The stone that kept moving
Nystagmus, large print books, and a career built on reading everything and checking it against the source.
The rock
I was on a school camp. We walked around a sea cliff, and the path crossed a gap where you had to jump from a large rock onto a stepping stone. Everyone in front of me jumped. I froze.
I could not pin the stone down. It kept moving under my eyes, and the waves moving around it made it worse. A fixed point that every other kid could see as fixed was, for me, never still.
I have congenital nystagmus. My eyes have moved on their own since before I can remember. Text slides when I read. I sit with a slight head turn that people notice before they notice anything else. My vision has a ceiling no pair of glasses has ever lifted. There is no cure.
The books
At school I could not keep up with the reading at normal print size. So I spent my pocket money on large print books. They were made for the elderly. I bought them anyway, because they were the only way I could keep up.
That is where the habit started. If I could not take something in at a glance, I would take it in properly. Read it, understand it, get it done. My eyes could not pin a moving thing down. Structure could.
Why technology
I went into digital transformation because it was work I could do on my own terms. On a screen I could read, understand and get work done, and the screen did not care how my eyes moved.
I co-founded Wizard Computer Training in the 1990s. It grew into Australia's largest privately held training company.
IBM and Deep Blue
I fell into IBM Australia by accident.
While I was there I was asked to evaluate whether Lotus Domino could serve as a free-text indexing and retrieval engine for Deep Blue, the IBM computer that played Garry Kasparov. I found it could not be used.
It is a small part of a famous story, and the honest version is the useful one. Part of the job of anyone who works with data is to say when the tool in front of them will not do the work.
Analytics at scale
Over the next decades the work grew from making systems run to making data decide things.
At IAG I worked on claims management. The old process was adversarial: repairers quoted against each other and assessors argued them down. We replaced it with a scored network. Every repairer was scored on their record. When a car was in an accident, the best-scored repairer in that region was selected automatically, the car was sent to them, and the repair was approved automatically. I was told afterwards that it saved the insurer 15 per cent of its underwriting costs.
Since then I have designed data platforms for government, defence and industry: data lakes that hold whole organisations' records, frameworks for sharing sensitive data between agencies, and the architecture that lets analytics run at that scale.
Growing up alongside the machines
I have worked alongside computing as it grew up. When I started, a computer could only do what it was told, line by line. Then it could search. Then it could learn from data. Now it can read.
That last step is the one that matters to me. For most of my life the bottleneck was reading. I compensated with large print, with structure, with patience. Now a machine can read thousands of papers and hand me the structure. The thing I built my working life around, getting understanding out of text, is something I can finally scale.
I also took a doctorate in leadership and decision making. It is about cognitive bias in leaders: why capable people make poor decisions, and why they defend them. The hardest thing to see clearly was never a stone in the water. It is your own judgement.
Pointing it at my own eyes
I run an automated trading fund. The work is turning large, messy data into something structured enough to act on, and letting machines act on it.
One weekend I pointed the same tooling at my own condition. I wanted to know what the literature actually says, rather than what a one-paragraph patient information sheet says. I built a corpus of 3,200 nystagmus papers. I had 1,404 of them read one by one into a structured table. That produced 669 intervention findings across 383 interventions, with the negatives kept in. I kept finding cheap, off-patent drugs that had worked in one place and had never been trialled anywhere else.
I am not a clinician, and none of it is medical advice. But the field had never been pulled into one place. That was a database problem, and databases are the day job.
Giving it away
The same pattern now helps other people read. I built a research system for a friend starting a doctorate: it keeps every claim tied to its source, reads the new literature each morning, and looks for the evidence against his argument as hard as the evidence for it. It worked well enough that I have released it as open source, as Commons, so any doctoral student in any field can use it with whatever model they have.
The point
I still could not tell you where that stepping stone was. But I learnt that if you cannot trust what you see at a glance, you can still get to the truth. You read it all, you structure it, and you check it against the source. That has been my whole career. The machines have finally caught up.