Machines

# Machines

AI Summary

Purpose:

• Tracks personal machines and local path conventions.

Key points:

• The same repo may live at different paths on different machines.
• Agents should identify the current machine before assuming paths.
• The wiki repo and shared career outputs live on Google Drive, synced across machines.
• Resume/portfolio PDFs go to the shared 내 드라이브/career/ folder, not

inside this wiki repo and not to machine-local paths like ~/hw/.

• This host's macOS username is james. A second Mac path under /Users/khw

remains recorded; never copy that absolute path onto this machine.

• Do not store secrets or credentials here.

Relevant when:

• Resolving repo paths.
• Setting up a new PC.
• Running CLI agents from different machines.

Do not read full document unless:

• The task depends on local machine paths.

Linked documents:

• ai/workspace/repos.md
• ai/workspace/active-context.md

Default Path Convention

Use the same path convention on every personal machine when possible:

~/wiki/llm-wiki
~/work/<repo-name>

Machines

Shared Output Folders (Google Drive)

The Google Drive root 내 드라이브/ is shared across machines. Verified paths:

Rules for shared outputs:

• Render career-facing PDFs into career/, not into machine-local paths like ~/hw/.
• Canonical sources stay in this wiki repo (human/resume/, human/portfolio/); career/ holds rendered PDFs only.
• Do not edit the same file from two machines at once; wait for Drive sync to finish before switching machines.

On a Mac whose username or Google Drive account-folder suffix differs, resolve a shared project path instead of copying an absolute path from another machine:

REMEMBER_REPO=$(find "$HOME/Library/CloudStorage" -maxdepth 4 -type d \
  -name 'remember-data-engineer-assignment-codex' \
  -print -quit)
printf '%s\n' "$REMEMBER_REPO"

If this prints nothing, confirm that Google Drive for desktop is signed in and that 내 드라이브/projects/ has finished syncing locally.

The lookup deliberately matches the ASCII repo name instead of the Korean path. This avoids macOS NFC/NFD filename-normalization differences between machines.

The Claude implementation of the same assignment lives under the career tree, where every path segment is Korean, so it carries an ASCII marker file for the same reason:

ASSIGNMENT_DIR=$(dirname "$(find "$HOME/Library/CloudStorage" \
  -name 'REMEMBER_CLAUDE_WORKDIR' -print -quit)")
printf '%s\n' "$ASSIGNMENT_DIR"
cd "$ASSIGNMENT_DIR" && bash resume-here.sh     # venv + tests + pipeline

resume-here.sh creates the virtualenv at ~/.venvs/remember-de-assignment, i.e. outside Google Drive. Never create a virtualenv inside Drive: it syncs thousands of files and its interpreter paths do not transfer between machines. The same rule applies to generated databases — warehouse/*.duckdb is regenerated by the script and is deleted before switching machines rather than synced.

Rules

• Do not commit secrets.
• Do not store API keys.
• Do not store private tokens.
• Use machine id instead of ambiguous names like main pc.