By skyforbes 
For the past month and a half, I’ve been using ChatGPT almost like an external memory.
I realized I can’t remember everything I do perfectly — I get distracted easily, and trying to manually document things every day felt impossible. So instead, I started using my everyday chats with ChatGPT as a way to preserve memory.
Whenever I studied something, got stuck, or just wanted to clarify a concept, I’d ask ChatGPT.
Sometimes I’d also tell it what I was planning to do that day, what I was struggling with, or even what my bigger goals were. Most of it wasn’t deep — just regular, messy study stuff: questions about machine learning assignments, debugging code, clarifying notes from lectures, or organizing coursework.
At the end of each day (or whenever I remembered), I would copy that day’s chat into a Word document, turn it into a PDF, and later upload all those files into NotebookLM.
It wasn’t perfect — I missed some days — but it gave me a searchable record of what I was doing, thinking, and asking over time.
When I asked NotebookLM to make summaries or timelines like “what did I do this week?”, the results weren’t very accurate.
It would often include things that didn’t actually happen — for example, ChatGPT sometimes gave me ideas or templates to use later, like structured note formats or systems to prioritize my work (the “fire, volcano, slow burn” idea). But I never really used those consistently. They didn’t solve the actual problem, and I eventually gave up on them.
Similarly, just because I asked ChatGPT to explain something — like the history of operating systems — doesn’t mean I actually learned it completely. NotebookLM would summarize it as if I did.
There were also moments where it misinterpreted context — like assuming my college projects (for the MAD-1 course) were personal side projects I created to deepen my learning for the “Learning Everything Project.” That wasn’t true; they were just required coursework.
So, the summaries and overviews sometimes felt off — too polished, too confident, like they were trying to make my scattered reality into a coherent story.
Right now, I have about 1.5 months of entries. Maybe once I’ve added 6 months or more, it’ll start showing clearer patterns — how my focus shifts or what ideas keep recurring.
NotebookLM also generated some video and audio overviews. They were interesting, but they all had that “inspiring podcast” tone — like they were narrating a journey of struggle and triumph. It didn’t really sound like me.
Still, I wanted to share one of the blog drafts NotebookLM generated, called “Learning in the Open.” It captures, in its own way, what this experiment looks like from the outside — even if my version of learning still feels messy and unfinished.
Click here to read “Learning in the Open”