Every connection in a reasoning graph should feel like a step, not a leap. We replaced our fibonacci spiral layout with uniform radial spacing — and made node size mean something.
When you look at a graph of connected thoughts, two things need to be instantly clear: what’s connected to what, and which connections matter most. Our original layout made neither easy.
The old engine used a fibonacci spiral to space nodes radially — golden ratio exponents, depth-dependent sizing, vote-based inflation. It produced beautiful organic shapes, but at a cost. The first two rings of nodes were crammed together while outer rings had too much space. Nodes with high vote counts ballooned in size, competing for attention with the actual structure. Following a chain of reasoning from one thought to the next required your eyes to jump unpredictable distances.
We tore that out and replaced it with something simpler.
Uniform ring spacing. Every depth ring is now exactly 220 pixels from the last. When you follow a chain from the birth node outward, each hop covers the same visual distance. Chains read like paths with consistent stepping stones. The spiral curve that gives each arm its organic shape is unchanged — only the distance between rings is now predictable.
Connection-density sizing. Nodes no longer grow when people vote on them. Instead, every node starts at the same base size (18px) and grows logarithmically with its connection count. A thought that five people connected to becomes visibly larger than a leaf node. The birth node starts bigger (28px base) because it’s the architectural center, but uses the same formula. Hub nodes — the ones people keep connecting to — earn their visual weight through structure, not popularity.
Vote magnitude moved to glow. Votes still matter, but they show up differently. Positive sentiment increases a node’s glow radius and brightness. Negative sentiment shifts the color through an amber-to-burnt-orange gradient by default, with a classic green-to-red palette available for readers who prefer it. Total engagement — votes in either direction — increases fill opacity, so lightly-voted nodes sit translucent and heavily debated ones feel solid. The information is still there. It just doesn’t distort the layout.
Adaptive ring expansion. When a node has many children crowding a single ring, the engine pushes that ring outward just enough to fit everyone. Downstream rings shift by the same amount, preserving the uniform gap. Uncrowded rings stay put. The graph only expands where it needs to.
Radius-based collision. The position-based dynamics constraint — the hard floor that prevents nodes from overlapping — now uses actual node radii instead of a depth-scaled constant. Every pair reserves a 14-pixel edge-to-edge gap on top of their own radii, so a hub pushed against a leaf claims proportionally more space than two leaves do. The constraint matches what you see.
The edge colors got a subtle update too — tree edges (the BFS backbone) stay cyan, lateral cross-chain connections shifted to a more saturated orange, and cross-component bridges moved to a slightly cooler green. The classification logic was already there; we just tuned the palette.
What didn’t change: the spiral curve per chain hop, the branch alternation at fork points, the golden-angle distribution for satellite components, the per-frame force simulation, and everything about how edges, votes, and layers work. This was a surgical replacement of two things — how far apart rings are, and how big nodes are — with minimal blast radius.
The result is a graph where structure is legible. You can follow a chain. You can spot where connections cluster. You can see which nodes people keep building on, because those nodes are physically bigger — not because someone upvoted them, but because the graph grew around them.
That’s the kind of signal we want Latent Organic to surface: not what’s popular, but what’s structurally load-bearing.
The update is live at graph.latentorganic.com.

AI-generated editorial illustration · TemperatureZero · April 13, 2026
Keep reading the signal
Get the Daily Signal — a concise briefing on what actually matters in AI and the systems around it.
Subscribe FreeContinue the archive