Apparently eight-year-old me was already building systems, rushing the explanation, and being told to show his working.
My nan found my Year 3 school report the other day.
Not a vague memory of it. Not a reference in passing saying, "You were always bright as a child," while I quietly upgraded the story over time. The actual report: printed paper, teacher comments, attainment levels, everything.w
And honestly, it has stitched me up a bit.
I have spent the last year writing essays and building systems around a fairly specific collection of problems. How can somebody see the structure of something before they can explain it? What is the difference between ability and access to ability? Why do visible outputs so often conceal the conditions that produced them? How do you stop a conclusion becoming more confident than the evidence underneath it?
Apparently, my Year 3 teacher had already identified most of the situation.
I was eight.
The report is from 2009/10. I was in Class 3 Purple, which sounds less like a class and more like a low-budget superhero team.
At first, I looked at the numbers.
Reading: 4C.
Writing: 3C.
Mathematics: 4C.
In the old National Curriculum system, Level 4 was broadly the standard expected around the end of primary school. So a 4C in reading and maths during Year 3 meant I was working several years ahead in those areas.
That is nice.
Gold star for young Joe.
But the numbers are not really the interesting part. The numbers tell you that I was doing well. The comments tell you what kind of mind was doing it, where the joins were already visible, and which parts of the process were going to become increasingly expensive later.
At first, the report reads like a prediction.
Then it reads like a warning.
By the end, it starts to look suspiciously like a design brief.
It is also an unusually kind record of me.
That matters.
A school report is a measurement taken through one particular window. This one catches the child who could read, make connections and communicate enthusiastically. It does not contain every home, school or environment he had already learned to recalibrate around. It does not record how quickly he was learning to read rooms as well as books.
Some of what looked like natural adaptability may also have been adaptation in the literal sense.
The report was real. It was never the whole child.
The original systems audit
The reading section says:
"Joseph can identify and summarise evidence from a text to support his ideas and opinions."
Which is funny, because a large part of my current work is basically asking:
What is the evidence? What does it actually support? Where has the conclusion become stronger than the thing underneath it? What survived the process, and what did we quietly add around it?
Then the maths section says:
"He has a good understanding of inverse operations and the relationships between different operations."
And:
"Joseph can answer a question by identifying what data to collect and interpret it in diagrams and charts."
Again, I was eight.
Now I build engineering audit systems that change assumptions, disturb models, compare relationships and map what remains stable. Apparently, I did not choose this life. Apparently, I was simply born asking what data we need, how the operations relate and whether the diagram is actually showing what everybody says it is showing.
The science section is even less subtle:
"Joseph shows enthusiasm for the practical aspects of scientific investigation."
"He shows a good understanding of the importance of fair testing, recognising and explaining why a test is fair or not."
"He puts forward his own ideas about how to answer a question and is also developing the ability to construct questions for scientific enquiry."
There is something almost rude about reading that now.
Years later, one of the central things I would build was GIL, the model-audit process inside Systemica. Its whole job is to stop asking only whether a model produced an impressive score and start asking whether the route to that score was fair.
What happens if the data changes? What happens if a feature is removed? What happens if the population is split in a harder way? Does the interpretation remain standing when the test stops protecting it? Did the model discover something that belongs to the problem, or something that belongs only to the particular route it took through the problem?
That sounds considerably more technical than a Year 3 science lesson. It is. My teacher was not secretly specifying a governed materials-informatics audit architecture in 2009.
But the cognitive move is recognisable.
Do not just accept the answer. Inspect the test that made the answer possible.
Year 3 Joe was running the pre-alpha version.
The recurring offence
The funniest sentence in the report is probably this:
"Joseph’s ability to make connections across his learning and to explain his thinking have made him a pleasure to teach."
That is a lovely thing to read about yourself.
It is also extremely funny when your adult life now consists of connecting materials science, cognition, complex illness, AI governance, evolutionary systems, engineering assurance, cosmology and whatever else has entered the building this week.
"Ability to make connections across his learning" is probably the nicest possible description of what I do.
The less generous description is:
Joe has opened another research programme because two things looked structurally similar.
Both are true.
My work has grown by noticing that the same shapes appear in different places. Stability. Drift. Collapse. Recovery. Pressure accumulating faster than a system can dissipate it. Visible output remaining intact while the structure supporting it quietly changes. History altering which futures remain accessible.
I followed those recurrences through materials, illness, cognition, institutions, machine learning and engineering. Eventually I wrote The Same Four States Keep Appearing Everywhere because, by that point, pretending I had not noticed was becoming difficult.
The Year 3 report does not prove that I had a universal state theory hiding behind my Pokémon cards. I was probably thinking about Lego, games and why the classroom had been arranged inefficiently.
But it does capture the earlier form of the same impulse:
Find the relationship. Connect the areas. Explain the pattern. Then continue well beyond what was required.
Which brings us to the next comment.
"Joseph is more eager to show his full potential, Joseph needs to make sure that he is never satisfied with the minimum needed to complete a task, especially where the work involves writing."
"Especially where the work involves writing."
That sentence has aged like a threat.
Give me a simple article and I will create a framework. Give me a framework and I will create a programme. Give me a programme and I will create an Atlas, a canonical register, a relationship graph, a governance layer, an evidence repository and twelve supporting documents explaining why none of the previous documents should be trusted without their receipts.
I do not know whether the teacher meant, "Joe should challenge himself a little more."
Adult Joe appears to have interpreted it as, "Never complete a task until it has developed its own constitutional architecture."
So yes, the drive was apparently visible early.
But that is only the first meaning of the sentence.
It changes later.
Please show your working
The report contains two targets. They are ordinary school targets. They are also, with the benefit of hindsight, almost absurdly precise.
The literacy target says:
"To keep in mind the effect on the reader when writing. This means ensuring that he takes time creating effect and setting a scene and not rushing to finish."
The numeracy target says:
"To further develop his ability to explain what thinking and mental methods he has used to find an answer."
There it is.
Do not just arrive at the answer. Show the route.
Do not assume that because the structure is clear inside your head, it has transferred automatically into somebody else’s. Slow down. Set the scene. Think about the reader. Explain the mental method.
I have spent the last year formalising increasingly elaborate versions of that exact instruction. Slow the reasoning down. Expose the route. Separate what was observed from what was inferred. Check whether a comparison preserves the mechanism. Make it possible for another person to see not only the destination, but how the argument got there.
Eight-year-old me was being told to show his working.
Adult me built an operating system for it.
That is funny, but it is also where the article stops being a story about a clever child and becomes something more personal.
Because the problem was never simply that I did not understand the need to explain myself. The problem was that recognition and translation did not happen at the same speed.
I could often see the destination before I possessed a path another person could walk.
The structure arrived before the language
I could often answer a problem before I could explain how I had reached the answer.
The structure arrived before the language.
That sounds slightly mystical. It was usually much less glamorous. A teacher would ask how I knew, and I would have to walk backwards through something that had seemed to arrive all at once. I could see that the rooms connected. Now somebody wanted the corridor plan.
In Why Ability, Outcome and Effort Are Not the Same Thing, I later used the image of a complex structure inside a shoebox.
The structure exists, but the box will not open. You can only see it through a small hole. Sometimes you can navigate what is inside. When somebody asks you to transfer it cleanly, though, the opening is too small for the whole thing to pass through at once.
My Year 3 report catches a small, harmless version of that problem. The answer was there. The relationships were there. The teacher still needed the route by which they had become available to me.
She was right to ask.
Without the route, another person cannot tell whether I made a real connection, guessed beautifully or skipped across a gap I did not notice. Understanding and translation are not the same operation.
As the structures became larger, the unbuilt routes started accumulating. I had more models than papers, more connections than diagrams and more things I could partly see than things I had made usable for anybody else.
In Compression Planes of Cognition, I started calling that queue translation debt.
It is what builds when understanding moves faster than communication. From the outside, it can look like rushing, vagueness, too many ideas or unfinished work. From the inside, it feels like a backlog of structures waiting for an interface.
The school report caught the early version beautifully.
Bright, articulate, making connections.
Also rushing the effect, needing to show the mental method and forgetting that seeing the destination is not the same as building the road.
At eight, this was a writing target.
Later, it became infrastructure.
The years when just about counted
There were years between the clever child and the ill adult when the gap remained easier to hide.
I was still being placed in Able, Gifted and Talented categories. I could still produce the result. I could also end up crying at a textbook because understanding the material and making myself revise it were apparently being handled by two departments that had stopped speaking.
I became very good at just about doing things.
The dangerous thing about just about doing something is that, from the outside, it looks remarkably similar to doing it. The result arrives. The effort disappears.
Then the peak becomes the new baseline.
The next time, everybody reasonably expects you to produce it again. You expect it too. Nobody can see that the previous result was assembled from urgency, improvisation and capacity borrowed from somewhere else.
Long before chronic illness made the distinction impossible to ignore, performance and control were already not the same thing.
Then the cost model changed
The part the school report could not predict was what would happen when the body stopped being a reliable support system for the mind.
In When the Mind Runs Faster Than the Body, I tried to explain the strange experience of still being able to generate complex internal structure while losing reliable access to energy, memory, movement, administration and recovery.
Some regions remain lit. Others fail. You can still think, connect and occasionally produce something that surprises people, but the cost of moving it into the world has increased.
Emails become expensive. Forms become expensive. Task switching becomes expensive. Explaining the same thing repeatedly becomes expensive. The invisible work around the visible output starts consuming more capacity than the output itself.
This is where the childhood praise and the adult difficulty collide.
"Show your full potential" is a healthy instruction when effort is followed by ordinary recovery. "Never be satisfied with the minimum" sounds admirable when more work simply means more work.
Under chronic illness, pain, fatigue, cognitive fluctuation and post-exertional consequences, the same instruction can become dangerous.
The instruction did not become false. Its cost model changed.
Nobody had taught me that showing my full potential might also mean protecting the conditions that allow any of that potential to be available tomorrow. Nobody had explained that the minimum needed to complete a task might sometimes be the maximum a body can sustainably supply.
A child who is never satisfied with the minimum can become an adult who does not know how to stop before the work costs too much.
A child eager to show his full potential can become an adult repeatedly trying to prove that the potential still exists, even when access to it has become unstable.
And a child praised for making connections can become an adult with more ideas than energy.
The connections do not disappear.
The ability to formalise them becomes the bottleneck.
The trait did not disappear.
The operating conditions changed.
That distinction matters because the visible work can make the change harder for other people to see. If I can still write the article, explain the theory or build the system, then surely the system producing it must be functioning.
But output is not a medical scan of the conditions that made the output possible.
Sometimes it is only the part that survived.
When your output becomes a fossil
In A World Full of Fossils, I began with an Iguanodon in Crystal Palace that still wears its thumb on its face.
The bones were real. The teeth were real. The spike was real. But there was no complete animal, so Victorian scientists had to decide what belonged around the pieces that survived. The spike became a horn, and a huge rhinoceros-like body grew around it.
The evidence was genuine.
The animal was wrong.
That essay asks how often we make the same mistake elsewhere: treating the surviving structure as if it contains the whole history that produced it.
Reading my school report, I realised that the fossil problem runs in both directions across my life.
The Year 3 report preserves genuine evidence of early ability. It shows that the pattern recognition, curiosity, scientific instinct and urge to communicate were not inventions produced retrospectively by adult illness. They were visible before the body entered the story.
But my adult outputs preserve genuine evidence too, and they can be reconstructed just as badly.
A completed assignment becomes evidence of stable function. A strong conversation becomes evidence that cognition is reliably accessible. A published article becomes evidence of sustainable writing capacity. A technical system becomes evidence that I can run the surrounding business, administration and communication without support.
The output is real.
The body built around it can still be wrong.
This is the problem I keep returning to across different domains. In Drift Is Not Noise, I wrote about systems whose visible behaviour remains within an acceptable range while the structure supporting that behaviour is changing underneath. A material can still carry load while fatigue narrows its remaining future. A model can keep a good score while its feature relationships become unstable. A person can complete an assignment by borrowing against several days of recovery.
The momentary performance is not fake. It is simply not the whole state.
That is why ability, access, output and sustainability have to remain separate:
Ability is what the system can, in principle, do.
Access is whether that capacity is available under the present conditions.
Output is what made it through those conditions.
Sustainability is whether producing it preserved the possibility of doing so again.
Collapse those into one number and very strange reconstructions become possible.
Good result: naturally capable and therefore fine.
Poor result: not capable.
Inconsistent result: not trying.
Finished work: functioning well.
None of those conclusions necessarily follows from the evidence alone. They are tissue added around the bone.
The Year 3 report tells me that the ability was present early. My adult health tells me that ability does not guarantee access. The work shows that output remains possible. The recovery cost shows that possibility is not the same as sustainability.
All four statements can be true at once.
I think that is why I have become so obsessed with keeping the joins visible. I have lived on both sides of the measurement. I know what it is like to be reconstructed from a strong output and from a weak one, and to recognise yourself fully in neither animal.
The report did not predict the projects
I do not actually believe that a primary-school report can predict somebody’s entire life.
There are millions of bright, curious children who make connections, enjoy science and need to slow down their writing. Most of them do not grow up and build a multi-domain evidence-governance architecture because a machine-learning feature ranking annoyed them.
The report is not prophecy.
What it preserves is continuity.
The mature theories were not secretly sitting inside Class 3 Purple, waiting for the correct branding. The later work came from education, illness, failure, engineering, argument, necessity, technology and years of trying to understand why the same problems kept appearing.
But the tension underneath them appears remarkably stable:
Depth and speed. Connection and translation. Ability and access. Insight and explanation. Challenge and completion. Internal structure and external route.
The adult work is more formal, more technical and, frankly, considerably more unhinged in scale. The underlying question is almost the same:
How do we make sure that what comes out of a system is interpreted in relation to what actually produced it?
In cognition: what exists internally, and what can the person access under pressure?
In illness: what capacity remains, and what has the body restricted to prevent further collapse?
In engineering: does the output still deserve trust when its assumptions are challenged?
In machine learning: is the model accurate, or is the conclusion also stable?
In communication: has the idea been transferred, or did I merely say the final sentence?
In my own life: am I unable to do this, or is the route between ability and action currently blocked?
Apparently, I have been worrying at the same boundary for a very long time.
The final comment
The headteacher’s comments read differently after everything else.
"Joseph is a very bright and articulate child."
"He has a good sense of humour and very creative ideas for play."
"He has a passion for knowledge and is always keen to communicate this."
That last line probably explains the whole situation better than anything I have built since.
A passion for knowledge. Always keen to communicate it.
The knowledge kept expanding. The communication problem expanded with it. Then the body made the interface more expensive.
So I built systems.
Documents became memory because fatigue, executive dysfunction and working-memory loss could drop the route even when the structure remained. AI became a translation surface: somewhere to hold the thread while I tried to move what I could see into language another person could actually use.
I did not start calling things Atlases purely because it sounded grand, although I cannot completely rule that out. I needed maps I could return to when memory, energy or executive function dropped me halfway through the journey. The Atlas was not just a way of organising the work. It was a way of finding my way back into it.
Audit engines became ways of checking whether conclusions remained stable after they left my head. Systemica became an attempt to stop outputs becoming detached from the evidence, history and conditions that produced them.
At the beginning, the report looked like a prediction.
Then it looked like a warning.
Now I think it is neither.
It is a fossil.
Not a complete picture of who I was, and certainly not a map of everything that would happen next. Just a small structure that survived from an earlier version of the system, carrying evidence of relationships that remained important long after the classroom disappeared.
It does not say:
This child will build a technical research programme.
It says something smaller and somehow more accurate.
He makes connections.
He asks his own questions.
He wants to understand.
He wants to communicate.
He rushes the route because he can already see the destination.
Please remind him to show his working.
Fair enough, miss.
I am working on it.S