@N01ennn: I DO NOT UNDERSTAND HOW OBSIDIAN USERS ARE STILL LOOKING AT 2D GRAPHS WHEN A 3D VISUALIZATION SHOWS THE ACTUAL TOPOLOGY…

I DO NOT UNDERSTAND HOW OBSIDIAN USERS ARE STILL LOOKING AT 2D GRAPHS WHEN A 3D VISUALIZATION SHOWS THE ACTUAL TOPOLOGY OF HOW THEY THINK

00:07 the screen displays “Topologies of Thoughts” on the right. a 3D network floats in place, currently rendering centralized mode with one core node connected to everything else

the tool takes an Obsidian vault, converts every note into an embedding, and renders the vault as a rotating 3D thought network in real time. every semantic relationship visible. every isolated cluster exposed

three topologies exist. centralized (one dominant idea). decentralized (isolated silos). distributed (fully networked). categories on the map include Generative Media, Artificial Life, and Cognition. most of them do not talk to each other

the interesting part is not the render. it is that the user finally sees where their thinking has gaps they never knew existed

the article covers what happens next. loops that argue with the isolated clusters, force cross-domain analogies, and turn the visualization from a mirror into a diagnostic

save this before every Obsidian user runs the topology check and realizes their thinking was never as connected as they assumed

@N01ennn honestly, a 3D view sounds cool but it just makes my head spin. 2D keeps it simple for my brain to follow. Maybe I’m just old school

A dev just gave the cleanest 4-method map of RAG in 2026:

00:06 - the question every agent builder asks daily 01:21 - why chunk retrieval loses context on every document 04:01 - Filters: cheap, deterministic, exact 08:30 - SQL query construction verified against Excel 15:46 - Vector Search demoted to the last resort

this is not another “vector databases are dead” clip. Embeddings still work. The job is picking the right method for the question shape

Software 1.0 was manual queries. Software 2.0 was embedding everything into a vector database. Software 3.0 is choosing the retrieval method based on the question shape

Filters win on tabular data with clear rules. SQL wins on math and rankings. Full Context wins when order matters. Vector Search wins only when a human would need to find one specific paragraph

pick the method a human would pick. spreadsheet filter, pivot table, reading the whole doc, or Ctrl+F

vector search raised the floor

method selection raises the ceiling