Your brain hasn't changed since you were a registrar

When I show a senior colleague one of the cognitive aids I've built, there's one reaction I get more than any other. They turn the phone over in their hand, nod, and say something like: “this would be great for a registrar. I wish I'd had this when I was training.”

It's meant as a compliment. I take it as one. And it is, word for word, the problem I'm trying to solve.

Because buried in that sentence is an assumption so deeply held that we never say it out loud: that tools like this are something you grow out of. That somewhere between registrar and consultant, the brain acquires capacity it didn't have before. That expertise expands the hardware.

It doesn't. Working memory doesn't get bigger with seniority.¹ The structure of the human brain doesn't change because you've passed your fellowship exam. A consultant at 0300 can hold roughly the same number of things in conscious attention as the registrar standing next to them. Which is to say, not many. What changes with seniority isn't the hardware. It's the expectation that you've somehow outgrown limits that every human brain shares.

---

Where the assumption comes from

We build it in early. From the first year of medical school, exams reward unaided recall, and the doctor who can quote the trial, the dose, and the guideline from memory is treated as the model of what a clinician should become. We celebrate unaided recall as a key marker of expertise. That's bizarre when you stop to think about it. Bizarre because recall is precisely the cognitive function we know to be fragile. It degrades with fatigue, interruption, stress, and time of day. Which is to say, under exactly the conditions in which medicine is practised. We have organised the status hierarchy of the profession around the brain's least reliable feature.

---

The line we've already crossed

Here's the strange part: we've already conceded the principle. Nobody expects you to carry amikacin dosing in your head, or the titration table for a bivalirudin infusion. The number lives in the guideline, and you look it up when you need it. You're expected to know the pharmacology cold: why the aminoglycoside needs a level, what renal failure does to clearance. But the number itself? We externalised that years ago, and nobody calls it deskilling.

So the principle is settled: understanding belongs in the head, arbitrary detail belongs in the system. The only question is where the line sits, and we've drawn it in strange places. The arrest algorithm belongs in your head, and you should know it cold. But the ECMO cutoff: 60 minutes of CPR, or 75? The apnoea threshold: a rise of 20, or 25? These are arbitrary numbers, and no amount of seniority makes them stick.

Memory has a third job too, one we barely talk about: keeping your place. You arrive at the arrest, take a thirty-second handover, sort the rhythm, then turn your mind to reversible causes. Hypoxia, being addressed. Potassium, pending on the gas. Two items in, the next rhythm check is called, or someone asks about access, or the family arrives at the door. The thread is gone, and you come back to it with questions: what have I excluded, what's still pending, what was I about to ask for? The algorithm was never the hard part. The hard part is holding your place in it while the room keeps taking it from you, and that's a working memory problem twenty years of experience barely touches.

---

The exam that proves the point

Here's my favourite example. As part of the intensive care fellowship exam, you are expected to be able to perform brain death testing from memory.

Consider what brain death testing actually is. It is not an emergency procedure. There is no time pressure. The guideline is written down, freely available on the ANZICS website.² The patient is not getting any more dead. There is no situation in clinical practice where the right thing to do is to perform it from memory rather than with the protocol open in front of you. And yet that is exactly what the exam demands. And the message it sends is that a real expert shouldn't need to look it up.

Now consider what actually needs a doctor in that room. Not the checklist. The thresholds and exclusions are in the guideline, and the guideline does not get tired or skip a line. What needs a doctor is everything around it. Talking to the family. Sitting with a mother who is looking at a chest that still rises and falls, trying to understand how this can possibly be death. Holding that space, then gently starting the conversation about organ donation. Those are the things only a clinician can do.

The cruel irony is that the culture of unaided recall doesn't just fail to help with the human work. It actively taxes it.

---

What experience actually buys you

This is usually where someone says: but experience matters. It does. Enormously. Experienced clinicians chunk.⁴ The rhythm strip that costs a junior doctor five slots of working memory costs a consultant one, because twenty years of pattern exposure has compressed it into a single recognisable object. That compression is real, and it's most of why experts are fast.

But notice what chunking is: compression of the familiar. Arbitrary numbers and one-off details never become familiar, so they never compress. Chunking lets the expert spend the same handful of slots more wisely. It doesn't add an extra slot. The hardware is unchanged, and on a bad night, under interruption and fatigue, the consultant's working memory fills just like everyone else's. It just fills with more expensive things.

A 2025 study in BMJ Quality & Safety analysed three hundred serious incident investigations from across three Australian states.³ Of the contributing factors investigators identified, 47 per cent sat at the level of the individual people involved. Of the recommendations that followed, only 6 per cent were rated strong, the kind of structural change that doesn't depend on humans behaving better. The rest were overwhelmingly policy reviews, training, and education. We keep responding to the limits of human cognition by asking the humans to cognate harder. We keep reaching for the weakest interventions on the list.

This is the thinking behind the tools I build. CodeClock keeps time and holds your progress through the arrest: what you've addressed, what's still outstanding, what's next. When the room pulls your attention away, none of it is lost. IntubAid carries the airway risks you've identified forward into a prompt, so they get verbalised to the team rather than left in one person's head. BloodGasPro shows its working line by line, and when a gas doesn't add up it tells you how much is still unexplained and makes you go looking before it lets you close. None of them decides anything. They hold what doesn't need a doctor and force the step that does, so your working memory is free for the judgement only you can bring. The dose lookup principle, applied above the line we currently draw.

So when a senior colleague tells me a cognitive aid would be “great for a registrar”, here's what I want to say, and am slowly learning to say out loud: it would be great for you. Not because you're not good enough. Because you're human, and the conditions are hostile, and the brain you're using tonight is the same one you had as a registrar. The only thing that's changed is that nobody expects you to admit it.

---