A showcase of my configuration stack and preferences.
Each root folder owns a README.md capsule documenting what I use and why.
Capsules record durable preferences, design rationale, and high-value snippets
that would be hard or ambiguous for a strong model to infer. An LLM can use
these to reconstruct or update local config, adapting to the current
environment.
Include config files or snippets only when prose would not get a SOTA model to a working result within a few iterations. Strip system-identifying details before committing — hostnames, usernames, and local addresses are excluded by default. Benign details are harmful in aggregate: individually minor, they collectively form a system fingerprint. Use generic placeholders where specific values are needed for context.
Do not track generated config, credentials, logs, caches, sessions, private paths, machine inventories, package/plugin state, or lockfiles that belong to local runtime setup.
Machine-specific values (hostnames, ports, partition labels, etc.) live on the machine itself and are preserved by the system backup. This repo documents stack choices and conventions, not machine inventory.