“What Did You Ask AI Recently?”
I couldn’t answer that question.
I ask it multiple times a day. Yet I couldn’t recall a single thing I asked today. I couldn’t remember what’s in my personal knowledge repository or why my current directory structure exists, or even how I manage it. The final blow was realizing that the reason I adopted that structure was simply because I had “offloaded research on AI-friendly directory structures entirely to AI”—I had no memory of why it was good or why I chose it.
Cognitive Offloading
I learned about something called cognitive offloading. Cognitive offloading refers to the act of externalizing cognitive processes—basically, relying on external aids like to-do lists or shopping memos. But lately, I’ve been externalizing not just to-do lists and shopping memos, but also online articles, search results, code snippets—absolutely everything. With AI added to the mix, the cost of externalization has effectively become zero.
Two Things AI Broke
As a result, two problems began to emerge:
- Processing information that is hard to externalize—like paper books—and outputting it became painful.
- Even the things I externalized left no trace in my mind.
Externalizing memories might have been acceptable, but when I outsourced even the organization of those externalized items to AI, I fell into a mindset of “I’ll just ask AI when I need it.” Because of that, I no longer grasp what’s in my notes or articles, nor even what I externalized. In the end, I feel it’s all wasted.
What Is True Quality?
The essence is not about externalizing and managing memories. Instead, I now believe what matters is:
How efficiently can I cram externalized information back into my own brain?
I feel we need to move from the phase of “using AI to the max” to the phase of “evolving alongside AI.” I thought, “I don’t need to remember things myself because AI can read and retrieve massive amounts of information.” But that made my brain empty—and then I wondered: would others find human value in someone like that? I’m not good at talking about myself, so writing this article is truly painful. Maybe that’s because I, as a person, am empty.
What I Will Do From Now
From now on, instead of building a harness (a system) for storing and organizing using AI agents, I want to build a harness for storing, chewing, and re-storing. By “chewing,” I mean summarizing in my own words and explaining. In short, I need to use my own brain more. This is still at the hypothesis stage; whether it works remains to be tested. I also plan to leave logs of my hypothesis testing and thought process.