Human Memory, Struggle & AI-Assisted Coding

Since I can remember, I've had some kind of memory problem. At least since I was 7 years old. I was constantly forgetting to pick up my stuff from school, missing assignment deadlines, and zoning out mid-conversation. If I had to semantically analyze the feedback I've received in life, the most common phrase would probably be: "Your memory sucks."

You can imagine how that plays out — in school, in friendships, and especially later when I became a software engineer, working on complex systems and even managing teams. Memory matters.

So how did I make it work?

Over time, I learned that memory isn't entirely out of our control. Sure, part of it is genetic and part environmental, but a lot of it comes down to skills. There's solid science on this: memory is something you can train.

🧠 Harvard Health – Train your brain: "You can train your brain to improve your memory, attention, and even your mood."

🧠 Harvard GSE – Think hard and dig deep: "Depth of thinking enhances memory retention."

I started absorbing every trick I could find — books, podcasts, YouTube rabbit holes. But one insight hit harder than all:
Memorizing something usually requires struggle.

Learning ≠ Easy Consumption

Let me give you an example: say you're learning piano. You find a great instructor and start lessons. Things move slowly. You hit the wrong keys again and again. But with each mistake, your brain adjusts. You fight through discomfort, unlock a level, and finally — it sticks.

That struggle is the learning. Your brain builds new memory blocks through friction.

So eventually, I trained myself to lean into struggle. At the micro level — like when I catch my brain drifting to weekend mountain biking plans during a technical meeting — I pull myself back. I know if I don't fight for focus, my brain won't register the details I need. And at the macro level — I accept that building memory is like building muscle. You can't skip the reps.

Then AI Entered the Chat

Like many engineers, I got deep into AI-assisted coding starting in 2022.

GitHub Copilot, ChatGPT, Claude, Cursor, Gemini — I was hooked. You could build functional prototypes overnight. It was mind-blowing.

In three days, I built GenShell, a CLI tool in Golang — a language I don't even normally use — and submitted it to a Google hackathon. Wild times.

But when the fun turned serious, things got messy.

The Production Phase Reality Check

I noticed a pattern: I could build POCs super fast. v1 was up, running, and I was pumped. But when it came time for iteration, debugging, or production stability — I felt lost.

Cursor was showing me the exact edits I had applied. I had reviewed them. I had tested the code. But in a moment of pressure — something broke in production — and I couldn't remember why we had made a certain architectural choice.

Worse: I had missed obvious things in code review.

The old feeling came back — that frustration, that childhood stress of "Why can't I remember this?"

But Neuroscience Says... No Surprise

There's a reason for this. When you're not actively engaged in the process — not struggling, not reflecting, not deeply encoding the why — your brain doesn't form lasting memory.

🧪 PMC – Exercise and cognitive function: "Physical exercise supports memory, attention, and problem-solving."

This applies to reading, too. If you just scan documentation with your eyes, don't expect much. To actually learn, you have to pause, reflect, summarize — you need active processing.

And AI tools — unless used carefully — remove that friction. They solve problems too fast. You stay passive.

With AI, My Memory Got Worse

I started making more obvious mistakes. And when they surfaced, I couldn't explain why I had done things a certain way. I had used the tools correctly. But I hadn't engaged enough with the why of my decisions.

I don't think AI is inherently harmful to cognition. But it definitely reduces the need for challenge. And that concerns me.

There's growing research about designing AI to augment rather than replace:

🤖 Stanford HAI – A Human-Centered Approach to the AI Revolution: “AI should augment human capability—not substitute it entirely.”

So I Changed My Habits

Here are some personal rules I follow now when coding with AI assistance:

1. For core services, I write it myself.
   Or at least define the schema, types, and comments before using AI to fill in the rest.

2. I make code self-explanatory.
   AI tools tend to "over-generate." That extra fluff leads to technical debt fast. Keep it clean.

3. I go offline sometimes.
   For certain tasks, I code old-school. No suggestions. No autocomplete. Just me and the editor.

4. No multitasking. Ever.
   Context switching kills focus — and memory. Think twice before you Slack mid-debug.

   📊 APA – Multitasking: “Multitasking negatively affects learning and productivity.”

5. I move more — I especially love racket sports.
   It sounds random, but aerobic exercise has a proven boost on memory and brain health.

   🎾 Harvard Health – Improving Memory: “Staying physically and mentally active helps maintain a healthy brain.”

An Open Question to Engineering Leaders

Here’s what I'm wondering now:

Are organizations seeing longer "time to resolution" for high‑stakes bugs? Not the easy ones AI can catch — but the messy, ambiguous, nobody-remembers‑why‑we‑did‑it‑this‑way ones.

I'd love to hear from other technical leaders. Are we challenging ourselves enough? Are our brains still in the loop when things get serious?

---

Final Thought

I'm not anti‑AI.
I'm just pro‑awareness.

There's power in letting AI handle the repetitive parts. But the moment we lose friction entirely, we lose something vital: memory, reflection, and the struggle that makes us grow.

Let’s be hybrid. Let’s stay human in the loop.