Thoughts on software architecture, intelligent engineering, and building DevNitro.
RSS FeedAI-assisted .NET development put Codex behind DevNitro’s full tool loop, generated working apps, and settled the UI foundation on Bootstrap in a .NET stack.
AI-assisted .NET development added a multi-database seam to GCT and automated versioned, obfuscated NuGet packages for DevNitro-generated apps in .NET.
AI-assisted .NET development connected DevNitro Core to its Docker Workspace Agent so the model could create, read, update, and delete app files in C#.
AI-assisted .NET development produced DevNitro’s first MCP tool call into a Docker container, proving the Core-to-Workspace Agent loop worked end to end.
AI-assisted .NET development focused on launching MCP servers, stabilizing integration tests, and switching from Claude to Codex CLI during the work now.
AI-assisted .NET development renamed CodeSlinger to DevNitro, moved the agent loop into Core, and produced a working CLI with permission prompts in .NET.
GPT-5 rebuilt a cost model and revived a .NET prototype during a week of AI-assisted .NET development, ending with a working CLI chat loop with Blazor.