Measured incidents, path-level storage research, and explicit limits. Every article separates what we observed from what we inferred.
We traced a 24 GB physical-footprint spike to a browser rollout reader retaining an almost 10 GB task log in 64 KiB chunks, then trying to concatenate it into one string.
Read ArticleThe .claude directory is not one giant cache. It mixes resumable sessions, project state, memory, settings, plugins, logs and a few exact rebuildable leaves.
Read ArticleSystem Data is a category, not a folder. AI models, build artifacts, package caches, logs and app runtime data can all land there — with very different cleanup risk.
Read Article