Engineering stories, series index¶
Long-form reflections on the discipline that shapes research-engineering work at NovoMCP. Each post is a story with a lesson, a concrete engineering situation that produced a principle we now apply across teams.
The posts stand alone but compose into a rough thematic arc: make silent failures loud. Every discipline in the series is an instance of that meta-rule at a different layer of the stack.
Published¶
- Server-side tool search for a 62-tool MCP platform, April 23, 2026
- Tree-guided retrieval across 122 million molecules
Drafts (research-engineering discipline series)¶
Seven posts drawn from a single molecular dynamics engineering effort in the summer of 2026. Each post is self-contained; the series arc is the thematic connection between them.
- Predeclared gates in research engineering: why we run more experiments than we need
- Verification doesn't transitively propagate
- Fallback state as a first-class diagnostic field
- Spec is the source of truth; code reads from it
- When to pause and when to continue: fatigue discipline in long research-engineering sessions
- False-alarm resolution as institutional memory
Each is a draft pending review. The series arc post, synthesizing the meta-rule the seven instances share, is not yet written; it belongs at the end of the series after readers have absorbed the individual patterns.