilarious! Such a good anecdote... Parentheticals feel like clarification. Footnotes seem to be the sweet spot.

There's just a few problems: Footnotes a just a ugly. They immediately draw the eye when the page. What's that? The text is usually obnoxiously small. And when you finally hit that superscript, you teleport to the end of the page, but when you're done? Well damn, where did that little superscript go... oh there it is. Ok, now I can pick up where I left off.

That almost 3 decades of internet has not improved upon this surprises me. So I brought to life a little idea I've had for a long time: Embedded Notes.¹ So what is an embedded note? It's a foot note that exists within the actual body of the text which is expandable and collapsable. You should be able to click on that superscript at the end of the previous sentence and the text will magically split, open and reveal the embedded note. But there's something else that I built into this new embedded note functionality.²

I'm fond of rabbit holes and red herrings. An unfruitful rabbit hole can feel like a red herring, and a red herring can prove to be a productive rabbit hole, if only you give it enough time, and the context changes enough, and you find out that the things you discovered or learned or figured out in the depths of that rabbit hole apply to something else you never expected.

Other updates: There's another book that's done. It's a very short one. Slim and sexy. It's called PAUSE, and it's a book about meditation. a few years before I started Tinkered Thinking, I was pushed by a handful of variables to commit seriously to developing a meditation practice. The experience was incredibly valuable and that commitment was one of the best decisions I've ever made. I'm currently updating the Tinkered Thinking Bookstore. It was my first project and it needs some overhauling to make it easier for me to get new books on it. I might get some other stuff on there too... We'll see.

There's also a meditation app which I have been "working on" forever. Which is to say, it got thrown on the back-burner faster and more often than pig decides to wallow in mud. A bummer, but it's literally in the final couple percent of progress. It's even already been approved for release on the Apple App Store.

I'm also realizing just how rich my life was with a daily writing practice. I've got a dozen chapters of a novel written, but it's not the same to work on a novel. A novel requires long stretches of time. Tinkered Thinking evolved to be a kind of gesture-with-words. A quick sketch. It is a skill all on its own with benefits that are a bit different from slower, more methodical forms of writing. But that's another topic, for another day.

I think Tinkered Thinking needs to continue evolving. I'm not exactly sure what that means other than more essays, more stories, more books. If you have any thoughts. I'm all ears. In the meantime: be careful of the context.

Oh, also. Tinkered Thinking now has dark mode. There's a very faint moon in the bottom right hand corner. Click it.

DATA DIET

May 18th, 2026

Many creatives are disturbed by the abilities of AI. Anyone who finds their lamenting to be childish or an eye-roll in some other way has likely never had the value of their identity threatened on a fundamental level. If you're a father with grown children, imagine finding out that the children are not biologically yours. If you're a mother, imagine losing that child. If you are an athlete, imagine becoming paralyzed. If you support a family with your hard work, imagine getting laid off, and your industry vanishing. If anyone loves you, imagine if they suddenly stopped loving you. These examples are extreme, and it might seem like a stretch to equate it to the loss of identity and artist might feel, but hyperbolic analogies are often necessary to bridge the vast gulf between two different groups judging one another for two very different experiences. Remember, no man is an island, and one day, the bell might toll for your sense of identity.

But why the creative stuff? Why incredible images? Why can't they make the AI do my dishes and fold my laundry and make my bed? Why can't it do all the boring stuff that limits my time to do the cool stuff?

First it's important to realize that none of the AI researchers and companies actually intended to do this. That might seem like a strange thing to say. They started companies, those companies make money — from the product we are discussing. How could this not be intentional?

When I use the word intentional, I mean it in the sense that water coursing down the side of a mountain, carving a river in a flash flood does not intend to decimate a small town in it's path.¹ There's no intent whatsoever, all you have is the structural particulars of a certain circumstance, and these particulars play out their natural course of events.

Individuals have free will, but I don't think groups of people do.

Actually I kinda doubt everyone has free will. It probably exists on a spectrum and part of what determines placement on that spectrum is situation: If you are in isolation in a max security prison, it's hard to claim much agency, and by extension, much free will. There is a gradient between that isolated bastard, and Elon Musk who can materialize the largest factories known to man by, well, deciding to.

But then think about everyone involved in the actual construction of that factory. Do they have free will? Well, certainly they can quit their job, and that's apparently an expression of free will, but for the overwhelming majority who don't quit, they are fairly locked in. Not max security isolation locked in, but fairly locked in. And there in lies my point: that group of people cooperating and coordinating to build a factory have very little free will as a group.

If you think about all AI companies as a group of people, do they have much free will over what they are doing? No, I wouldn't say so. And the reason why is the Data Diet.

For one very big obvious reason, humans did not find it worth their time to meticulously record hundreds — millions of hours of POV footage of themselves washing dishes and uploading all of that footage to youtube. Strange isn't it? No, it's not strange at all, because who the fuck would want to watch that? Same goes for folding clothes or making the bed or tidying the house. No, humans like beautiful, entertaining and interesting things, so we post beautiful, entertaining and interesting things. We don't like tedious, boring and mundane things, so we don't post those things. What we have posted to the internet is the data available for the diet of AI.

AI companies used what data was available to train the models that we have. And what was available was all the books in existence and all of the pictures we've posted online. So that's why AI has become good at those things first. Key word their... "first". Because now that AI companies know that this model architecture works, now they are going to try and find more data to create new models that can do other things. There are already data sweat shops filled with people sitting at desks, folding clothes with cameras strapped to their foreheads and wrists. We will eventually have those robots doing our tedious chores, but again, it's a question of diet, and what data we have available for it.

This insight raises a very useful question: what can't be captured in a data format? Perhaps nothing, given enough time, but in the same way that locks, keys, safes and security systems are just deterrents against thieves, what useful aspect of our personal experience would be the biggest pain in the ass to capture in a data format?

Many seem to think taste is the final moat of human cognition, but I honestly question this. Good taste only exists if some exclusive cadre of high minded people agree that it has taste. And if it exists across more than one person, then I sincerely believe that it can be modeled. It might not be modeled as efficiently as a tasteful human brain can right now (in the same sense that the brain only requires 20 watts of power, which is nothing compared to running a single GPU for a query), but modeled I believe it will be.

So what's the biggest pain in the ass to get data on? I can think of one thing: your unique perspective as it relates to your singular placement in space and time, and the path you've taken to get there and how the experiences of that path have shaped the way you look at things.

Some of that could one day be modeled. And perhaps powerful inferences could be made regarding how you would interpret those experiences based on your particular genome, but it will never capture the full resolution of your lived experience. (knock on wood or slap me across the face if it turns out I'm typing this inside of a Matrix style simulation.)

Regardless, it'd be an enormous pain in the ass to gather this data. So what sort of practical use is this insight? Well, for one, the more normal your life, the easier it's going to be to make a simulation of you as a person, because there will be plenty of "like" examples. But the more wayward, random, unexpected and borderline batshit crazy your life's path is, the harder it will be to model, and likely the more tangential your perspective and well of insight may be. Simply put:

The weirdest life will be the last to be automated.

BISTABLE HYSTERESIS

April 4th, 2026

There exists a pernicious and particularly insidious problem that I've been trying to solve for years. Many attempts at solution have yielded a variety of effects that have slowly mapped out a model, or a thesis for the system that I'm trying to interact with. The variables are numerous, the interactions are countless and the sweetspot is elusive. Recently, I discovered the term to describe the system that I've been playing with: Bistable Hysteresis.

Life has some rare delights, and coming across a word or term that perfectly describes a model that's organically arisen in the mind after many hundreds of hours of concentration is one of them. I remember it happening often as a kid, but as an adult it's far more rare. The last time this happened, I believe, was with the word iatrogenic. Unfortunately in adulthood, this rare delight is almost always going to be accomplished by technical or foreign words. Tsundoku, hyperbolic discounting, hormesis, saudade, agathokakological. Ten cent words for sure that are more likely to elicit an eyeroll from a listener or reader who doesn't know wtf you're talking about. It's unfortunate that the predominate education system has such a talent for suppressing education - that innate ability that kids have.

Now that the self-conscious preamble to discussing a stuffy conceptual term is out of the way ( Deciding on the fly here that I'm probably going to code in expandable footnotes more like embedded notes ) into Tinkered Thinking - I love footnotes and methinks building expandable footnote capability right into a composer would actually help a lot with first drafts because writers would be able to go off on tangents without the sense that they've sacrificed the main vein of the essay, as I've clearly done here, now multiple times. To be sure, a parenthetical is something that couldn't make the cut as a true sentence, and a footnote couldn't make the cut as a parenthetical and an endnote couldn't make the cut as a footnote, and what gets left on the cutting floor, well, hopefully that's just bad writing.)

Imagine you are stuck in a valley. This is a bad valley. And you want to get out. The natural tendency is to just keep rolling down the valley. This is the design of the situation. And not only that, but staying in the valley makes it deeper over time. In this situation, what do you do?

The orthogonal solution. The only way out is to turn and go up the slope. This is very difficult, particularly if you are not accustomed to traversing steep slopes. The balance required is difficult and precarious. Each step requires a lot of effort. And It's easy to take a wrong step, slip on the footing and tumble back down into the valley. The process of getting out of the valley is simultaneously very difficult and very delicate — not a combo that people generally find themselves equipped with an innate talent to handle.

Think of this in the context of a real life situation: kicking a bad habit. You have finally had enough. You are sick of cigarettes or vaping. You throw out your stash. You feel triumphantly angry. Finito. Done. Voila! Your rack up a few glorious days. Ha, this is easy! And then something annoying happens. It's not just annoying, it's a tiny inconvenience that actually derails your whole day. It's not life threatening but it pisses you off. You want relief from how you're feeling. And then that craving, that desire for that sweet sweet release, that single puff that just brushes all of life away for a few moments. The memory of that feeling fills you and now you're thinking about how far away the nearest store is. You can't concentrate on the annoying problem at hand that needs to be solved. And It's boring in comparison to the good work you had planned. The craving is now hijacking your logic. You deserve it because you had the best of intentions to get good work done, but you hit a blocker that wasn't your fault. You are being unfairly punished. You deserve something to balance the cosmic scales of justice. Just a little taste. Just one won't hurt.

That's the misstep. The combo of delicate and difficult becomes too much. That single puff sends you tumbling back down into the valley, and you dig in the depth of that valley, making it a little deeper, making the slope a little higher.

It doesn't have to be an addiction. It can be something like exercise, and getting yourself to do it consistently. A couple of days in, you feel good, you're doing it! But then one off night of sleep, or you're too sore, and suddenly the excuses start piling up. I should just do a little cardio, maybe just sit in the sauna to help maintain the habit. But you say fuck it. That night you binge on some ice cream - you deserve it. The crazy glucose spike right before bed ruins your sleep and you wake up feeling miserable. Well you won't have a good workout after such a bad sleep, so it should be another rest day. And before you know it, you're back down in the valley.

A person can spend many months, years, even decades trying to climb that slope, and each failure can be demoralizing: the definition of insanity is doing the same thing over and over and expecting a different result. The valley metaphor also shows how a bad situation can be self-reinforcing in more than one way, it can be psychologically self-reinforcing because any attempt at change seems hopeless.

The funny thing is that a lot of skill acquisition looks like insanity. Trying to ride a bike, or a unicycle or ice skating, often requires falling over and over until finally, a different result emerges.

Now before we extend the metaphor and finish it, let's explore a different one: a train on its tracks.

We are (ideally) born with good health. And from a the perspective of a biological system, its functioning is much like a train riding along on its tracks. But as we get older, bad habits begin to have an effect. Like filing off the inner lip of a train wheel. If it degrades enough the train can slide off the tracks. Ask yourself, what's easier: to get a train off the tracks? Or, to get a train back on track? Both are pretty difficult, but surely if you had to pick one, it would be getting a train off the tracks (just go full speed on a tight turn and voila. But getting a train wreck back on track? uhh. good luck). Because of the design of the the system, the way on and the way off have totally different mechanisms. Just like climbing out of a valley and sliding down into a new valley are two different directions and methods. This is hysteresis: a system's current state depends no just on its current inputs but on its history and therefore the path to get into one state is different from the path to get out of it.

Bistable just refers to each state (assuming the system has just two states) is self reinforcing. Meaning, the longer the system is in one particular state, the harder it is to switch states. Or put in practical terms: the longer someone is unhealthy, the harder it becomes to get healthy. And the opposite is also true: Someone who has impeccable health can withstand more unhealthy slings and arrows than someone who has just entered a positive state of health.

Hormonal health is a great example of bistable hysteresis: hormonal profiles are often self-reinforcing — even the bad ones! We often have an assumption that the body drifts back towards a healthy state by default. But that's not the case. Because of bistable hysteresis, a bad hormonal profile, or hormonal health that has gone off the tracks, can self-perpetuate. Here's an example: Low Free Testosterone shifts body composition to have more adipose fat tissue, particularly visceral fat (fat around organs). This fat is not just stored energy, it's an endocrine production site, meaning it produces stuff that influences hormones. What does adipose tissue produce? Aromatase. What does aromatase do? It converts Testosterone to Estrogen... which lowers testosterone even more, which shifts the body to have even more adipose tissue, which produces more aromatase, which converts more testosterone to estrogen, which... you get the picture. It's a vicious cycle.

Sleep is also going to suffer in this situation, and good sleep is crucial for a good hormonal profile, so if a bad self-reinforcing hormonal profile is further exacerbated by bad sleep, well, it's like an Ouroboros lapping itself.

So back to our valley metaphor. Getting out of the valley isn't the only hard part. If you manage to get to the crest, you don't necessarily fall into a valley of great health. It's more like the valley is a canyon in a plateau and it doesn't take much to aimlessly wander back into the bad valley.

If you want to stay out of the bad valley, you have to dig a new valley.

This is the part of hysteresis where getting into a new state requires a different method from getting out of the old state.

It's a bitch to dig a valley, but after digging just a little bit, you're already making it difficult to get out of. This is a good thing, it's self-reinforcing in a virtuous way.

So what's useful about understanding bistable hysteresis and having a word for it? Two things: one is understanding the process of getting out of a bad valley — it's delicate, difficult and easy to backslide back to zero. The second is the psychological reality of getting out of the valley and reaching the plateau. It can feel like victory, but unfortunately it's only one half of the solution. It's like when you finally figure out how to lose weight. Maybe you get to the point where your abs are showing. You have more energy, you feel amazing, and you love the way you look. It's a victorious feeling, and along with that comes an urge to celebrate. Do you celebrate by eating a salad an hitting the gym? Or do you feel like you've earned the right to let lose a little and party? A cocktail, a late night, some fast food. Hell, you deserve it!

Paradoxically, the most appealing thing to do when you arrive at the plateau is to turn around and look at the slope you just climbed as a fun slide.

The feelings of victory induce a desire to undo that victory. But the correct way to react to the feeling of victory is to double down on the strategy that got you there: to do the work to dig a new valley, one that'll protect you from sabotaging your own gains.

What's most interesting about bistable hysteresis in human systems, particularly biological systems is how accurately it represents reality, and how at odds it is with human psychology. You are constantly working against hardwired incentives, and true success requires an understanding of an abstract process for a goal that's hard to feel. Few people seem capable of changing their behavior based solely on a conceptual understanding, but for those who can, bistable hysteresis is an excellent model. I'm starting to see it unexpected places: human relationships, business incentives and corporate structures, sycophancy and delusion and even AI psychosis.

Albert Camus once said that we must imagine Sisyphus happy. I prefer to imagine that he found the plateau.

Older

subscribe

rss Feeds

Most Popular

A LUCILIUS PARABLE: THE LAST THOUGHT

July 26th, 2026