The Vibe Father
GitHub is retiring the Spark experience on github.com. New users and new project creation stop August 4, and the product retires August 31, 2026. GitHub says existing deployed Spark apps will continue to run, but the authoring surface and GitHub-provided AI infrastructure behind Spark will not remain available after retirement.
This is not a reason to panic-delete a working application. It is a deadline to separate what belongs to your codebase from what belonged to the builder platform. The most important question is not “can the page still load?” It is “can we change, test, deploy, and operate it once the original builder is gone?”
Inventory before exporting
| Surface | What to capture |
|---|---|
| Source | Repository, branches, build instructions, environment variable names, and dependency lockfiles. |
| Deployment | Current URL, hosting configuration, domains, data stores, logs, and rollback path. |
| AI features | Every call to Spark-provided inference or the retired GitHub Models surface, plus prompts and tool schemas. |
| Data and access | Exports, backups, ownership, user roles, and any sensitive data exposed in the published app. |
Export is the start, not the migration
Export the application early, then prove the exported project can build outside Spark. Run the test suite, start it locally, create a clean deployment, and exercise the highest-value user flow. An export that compiles but still assumes a platform-managed model, database, secret, or deployment feature is a snapshot—not an operable application.
For AI-backed features, write down the provider contract before replacing it, model identifier, request shape, streaming behavior, authentication, rate limits, content safety expectations, error behavior, cost limits, and fallback. Move those calls behind a small adapter so the rest of the app does not become permanently tied to a single vendor response format.
Migration checklist
- Export every active Spark project and make a second, independent backup.
- Install dependencies in a clean environment and record the exact build and test commands.
- Deploy one non-production copy under infrastructure you control.
- Replace any retired GitHub-provided inference with an approved provider or a feature flag that fails safely.
- Move secrets to the new host’s secret manager, never copy them into the repository during the export.
- Verify data ownership and backups, especially for a published app’s shared store.
- Run a production-like smoke test and document how to roll back.
- Complete the migration before August 31 rather than discovering a hidden dependency after the editor disappears.
Trend signal
The Kimi WebBridge Google Trends snapshot from August 4 showed continuing relative interest in broad AI-coding terms. Those values are not a migration priority score. They are a reminder that app builders and coding agents change quickly, so a durable harness keeps source, data, deployment, and model access portable.
Bottom line
GitHub Spark’s retirement does not automatically take down deployed apps, but it does remove the original place to create and edit them. Export early, test outside the platform, and replace any managed AI dependency with an explicit, observable integration. Portability is what turns an AI-built prototype into software you can keep shipping.
