Skip to content
Brian Sithu
Go back

The brain gets better at what you actually do

Cover image for The brain gets better at what you actually do

Neuroplasticity is real, but it is narrow and it fades. What that means for how I learn new tools, and for what I let an agent do for me.

Brian Sithu

Written by

Brian Sithu

View profile

On this page

06 sections

Neuroplasticity gets sold as a promise. Your brain can rewire itself, so you can become anyone. The research says something smaller and more useful. The brain does change with practice. It changes in the parts you use, for the things you practice, and it drifts back when you stop.

That second version is the one I try to work with.

What the studies show

Three studies come up whenever this topic does, and they are worth reading in their own words rather than in a wellness blog’s.

London taxi drivers spend years memorizing the city’s streets. Maguire and colleagues scanned them and found a larger posterior hippocampus, a region tied to spatial memory, than in people who did not drive taxis. The size tracked how long they had been driving. It was a small group and a snapshot, so it cannot prove the job caused the change, but it fits.

Draganski and colleagues went further and ran an experiment. They taught adults to juggle over three months and scanned them before and after. Grey matter grew in areas that process visual motion. Three months after they stopped practicing, much of that growth had receded.

Walker and colleagues trained people on a finger-tapping sequence. Those who slept before the retest improved by about 20 percent. Those who were awake for the same twelve hours barely improved. Some of the learning happened after practice ended, during sleep.

None of this makes the brain infinitely flexible. It says the brain adapts to demand, specifically, and keeps the adaptation only while the demand lasts.

Narrow, not general

The specificity is the part most people skip.

A large review of brain-training games found that people get better at the games. Evidence that they get better at closely related tasks was weaker. Evidence that they get better at unrelated tasks, or at everyday thinking, was very limited.

I find that useful rather than disappointing. It means there is no shortcut skill that makes everything else easier. If I want to get better at reading unfamiliar codebases, I have to read unfamiliar codebases. If I want to get better at debugging distributed systems, puzzles will not do it. The practice has to look like the work.

How I use it

Learning a new tool, I try to use it on real work from the first day, not on a tutorial project I will throw away. Tutorials are a different task. They teach me to follow tutorials.

I spread practice out instead of cramming. A focused hour, then something else, then sleep, then another hour. It feels slower. Going by the sleep research, I suspect some of the work is happening in the gaps.

I keep doing the things I want to keep. The juggling result is the one I think about most. Skills I stop using do not stay where I left them.

Practice is not the whole story

I want to be honest about the limits here. A meta-analysis by Macnamara and colleagues found that deliberate practice explained about 12 percent of the differences in performance across studies. In games it was about 26 percent. In professions it was under 1 percent.

That last number is humbling, though part of it is how hard practice is to define and measure in a job. Still, practice is not destiny. Other things matter: what you already know, the problems you get exposed to, the people around you.

But practice is the part I control. I cannot choose my starting point. I can choose what I do with tomorrow morning.

The part that worries me

This connects to something I wrote about AI. If the brain gets better at what you actually do, then what you hand to an agent is what you stop practicing.

That is fine for the parts I do not need. I do not need to stay fluent at writing boilerplate. It is less fine for the parts that make up judgment: reading a stack trace, holding a system in my head, noticing that a design is wrong before it is built. If an agent does all of that for me, I expect those skills to fade the way the jugglers’ grey matter did.

So I still debug some things by hand. I read the diff before I accept it. I sketch the design before I ask for one. Not out of nostalgia. Because the curiosity that makes AI useful has to be kept in shape by something, and that something is doing the hard part myself, often enough.

What I take from it

The brain is not clay you can reshape at will. It is closer to a muscle that grows where it is loaded and shrinks where it is not.

Practice the real thing. Spread it out and sleep on it. Keep doing what you want to keep. And be deliberate about which skills you let a tool carry for you, because the ones you never exercise are the ones you lose.



Previous Post
Your agent should never see the key
Next Post
AI is an amplifier, not an equalizer