Why broca? On Aphasia, LLMs, and Finding Language

It was 1861, a 51-year-old patient named Louis Victor Leborgne was admitted to a hospital in Paris with gangrene. At that time, gangrene was nothing unusual; his history, however, was quite interesting. He had lost the ability to speak for over 20 years. He can understand spoken language perfectly, but was unable to say anything beyond “tan”, which eventually became his nickname “Patient Tan”. The only way he can communicate was through sign gestures.

Eventually, he succumbed to gangrene and died at the hospital. The physician treating him performed an extensive autopsy (apparenly this was normal back then) and discovered pathology on the left hemispere of his brain, “a cavity filled with serous fluid in the left perisylvian region the size of a chicken’s egg”. It was concluded that the lesion in the frontal lobe of the left hemisphere was responsible for the patient’s speech disturbance. The physician treating him was Paul Broca, a French surgeon and anthropologist whose name was immortalized as the name of the neuroanatomic region responsible for speech production.

Digital aphasia

I am a man of ideas; perhaps too many ideas. When I started my private practice at 2019, the thought of maintaining conventional medical record with stacks of wrinkled pages disgust me. Back then, the term “electronic medical record” only existed on journals and seminars, rarely found on the wild. I began exploring the rabbit hole of web design only to hit a wall. I had the picture ready; I can showcase the design on an excel worksheet, and even the entire database structure. But I had no way of articulating it.

I had aphasia. Digital aphasia.

Eventually, I gave up and hand over the design to a friend. BrocaEMR was then built in less then a month.

On Large Language Models

Over the years, I tried learning to code by myself. I was introduced to Filament, which was quite intuitive to implement. But still, I spent hours and hours debugging error messages with google search.

LLM then take over by storm. The first prompt I used on chatgpt was how to make a python data analysis app; have no idea how to implement it but it gives me the impression that LLM is going to change the way we code. LLM was excellent on debugging codes, where I can paste the error code and get to the problem in no time.

Then it started. I built several small apps for my workplace; a mass file copying tool to search and copy thousands of pdf within minutes instead of days, mass request app to automatically generate pdf, or merge tens of thousands pdf files in an instant. Bear in mind that I still have no idea how those apps work.

Artificial intelligence have actually existed for quite some time. IBM’s Deep Blue have beaten chess world champion Gary Kasparov on 1997, or Watson that beat human on Jeopardy at 2011. However, it was unavailable to us commoners. Interacting with the AI simply means typing lines over lines of code on a supercomputer.

Large Language Model arrived with a different kind of intelligence. It does not require us to learn their language, they were trained on our languange instead. LLM somehow felt like a digital version of Broca’s Area; not to think by itself but to translate our ideas to lines of code.

Not a replacement for thinking

I should be clear about what I’m not saying. I’m not saying LLMs replace thinking. Broca’s area doesn’t think for you — it doesn’t decide what to say, doesn’t weigh the ethics of saying it, doesn’t know whether the thought is true or kind or worth expressing. It just provides the language.

LLMs are the same. They don’t have the ideas. They don’t know what’s worth building or why. They can’t replace the clinical judgment that comes from years at the bedside, or the philosophical instinct that tells you which question is the right one to ask. What they can do is bridge the gap — turn the thought you already have into the form it needs to take.

That bridging function is not trivial. Language is not trivial. The difference between having an idea and being able to express it is the difference between having a map in your head and being able to give someone directions. One helps only you. The other changes what other people can reach.

Why this name matters

I named this site broca.id because the name carries both meanings I care about. It’s a medical reference — a nod to neuroanatomy, to the clinical training that shaped how I think. And it’s a metaphor for what makes this moment in technology different from the ones before it.

The printing press made ideas reproducible. The internet made them distributable. LLMs make them expressible — they lower the cost of turning a thought into something that exists outside your head.

That’s what this site is. Ideas, given language. A blog that exists because the gap between what I meant and what I could make finally closed.

If you’ve ever had an idea you couldn’t express — in code, in words, in any medium — you understand why this matters. Welcome.