tj-pak — the-learning-machine % cd ..

% open the-learning-machine

The Learning Machine

Feed it a video; it builds my brain vault and writes me a personalised read every morning.

ai live

PROBLEM

I consume a lot to learn, videos, articles, podcasts, and almost none of it sticks. A week later it's gone, and two ideas I picked up a month apart never get connected because they live in different corners of my head. The watching feels like progress. It isn't.

WHAT I BUILT

A personal learning machine. I feed it a video, it pulls the real content out, files it into a growing brain vault, and finds the connections to everything I've fed it before. Every morning it reads the vault and writes me one fresh piece to learn from, built around what I'm working on that week, in Korean, the way I actually absorb.

HOW IT WORKS

One skill ingests a reference, extracts the full content out of a video so I don't have to watch it, and breaks it into atomic idea-notes. Each new note is cross-linked against the whole vault, so old notes get more useful when a new one lands. A second skill reads the vault each morning and writes one long-form piece around the week's priority, ending in reflection prompts. I journal back, and that journal becomes new notes, which feeds the next read.

WHERE IT'S AT

Live and running daily. Amplify, the vault, and the morning read all work and I use them. A V2 that mines the journals over time and grows its own links is deferred until I've run V1 for a while.

WHAT I LEARNED

Most learning tools are storage, you save the thing and never see it again. The value here is the opposite: it connects and resurfaces. The compounding is the whole point, every new note makes the old ones worth more. It's the first system I've built that gets more useful the longer I leave it running, instead of decaying.

STACK

Claude Code skills (amplify + morning-read) · Obsidian vault · caption/transcript extraction

The Learning Machine output
Private — runs daily.

% cd ..