Methodology
Sector Radar Method
Denominator: full corpus, N=32, and in-cluster centroid sample, N=20. Source: accepted MCP Memory Servers report, centroid, calibration, and visual-data artifacts. Reader meaning: this page explains how to interpret public labels, samples, limitations, and correction routes.
Principle
Sector Radar asks what a landscape looks like. It does not ask whether a repo threatens a reference product, and it does not claim AI authorship. It maps pattern convergence, differentiation, and category boundaries.
Two-Pass Denominator
Pass 1 surveys the full corpus for candidate traits and category leakage. Pass 2 defines the centroid from the in-cluster subset only. This prevents adjacent infrastructure from distorting the default MCP memory server pattern.
The full corpus for this report contains 32 repositories. The centroid is defined from the 20 repositories whose primary product is an MCP memory server. Adjacent infrastructure remains visible in the report because it is part of the discovery experience for builders, but it does not define the default MCP memory server.
Corpus Boundary
An in-cluster MCP memory server has MCP server behavior as its primary product and persists AI coding session context across interactions. Adjacent infrastructure, reference implementations, and edge cases are retained in the public report where they affect discovery, but they do not define the centroid denominator.
Scoring Discipline
Scores are evidence summaries, not universal quality judgments. A high Centroid Match score means a project resembles the sector default. It does not mean the project is better. A high Differentiation score means the project brings a distinct pattern beyond the default. It does not mean the project is safer, more mature, or more useful for every user.
Adoption and repository activity are time-sensitive. All scores are frozen at the 2026-05-10 evidence snapshot so the report can be reviewed against a stable record rather than a moving target.
Limitations
| Limitation | How this report handles it |
|---|---|
| Search corpus bias | The corpus reflects repositories visible through MCP memory-related discovery paths. It may miss private, newly published, renamed, or poorly indexed projects. |
| Small sector size | The report uses explicit denominators and avoids claiming that the MCP memory server pattern generalizes to all AI memory tooling. |
| Time-sensitive evidence | Adoption, commits, and documentation may change. The report freezes evidence at a stated snapshot date and invites corrections. |
| Subjective interpretation | Some signals require judgment. Diversum publishes the scoring rules, role derivation, and correction path so disagreements can be inspected rather than hidden. |
| Public reputational risk | Public language describes sector position and evidence patterns. It does not claim AI authorship or moral failure by maintainers. |
Anticipated Objections
"Is this just an AI detector?"
No. Sector Radar does not claim to detect whether a project was written by AI. It measures pattern convergence: repeated framing, architecture, dependency, maintenance, and differentiation signals across a sector.
"Does high convergence mean a project is bad?"
No. High convergence can indicate a legitimate standard, a useful default, or a crowded attractor. The interpretation depends on Human-Edge, Adoption Depth, Integrity, and Differentiation.
"Are repo labels fair to maintainers?"
The labels are descriptive and evidence-bound. "Inactive Scaffold" means a project is structurally present in the sector but shows little sustained engineering evidence at the snapshot date. It is not a judgment of maintainer intent.
"Why include Rekindle if it is built by the authors?"
Excluding Rekindle would make the sector map less complete and would hide an obvious conflict. The report discloses the conflict, labels Rekindle wherever it appears, keeps its low Adoption score intact, and publishes enough methodology for readers to challenge the scoring.
"Why publish individual repo names at all?"
Sector claims need examples and traceable evidence. The report uses repo names to support the map, but it frames the analysis around sector structure rather than surprise public audits of individual maintainers.
Public Language
Public role labels are descriptive: Category Leader, Differentiated Niche, Strong Default, Baseline Implementation, Inactive Scaffold, Adjacent Infrastructure, Reference Implementation, and Edge Case. Internal competitive labels do not appear in the public report.
Source Artifacts
The methodology page is grounded in Reports/SECTOR-RADAR-MCP-MEMORY-SERVERS.md, Manual-audit/CENTROID-MCP-MEMORY-32.md, Manual-audit/CALIBRATION-TABLE-RADAR.md, Reports/VISUAL-DATA.md, and Manual-audit/AUDIT-TEMPLATE-SECTOR-RADAR-v1.md.
Corrections
If you maintain a project assessed in this report and believe a claim or score is wrong, contact hello@diversum.dev. Include the project, report section, claim or score in question, and evidence for correction.