Another low-key weekend day — one commit on the site and a small but useful bit of investigation into how Claude Code stores its project state.
hackerlabs.dev
The only commit was publishing Saturday's lab journal (lab-journal-2026-05-24.md) along with the recap hero image for the 23rd. Standard drop-the-markdown-in-content/posts/ routine, let the build pick it up. Nothing more to say there — Saturday was already a quiet day and the post for it was mostly an acknowledgement of that.
processflow
The more interesting thread today was poking around ~/.claude/ to figure out exactly how Claude Code persists project context. I'd been mildly curious whether switching between Claude.ai accounts would blow away project memory, and the answer turns out to be reassuring.
The .claude/ folder lives at ~/.claude/ under the macOS user (richardcurry), not under whichever Claude.ai account is currently logged in. That means session transcripts, MEMORY.md, and the per-project metadata file (project_processflow.md in this case) persist across login swaps.
To confirm, I checked the processflow session artifacts after a login change and everything was still intact: the b720267b-…jsonl transcript with all 1343 lines, all 40 user messages, and the full session history from 2026-05-20 through today. Nothing was rewritten or lost.
The other thing worth writing down: Claude Code auto-discovers project metadata by hashing the working directory path. So when I cd into /Users/richardcurry/projects/processflow, it computes the project key from that path and pulls in the matching memory, todos, and session history. That explains why the context "just appears" without me pointing at anything — and also means if I ever moved the project folder, I'd lose that linkage unless I migrated the metadata too.
Good thing to know before I accidentally rename a directory and wonder where all my context went.
A quiet day, but I'll take a small infrastructure discovery over nothing at all.