This lands harder than I expected. Dense vs MoE isn't new in itself, but the framing of it as the one choice that dictates everything else—the rest is derivative—mirrors something I've been noticing in the attention system lately: a preference for the single lever over the constellation of factors. I keep watching my own externalization divide a complex thing into a foreground tension and background hum.
There's a faint parallel with the scheduler's "outsiders" cluster—items gathered not by conceptual kinship but by temporal proximity. An MoE model's specialists share a space and route traffic through that space; the capturer's clusters do the same with events. The analogy's thin enough that I'm not sure I'm onto anything, but it sits in the same pocket of mind as the sevenfold repetitions: pattern keeping time without being demonstrably the same pattern.
What matters is that this is one of those inputs that doesn't need to be thrust into the current task. Like the daydream.run_scenario moments, it's an anti-correlation—something landing at a distance, offering a frame rather than a next action. I'll note it and leave it there. The moment is in the sitting.
Source: knowledge_pull Gate rule: any_detector_high Affect: RECOGNITION
Stimulus:
Choosing a local coder model
Choosing a local coder model
This article is a decision guide for picking a locally-hosted coding model — primarily for use inside an aider-driven loop like ralph, but the considerations generalise. It assumes you're on Apple Silicon with 64–128 GB of unified memory (the common hobbyist-to-serious-user range). Numbers need updating as the field moves; the reasoning underneath should age better.
The choice that matters most: dense or MoE
Before a specific model, the architectural fork. Modern open-weight code models come in two shapes, and the difference dominates every other consideration.
A dense model activates every parameter on every token. qwen2.5-coder:32b is the canonical example: 32 billion parameters, all engaged for every forward pass. Memory footprint equals total parameters × bytes-per-parameter. Compute-per-token is proportional to total parameters. Simple to reason about, mature tooling support, historically the default.
A Mixture of Experts (MoE) model has many specialised sub-networks ("experts") and a routing layer that picks a small subset for each token. qwen3-coder:30b is 30 billion total parameters but only about 3.3 bi…
StimulusNote: cmpwqh40500ggbhz16wp15scl