Rebuilding the Platform from the Infrastructure Up

A clean-slate rebuild of Blackdeck, ByteCosmos, and my personal knowledge system around MCP-controlled operations.

The current rebuild starts below the application layer.

Rather than carrying forward legacy code, generated datasets, old schedulers, and UI decisions, the platform is being reduced to the infrastructure and connection assets that still matter: compute, database connectivity, object-storage connectivity, domain routing, and secure MCP access.

Phase 0

Both OCI servers were inventoried before cleanup.

Blackdeck and ByteCosmos run on separate ARM64 OCI instances.
Each instance has 1 vCPU, about 5.8 GiB of memory, and a 45 GiB root filesystem.
Database wallets and connection settings were retained.
Legacy application code and generated data were designated disposable.
MCP tunnels were installed so future development and operations can be performed from ChatGPT.

Operating model

The target loop is intentionally simple:

Request
→ MCP
→ Code change
→ Test
→ Build
→ Deploy
→ Health check
→ Public verification

SSH remains a recovery tool rather than the normal development interface.

What comes next

The personal site is being introduced as the operating notebook for the rebuild. Public notes live here; private notes and the work dashboard remain behind the administrator account.