Why Your Brain is Getting Weaker in the Age of ChatGPT

The brain is a "use it or lose it" organ with zero customer support. When something else thinks for you, parts of you quietly go offline. This is not a metaphor — it is a description of how the brain actually works. Cognitive neuroscience is depressingly unambiguous about this.

The good news is the same fact cuts the other way. The brain rebuilds itself when you load it. This post is about the three specific muscles that weaken in the AI era and the three small, stubborn exercises that rebuild them.

Muscle 1: Working memory

Working memory is your ability to hold several moving parts in your head at once. Reading a paragraph and remembering what it said. Writing a long email without losing the thread. Debugging across three files and keeping the whole picture in mind.

AI has quietly become our external working memory. You do not have to hold a whole system in your head anymore; you can paste it into a model. That is great for output and bad for capacity. Like any capacity, it shrinks when unused.

Muscle 2: Tolerance for confusion

Confusion is a feature, not a bug. It is the feeling you get when your brain is actively rebuilding its model of something. It is uncomfortable, and it is supposed to be — if it were comfortable, we would not do the rebuilding.

AI removes confusion by giving you immediate pseudo-understanding. "Oh I get it now" without the part where you actually get it. Over time your tolerance for sitting in confusion drops to near zero, and the learning that used to happen in that confusion stops happening entirely.

Muscle 3: Attention span

Every time you open an AI tab, you introduce a micro-break into whatever you were doing. Over a workday, those micro-breaks are larger than the actual lunch break. Your brain learns that sustained attention is optional. Next time sustained attention is required — a hard conversation, a long meeting, a difficult piece of writing — you cannot summon it.

The three exercises that work

I am not inventing these. I am reporting what actually works after trying many things that did not. None of them are magic. All of them are boring. That is the point — the boring ones are the ones that compound.

• Write one page longhand each morning, no edits, no pauses. This trains attention and first-draft courage
• Solve the problem before opening the AI tab, even badly. This protects working memory and confusion tolerance
• Read one long article a week and summarise it for a friend from memory. This trains both capacity and compression

Why these exercises and not others

Longhand writing forces you to think slower than you type. That slowness is where deeper connections happen. Your hand is a bottleneck on purpose.

Solving before prompting preserves the struggle where learning lives. Even if your solution is wrong, you have now built a map of the terrain, so when AI gives you the right solution, you can actually understand it.

Summarising from memory — not paraphrasing while looking — forces your brain to have actually absorbed the material. If you cannot summarise it, you did not really read it. You just moved your eyes over it.

A warning about the rebuild

It will feel slow. It will feel worse before it feels better. Your first week of longhand pages will produce garbage. Your first month of pre-prompt problem-solving will look slower than your AI-first colleagues. This is normal. You are rebuilding load tolerance, not doing a stunt. Load tolerance is always slow to rebuild.

By month three, you will notice you can sit with a hard problem for 45 minutes without reaching for a tab. That is a superpower. By month six, you will notice you are seeing things your AI-heavy colleagues are missing. That is a career.

Confusion is where thinking lives. Stop skipping past it to the nice clean answer.