"How large is Sonomundi?"
The product core was measured on July 8, 2026 with cloc 2.09 on branch docs/preprod-v046-verification-20260708: 274,225 application-code lines across 1,589 files, excluding tests, docs and dependencies, including 136,791 PHP lines, 102,986 Vue lines and 29,366 TypeScript lines. Product tests represented 130,623 lines across 813 files. On July 16, 2026, the project has 3,320 commits on main. These volumes remain secondary metrics: maintainability is assessed through module boundaries, tests, migrations, API contracts, documentation and the ability to isolate a change.
"How much was generated by AI?"
The project assumes significant AI assistance, without a reliable percentage to announce. Changes go through isolated branches and worktrees, Docker/CI tests, pull-request auto-review processes and structured documentation that simplifies control. There is no auto-merge: I remain accountable for framing, architecture, trade-offs and the integration decision. I present Sonomundi as a long-term Laravel/Vue side project, not as code accepted without verification.
"How do you validate the code?"
In layers: PHPUnit back-end tests, Playwright UI tests, role-based access matrices, OpenAPI/Scramble for contracts, pull-request auto-review, validation documentation and a check of the demo path. The static inventory of main on July 16, 2026 counts 2,189 declared PHPUnit tests across 401 PHP files (1,687 test* methods and 502 #[Test] methods). This describes the codebase, not a successful full run; data sets may also change the number of executed cases. Functionally, the documented Playwright matrix shows 683 passed / 0 failed across unlogged, public, artist, label and admin roles. For a given task, I look for the most targeted validation possible, then broaden it if the change touches a shared contract.
"What proves the project is structured?"
There are migrations, API resources, role middlewares, jobs, tests, architecture documentation, OpenAPI and access mapping. These are markers of a structured product. Some surfaces remain frozen or partial.
"What would you redo if you had to industrialize the project?"
I would freeze a smaller MVP, remove off-track workstreams, better separate services that deserve distinct ownership, strengthen CI/CD, secrets management, observability, backups, monitoring and demo data. I would also treat economic flows and real audio as dedicated workstreams with stricter non-regression tests.
"How do you avoid the black-box effect of AI agents?"
Through explicit specs, separate worktrees, authorized scopes, required validations, pull-request auto-review processes and short handoff documentation. An agent produces a proposal, a diff or an analysis; it has no merge authority. The source of truth remains the repository, tests, documented contracts and the integration decision for which I am accountable.
"Could this side project interfere with a full-time role?"
No. Sonomundi is a long-term personal project developed outside working hours, with no rushed launch schedule. It can be paused, and I remain fully available to my employer. My goal is to automate tests, PR reviews, documentation updates, deployments and monitoring alerts as far as practical to minimize maintenance and interruptions.
"What is demonstrable and what is not?"
Demonstrable in screen sharing: /events, selected region, event/artist/track opening, OpenAPI, anonymized diagram and test results. The product also covers multi-role auth, public/authenticated/admin surfaces, API contracts and workers. The reader has nothing to run. Payment/tokens/settlement, infrastructure industrialization, mature production observability, autonomous AI agents and admin surfaces open to external users remain distinct workstreams.
"Why start with this path when there are more features?"
Because it gives a quick reading of the product core: event discovery and event/artist/track navigation. Other areas can then be detailed depending on the reading depth: roles, public/authenticated/admin surfaces, workers, API contracts, tests and AI workflow.
Conclusion Angle
The summary to keep in mind: "I use AI to accelerate a long-term side project, with pull-request auto-review, tests and validation documentation. I remain accountable for the frame, invariants and integration decision." This is the central point of the Sonomundi workflow.
This page is a presentation support. It complements the local project demonstration.