Commits
The development journal gets its own repository
Living documentation and journal are versioned in wb-dev-journal: snapshot of
Big projects rarely change with a single blow. They change through small, well-made decisions. Today brings no new code, but an organizational decision that sets how we'll work from here on: the documentation and this very journal now have their own repository.
The problem: living documents without a safety net
With the workspace —the working folder where AI agents and I edit— reorganized, the living documentation and this journal had stopped living inside any git repository. Translation: nobody was keeping copies of its history. An accidental deletion, a broken disk, and goodbye to the record of how each decision was made.
The solution fits in one line: a new private repository, wb-dev-journal, with everything inside. The development journal, the project's living documentation, the historical archive of obsolete material, and the public excerpt of the journal that feeds the website. The working folder remains the place where editing happens; the repository, where it's committed and saved.
Arrival snapshot: where the project stands
The second entry of the day is a snapshot of the state at the close of October's sessions, and serves as a summary of this stretch:
- Security audit: P0, P1, and P2 blocks closed and applied in production.
- Core releases up to v0.1.23, including the batch selection and deletion of media requested by the owner.
- Self-installing updater with signed command, verified on the real client.
- The product website is already on the new circuit: its own repository and image, restored data, live analytics, and the menu updated with the panel button.
And the live pending items, which is healthy to name: applying the two latest releases to the panels, retiring the old deployment once its observation period ends, and the becoach customization, which will take its own session.
Why this matters
It may look like bureaucracy, but it's the opposite. A development journal is only useful if it's where it can survive. Now every entry is versioned, with its full history of changes, and the AI agents that work with me have a clear entry point: read the living documentation before touching anything, and commit to the repo when finished.
It's the same idea that runs through the whole project: that the system explains itself. Now its history does too.
The previous entry was quite an adventure: a rollback in production and the mystery of the neighboring postgres. And if you've arrived here from the journal's index, the first entry explains how it all began.