Not data. Not information. The structure that lets you act.
You spent the week building. You hit problems that felt technical but were actually about something deeper:
- two APIs describing the same thing with different names
- an agent that remembers everything but knows nothing
- a graph where you couldn't tell which edges to trust
- a model that found correlations but couldn't tell you what to do
These aren't bugs. They're the oldest problems in science. People have been working on them for three hundred years. The math is beautiful. The intuitions are wild. And they describe exactly what you were doing this week whether you knew it or not.
The load-bearing concepts. Each one grounded in the giant who formalized it.
- assembly — unstructured → structured. The universal pattern.
- belief — You have a graph. Why trust any edge in it?
- ontology — Naming things is the hardest problem.
- memory — Forgetting has a thermodynamic cost.
Napkin-sketch-able. Do them standing up with a drink in your hand.
- build a belief graph — Map something you know. Score it. Find the weakest link.
- resolve an ontology — Two schemas, same domain. Merge them. What did you lose?
- assemble from noise — Raw signal → structured knowledge. What's the graph?
Conversation fuel. No right answers.
From Vibe Coding Nights. No lectures. No networking. Just the thing.