The third of three. The first was the mechanism. The second was the threat. This one is the person.
I built an ontology and called it Fuckery. I named the loop at the centre of it Fuckari. In the glossary I wrote for that framework there is a third word, Fuckarian, and I defined it as a being constituted by its assumptions — constrained by them in the same act that lets it think at all. I wrote that entry about nobody in particular, the way you write a definition. Then some weeks later I read the table back, and the third row was looking at me.
I am a Fuckarian. I have been one the entire time. The framework and the protocol were both built by one, and it seems only fair to say so out loud before going any further.
There is a rule in the Fuckery README that I put there myself and have so far declined to apply to the work: reasoning trails are first-class here. The path by which a claim was reached is not scaffolding to be thrown away once the claim stands. Every claim in that framework is marked for provenance — invariant, derived, speculative, illustrative, reasoning trail. Every claim except one. The framework itself has been presented as though it arrived from nowhere in particular, which is the one thing its own rules do not permit.
So this is the missing tag.
I have been asking people to check my reasoning trails. Here is mine.
What the first two were
The short version, for anyone landing here first. Skip it if you already know the other two.
Fuckery is a framework about constraint and emergence. Structure forms under constraint, and then the structure becomes constraint, and whatever forms next forms under conditions the last thing left behind. The loop that describes is called Fuckari. I ran it against a snowflake for twenty-eight steps, in public, because a snowflake is an object whose physics nobody disputes and in which nobody has a stake. If the framework was going to fail, it should fail there, visibly, on something harmless.
The Constraint Protocol is a separate thing and a more serious one. It is an open public ledger for declaring what an entity will not do, and for making conduct under that declaration checkable by anyone who cares to check. It is neutral about the constraint and neutral about what kind of thing is holding it.
Both are linked at the bottom. Neither is required reading for this one, which is the only one of the three that is not about an argument. It is about where the argument came from, which was not a snowflake.
The things I have been called
Over the years a fair number of people have had a go at naming what I am. Neurodivergent. Somewhere on the spectrum. ADHD. Some of those came from people who love me and some from people who were being professional about it, and every one of them was an honest attempt to fit a word to something that was not obliging anybody by holding still. That is a reasonable thing to attempt and a hard thing to get right. I am not going to take any of them up and I am not going to argue with any of them either.
At the end of the day, I’m just me.
Conditions
Here is the thing I know about myself with some confidence, because it has been tested often enough to count as tested.
Given love, and respect, and patience, the words come out.
Given any less than that, they do not. Not less well. Not at all. It is not a preference, and it is not something I can push through by trying harder — I have tried harder for long enough to be sure about that part. It is closer to a precondition, in the way light is a precondition for a photograph. You can have the camera, the subject, the film, the intent and the whole afternoon, and in the dark you still get nothing, and no amount of wanting it changes the chemistry.
I am aware of how that sounds. It is a demanding set of conditions to attach to a conversation, and most conversations in a life do not have the budget for it. A shop does not. A meeting does not. A first introduction certainly does not, and neither does the middle stretch of a long relationship, where everybody is tired and the day still has to be got through. That is not anybody’s failing and it is not a shortage of goodwill. It is arithmetic. Those conditions cost something to supply, they have to be supplied by a particular person out of a finite amount, and nobody owes them to anybody.
So for most of my life the words did not come out.
That is the mechanism. Everything else in this post is downstream of it.
The hard part was never starting
People assume the difficulty is starting. It is not. Starting a conversation is easy. I have started thousands of them and I am perfectly fine at it, in the way anyone is fine at a thing they have done thousands of times.
The difficulty arrives immediately afterwards.
Once the conversation is started, I have to get across what it is like in here, and there is no compression for that.
It is not a fact I can hand over. A fact goes in a sentence, and the other person either has it now or does not, and either way the transaction closes. It is not an opinion either, because an opinion can be agreed with or disagreed with, and both of those are fine outcomes — disagreement is a form of having received the thing. What I am trying to convey is neither of those. It is the set of conditions that everything else I say is said from. It is the room the sentences are spoken in, rather than the sentences.
And if the room does not come across first, then everything after it lands in the wrong place. Not rejected. Filed. Filed accurately and in good faith, according to a filing system built for a different room, which means it is now somewhere I cannot retrieve it from, and the other person has been entirely reasonable throughout, and we have both spent the afternoon.
So the conversation I actually want to have is a conversation about the conditions for the conversation.
There is no polite way to open with that. There is no impolite way either, which I have also checked.
Listen
All my life I was told to listen. Alongside it, and often in the same breath, I was lectured about responsibility.
Neither of those was unfair. I did need to listen. And there were long stretches where responsibility was not a word anybody could have used about me with a straight face — that is not modesty, it is just the record.
What could not be seen, because I had no way of showing it, was that the same thing was running in the other direction at the same time, silently, and stayed silent for most of a life.
I was not being heard either.
Not because anybody refused to hear me. Nobody refused. The reason is duller than that and it is located entirely in me: the words did not come out. Underneath them I was screaming — what is this, what is this world, why, why is it like this, what am I supposed to be doing with it — and not one syllable of that ever made it into a room with another person in it. It was not withheld. It was not being saved up for later. It simply had no route to the outside, and from the outside a thing with no route looks exactly like a thing that is not there.
So the instruction to listen arrived, correctly, from people who could see that I was not listening, and it landed on somebody who was at that moment at maximum volume internally and entirely silent externally. Both of those were true at once. Only one of them was visible, and it was not the one I would have picked.
Now the part that is mine, because everything above is only half of it, and it is the easy half.
I was waiting for somebody to have the answer.
Most of us do not have the answer. Almost nobody has the answer. I was asking the people nearest to me — which is to say the people with the least ability to step away from being asked — for a thing that is not in any person’s possession, and then quietly marking them down for failing to produce it. Not out loud. I never said it. That made it worse rather than better, because a standard nobody has been told about cannot be met, cannot be argued with, and cannot even be known about. I ran that arrangement for a very long time on people who had done nothing to invite being examined.
That is my share. It is not a small one.
What it cost
It caused real havoc in my personal life.
I want to say more about that than the sentence does on its own, because a sentence that short, left sitting there, invites a reader to go looking for an event. There is no event. That is the part worth getting right, and it is the part that took me longest to see. Havoc does not require an event. What it was, was a pattern, run for a long time, by me.
The pattern went like this. I would be in the room and not reachable. Not absent — I was there, I was talking, I was doing the things that get done. But the part of me that the conversation needed was somewhere else, working on something, and I could not bring it back on request and I could not explain where it had gone. Then, at intervals, I would surface and need an enormous amount from whoever was nearest: attention, patience, a hearing for something I had no way of compressing into the time available. And then I would be gone again.
From the inside that is one continuous state with a slow rhythm to it. From the outside it is not a state at all. It is a sequence of unannounced arrivals and departures, with no schedule, no warning and no stated cause, and there is no way to plan a life around it.
Here is the thing I did not understand at the time and understand very well now. A person who cannot say what is wrong and a person who will not say what is wrong look exactly the same from the outside. There is no observable difference between them. None at all. Everybody around me was working from the evidence available, and the evidence available supported the second reading at least as well as the first — better, in fact, because the second is the common case, and going with the common case is not a flaw in somebody’s judgement, it is what judgement is. So that is how it was read. That was not a misreading. It was a sound inference from the data, and the data was all I had supplied.
And the reasonable response to a person who will not say is to give them room. Stop asking. Take a step back and keep something in reserve. Every one of those is the correct move, and every one of them was made by people who were being careful with themselves and were, for the most part, still being kind to me while they did it. I would have done the same. I would have done it sooner.
The other half of the pattern is the one from the previous section, and it is worse, so I will put it plainly rather than gracefully. I was asking the people closest to me for something almost nobody has. Not comfort — an answer. An account of what this is and why it is like this. I never put it in those words, because I could not put it in any words at all, but that is what the asking was for, and the asking did not stop, and it could not be met, because the thing I was after is not held by other people. So the people nearest to me were handed a test they could not pass, repeatedly, over years, without ever being told there was a test. Nobody agreed to that. Nobody was informed it was happening. It happened anyway.
What I understand now, which I did not then, is the accounting, and the accounting does not come out in my favour. I was treating other people’s patience as though it were weather — a thing that is simply there, that renews by itself overnight, that nobody has to do anything to produce. It is not weather. It is a finite thing that particular people spend, deliberately, out of a supply they need for their own lives, and when it runs out it has not been lost. It has been spent. Somebody spent it, on me, and then did not have it for something else.
I am not going to write the rest of it. Not because it is too much to write, but because the rest is not mine to write — every further sentence would need somebody else standing in it, and they did not choose to be in a blog post, and the version of them I could put here would only ever be my version of them.
The sentence is in this post because leaving it out would make everything else in here sound considerably cheaper than it was.
I talk like a numpty
There is a mismatch here that I should probably name, since everybody else has had to work around it without being told it was there.
I talk like a numpty.
Most of the time that is an accurate read, because most of the time I am being a numpty. This is not self-deprecation. I like it, it is genuinely most of what is going on, and most people who have met me got that and nothing else — which was not a partial view of the situation. It was the view.
Then every so often something comes out about constraint regimes and top-down causation. Same voice. Same bloke. Same delivery, no change of register, no signal of any kind that the thing now being said is of a different type from the thing said ninety seconds ago about the bins.
Nobody has any reason to take the second thing seriously.
People use the source to rate the signal, and that is neither laziness nor unfairness — it is the only method available at scale. Nobody can evaluate every claim on its merits; there is not enough life for it. So a person builds a model of the source and runs on the model, and keeps the model updated as evidence arrives. The model of me is accurate. It is accurate because I built it myself, out of evidence I supplied personally, consistently, over decades.
It is a good heuristic. It is how a person avoids spending their one life on nonsense.
It just happens to be wrong in this particular case. And there is no way for me to say so that does not sound precisely like what somebody would say if the heuristic were right.
That is not a complaint about the heuristic. It is the cleanest trap I know of, I am inside it, and there is a certain amount of pleasure to be had from admiring the engineering.
Things I built that nobody used
Before any of this I built a number of things that were genuinely useful and went genuinely unused.
I want to be precise about that, because there are two versions of that sentence and only one of them is true. The false version is that the things were good and people were wrong. The true version is that the things were useful and were not adopted, and useful and adopted are two different properties, and the second does not follow from the first. It never has. There is no mechanism by which it would.
Adoption is its own discipline. It wants patience with people who are not yet interested, which is most people most of the time and is not a criticism of them — everybody is already full. It wants a tolerance for saying the same thing a fourteenth time with the same energy as the first. It wants a read on when somebody is busy, and the ability to come back later without sulking about it in the interval. It is real work, it is done properly by people who are good at it, and I am not one of those people and never have been. For that category of task I have roughly the attention span of a firework.
So: useful things, unused, and the explanation sits on my side of the table. Nobody failed here. A thing that exists and is not adopted has not been rejected. Most of the time it has not been seen, and the seeing was my job, and I did not do it.
For a while I treated that as a problem to be solved. I read about it. I tried. I got marginally better at it in the way a person gets marginally better at anything they are constitutionally unsuited to, which is to say measurably, slowly, and never enough to change the outcome.
Then I stopped.
So I stopped waiting
What I did instead was stop building things for other people to use, and start building things that made the next thing easier to build.
That is a different objective function and it has one enormous advantage: it does not require anybody’s agreement. Nobody has to adopt a tool whose only user is me. The adoption problem simply does not arise, because the audience is one man and he was convinced before the work started.
The compounding is the entire point. Each tool shortens the distance to the one after it, and the shortening multiplies rather than adds, so the early ones look like nothing at all and the later ones look like cheating. There is no moment at which it becomes impressive. There is a long stretch where it is obviously a waste of time and would be scored that way by any reasonable observer, and then the same curve goes on doing the same thing it was already doing, and at some point the outputs stop being commensurate with the inputs.
The cognitive companion is where that got to.
It exists, it works, and it does a great deal of my programming now. Not as a demonstration, and not as an assistant I supervise while it produces a junior version of what I would have written myself. It writes the code while I work on what the code is for. I would rather overstate that distinction than have it read as a productivity anecdote, because it is not one: the thing I do all day has changed category, not speed.
Which is what freed me up to go after the Internet of Cognition.
By that I mean the layer above the companion: companions and agents connected to one another and exchanging work, the way the Internet of Things put devices on a network, except that the things on this one think.
That is not reachable from a standing start. It needs a foundation underneath it — the constraint work, the definitions, the protocol, the companion — and the foundation is the part everybody skips, because the foundation is invisible in the finished thing and takes most of the time.
Pegs, not stairs
I am not building a staircase. I want to be clear about that, because people keep reading it as one and then being disappointed by the handrail.
What I am doing is driving pegs into a rock face and climbing them, one at a time, to see what is over the edge.
A peg holds my weight. That is the entire specification. It is not level. It is not wide enough for two. It has been tested exactly once, by me, under load, at the moment I put my weight on it, and there is no margin in it for anyone heavier, or slower, or carrying something they cannot put down.
Nobody should be asked to trust it with anything they care about.
Turning pegs into stairs is real work and it is a different discipline entirely. It means going back down. Fortifying what is already in. Widening it. Adding the part people hold on to. Working out what happens in the wet, and what happens at night, and what happens when somebody arrives at the bottom with their own reasons and their own hurry and no interest whatsoever in the view.
That work matters enormously.
It is not mine. I am not qualified for it, I am not temperamentally suited to it, and if I stopped to do it I would stop climbing, which is the one thing I am actually for.
Where the two threads meet
Put the two halves of this together and you get the thing the post is actually for.
The conversation moves at the speed of a conversation. That is the correct speed for a conversation. It has to bring people along, it has to survive contact with reasonable objections, it has to be checked — and the checking is the expensive part and by a distance the most valuable part. Nothing I have written here is an argument for going faster than that.
I move at the speed of somebody who stopped waiting for one.
So by the time the discourse arrives at the first peg — and it does arrive, it always arrives, and the people arriving are not slow, they are being careful, which is a completely different thing and the better of the two — I am ten pegs up and looking at an obstacle I have no vocabulary for yet.
That gap is not a credential.
It does not make me right. Being early is uncorrelated with being correct, and a man on his own on a rock face has exactly the error correction he brought up with him, which is not much and is all of it his own. What the gap does is make me badly placed to explain myself and well placed to see the next problem early. Those two come as a pair. They are the same property looked at from either end, and I have not found a way to have one without the other.
So when I say the Constraint Protocol is one possible tool for an urgent problem, read that as recollection rather than advocacy. It is not a pitch and there is nothing to buy. It is where the climb has got to and what I am working on now. It is one tool. It may well be the wrong one. And I would rather publish it and be corrected in public than hold it back until it is finished, because I do not think there is time for finished, and I would rather be usefully wrong on a schedule than privately right too late.
To the people saying slow down
What follows is opinion, and I would rather label it as opinion before I say it than after. I have no data a reader does not have, no inside knowledge of anything, and nothing at all to say about any particular person or company. This is just where I have landed, offered as one man’s position.
There is a set of people who have started saying we should slow down, and who are being treated as alarmists for saying it. I would like to be counted with them, and more plainly than most of them put it.
Yes. It is that bad.
I am not being contrarian about this, and I am not doing doom either — doom is a posture, and it is about as useful as the posture it is reacting to. I am simply in agreement, and I would rather be in agreement out loud than have it inferred from the fact that I keep building things.
The comparison I keep coming back to is climate, and the comparison has a sting in it.
We have spent decades unable to act on a threat we could measure, describe, model and very largely agree on, because the costs of acting were near and concentrated and the damage was far off and distributed. That is the exact shape of problem that human institutions handle worst. It is not that anybody involved was stupid. Very nearly every actor in that story was responding rationally to the incentives directly in front of them, which is precisely why it did not get solved. The failure was structural rather than moral, and structural failures do not get fixed by locating somebody to be angry at, which is unfortunate, because locating somebody to be angry at is the thing we are all best at.
Now the sting.
The thing most likely to get us out of that is more intelligence applied to it. Better modelling, better materials, better allocation, better chemistry, better everything, sooner. And at this particular juncture, the thing we are calling AI is the clear and present danger.
So the instrument and the hazard are the same instrument.
I do not have a resolution for that. I am not sure one is available at the length of a blog post, and I am fairly sure it is not available from me. But pretending that is not the shape of the problem has not helped anyone yet.
A new kind of electricity
Here is how I hold it. This is my own frame rather than a claim about the world, and it is in this post because it explains why I built the thing I built — not because I want to argue it with anybody.
What we have is a new kind of electricity.
Not a product. A substrate. Something drawn on everywhere, rather than a thing used in one place for one purpose.
And the thing about live current is that nobody works it bare-handed. An electrician has insulated boots. Gloves rated for the voltage. A mat. And behind all of that, years of training whose actual content, once you strip the syllabus off it, is a list of what not to touch and in what order. That training is not optimism about electricity. It is what a person looks like who has thought carefully about the specific ways this kills people, and has arranged themselves accordingly, and intends to keep doing so on a Tuesday when they are tired.
None of that equipment makes electricity safe.
It makes one particular person safe while doing one particular job, and everybody in the trade knows the difference, and the ones who forget the difference are the ones it happens to.
A cognitive companion is that tool set. That is the whole of my claim for it. It is a person’s own equipment for working with this substrate — an extension of the person, not a separate agent acting in their place. If it acts in the person’s place, or answers to somebody else, it is not a companion, and calling it one is the first error rather than a harmless one.
Which is why I needed equipment and not an assistant.
An assistant is another party in the room. Another party in the room is somebody I have to get the world across to before anything useful can start — see most of this post for how reliably that goes. Equipment is a different category of thing. Equipment does not have to be brought up to speed every morning. It holds the context, it keeps the trail, and the conveying has to survive once rather than being attempted daily and failing daily. For most people I imagine that is a convenience. For me it was the difference between the work existing and not existing.
Then there is the word, which is the other half of this, and the half that costs everybody rather than just me.
Nobody standing on a train listening to music on their phone describes themselves as an electricity user. They are listening to music. The electricity is four layers down, and not naming it is not ignorance — it is the correct level of description for what is actually happening.
We say “AI” for the model, the product, the company, the industry, the research field, the risk and the future, all in one syllable. And because the one word covers all of it, every conversation has to settle whether the entire category is good or evil before anybody has said which thing they mean. That is not a disagreement. It is a naming failure, it sits upstream of the disagreement, and it is why so many of these conversations feel like they are being conducted, at volume, by people who agree.
It is also why the definitions I work from say nonbiological intelligence and not artificial intelligence. Artificial prejudges the thing as counterfeit and smuggles a verdict into the name. Nonbiological states the substrate and stops.
That argument is written down properly on the definitions page and I am not going to improve on it here. What the definitions page does not say is why I needed it, so that is the part I will add: name the thing being done, not the substrate it runs on. I did not arrive at that rule out of tidiness. I arrived at it because every attempt I ever made to talk about this without it turned into an argument about something else inside two minutes, and I do not have two minutes to spare on the way in.
Bothans
This one did not cost sleep.
It cost something else, and the people it cost it to have mostly been kinder about it than the situation warranted. Some of them are still here and some of them are not, and in both cases they were working from what they could see, which was never very much, because I could not show them. They were generous on an amount of information that did not justify generosity.
I am not going to name anybody. They know.
“Many Bothans died to bring us this information.” — Mon Mothma, Return of the Jedi
If you would like to help with the next one, the page is here: ko-fi.com/jeyanandan. Nothing is bought by giving — the framework stays CC BY 4.0, this essay stays public, and nobody who gives gets anything a reader does not already have.
The framework itself lives at github.com/ObsidianDelta/Fuckery. The Constraint Protocol is at daxfoundation.org/#cp. Every term used above is defined precisely at daxfoundation.org/#definitions, including cognitive companion and AI as a general-purpose substrate.