Human Mind as a Tape Machine
4) Long-term knowledge
▲
│ (consolidate)
│
┌───────────────────────┐
2) Forget ◄──┤ Sliding Window │◄──── 1) Input tape (feed)
│ (working memory) │ ▲
└───────────┬───────────┘ │
│ │
▼ │
3) Generate new events ─────────┘
(feeds back into 1)The Brain Rot Loop:
Algorithms hack 3) → generate more 1) endlessly → filled window amplifies 3) → infinite loop.
How to hack back:
Reset your context window. Use 2) deliberately.
Activating 2) Forget
a) Kill rehearsal
Don’t replay it. Every replay pushes it back into 1). Label it, move on.
b) Shift context
Memory is location-bound. Change room, go outside. Move the read head off the tape.
c) Overwrite
Not passive “better content” (that’s still 1). Use active load: build, solve, move.
d) Sleep = GC
Brain consolidates last-in. Feed before bed = save the junk. Cut feed early = let GC run.
e) Break cue links
Phone = retrieval cue for all of it. Rearrange, change apps, change triggers.
Building 4) Long-term Knowledge
- is a graph database. Nodes of ideas, edges of context. Externalize it: Markdown, Notion, Obsidian, whatever works.
What matters:
- Structured — Unstructured notes are just 1) pretending to be 4). Structure forces processing.
- Progressional — Not a flat pile. A list that grows forward. First encounter → current understanding.
- Searchable — Can’t find it = don’t have it. Free your sliding window. Let the tool remember.
Key: 2) is not delete. It’s access control. You don’t erase the tape — you move the head and stop going back.