AI Tooling Convergence Field Brief

What This Is

This is a public field brief built from accepted Diversum artifacts only. It summarizes patterns across the first three published AI tooling Sector Radar reports and keeps the source trail visible for readers who want to inspect the evidence.

What This Is Not

This page does not change scores, corpus membership, cluster roles, centroid construction, calibration interpretation, or methodology. It is not a claim that any project was AI-generated, not a public leaderboard, and not a repo-specific buying recommendation.

What Diversum Has Mapped

Diversum has published three Sector Radar reports covering 78 public repositories. Each report freezes an evidence snapshot, defines a sector boundary, derives a synthetic centroid from accepted artifacts, and shows where public repositories cluster, diverge, or sit at the boundary.

Recurring Patterns

1. A Practical Default Appears Quickly

Each sector shows a visible default pattern. MCP Memory Servers center on local-first MCP servers that preserve coding-session context across turns. Chrome Extension Starters center on Manifest V3, TypeScript, React, Vite, popup/content/background surfaces, and hot-reload expectations. VS Code AI Coding Agent Extensions center on editor-native agents with workspace/file context, chat or agent surfaces, provider/model configuration, tool or process bridges, and marketplace-style distribution.

The useful finding is not that every project is identical. The useful finding is that builders can see the default they are entering before they spend weeks rediscovering it.

2. Differentiation Moves To The Edges

The center of a sector often converges faster than the edges. Chrome Extension Starters show this most sharply: the Differentiation Index is 1.79/5 while Convergence is 4.00/5. Once a builder chooses MV3 + TypeScript + React + Vite, many starter repositories become hard to distinguish.

MCP Memory Servers look different. The in-cluster Convergence Index is higher at 4.35/5, but Differentiation remains 3.1/5 because several projects pursue distinct memory architecture, retrieval, and workflow choices beneath similar public framing.

VS Code AI Coding Agent Extensions sit between those cases. The accepted report records a 3.55/5 Convergence Index and 3.05/5 Differentiation Index: a strong editor-agent default exists, but package surfaces still vary through protocol bridges, local models, Claude Code workflows, markdown chat, ACP/MCP clients, and multi-agent orchestration.

3. Discovery Boundaries Matter

All three reports show that a search result is not the same thing as a sector. MCP Memory Servers have the clearest discovery-boundary problem: 8 of 32 full-corpus rows are adjacent infrastructure rather than retained MCP memory servers. Chrome Extension Starters have a lower but still visible boundary issue around frameworks, plugins, archived projects, official samples, and finished extensions. VS Code AI Coding Agent Extensions include adjacent infrastructure rows where packages influence the ecosystem without being the same kind of editor-agent extension as the retained corpus.

For builders and operators, this means a sector map is valuable before any recommendation is made. It separates true peers from references, adjacent infrastructure, legacy artifacts, and edge cases.

4. Adoption Is Usually Uneven

The reports repeatedly show top-heavy adoption. MCP Memory Servers have high surface convergence, but only a small retained subset reaches high adoption depth. Chrome Extension Starters include older high-adoption projects alongside newer default-pattern projects, so historical stars need interpretation rather than blind reuse. VS Code AI Coding Agent Extensions are especially top-heavy: Cline, Continue, and Kilo carry much of the visible adoption while many long-tail repositories remain strategically informative but lightly adopted.

The reader should treat adoption as one evidence field, not the conclusion. High adoption can indicate maturity, history, visibility, or default fit; the calibration tables keep those signals separate.

5. Integrity And Viability Differ By Sector

Chrome Extension Starters show unusually strong viability: 21 of 24 repos show sustained engineering and no retained starter is classified as an inactive scaffold in the accepted report. That makes sense for a sector where a starter must produce runnable extension surfaces to stay useful.

MCP Memory Servers are younger and noisier. The accepted report records a 25% in-cluster scaffold rate and a 25% full-corpus category-leakage rate. This does not make the sector weak by itself; it means a reader needs help distinguishing structural presence from sustained implementation.

VS Code AI Coding Agent Extensions have no inactive scaffold rows in the accepted public report, but the sector remains uneven because adoption, agent depth, and protocol coverage concentrate around a few visible projects.

What Remains Uncertain

• Repository activity, archive state, stars, forks, marketplace adoption, and README claims can change after the snapshot date.
• Three sectors are enough to identify recurring public-analysis patterns, but not enough to claim universal laws for AI tooling.
• Sector boundaries remain judgment-bearing. Diversum publishes boundary rationale and correction routes so disagreements can be inspected.
• Public Sector Radar reports stay sector-level. Repo-specific strategic advice belongs in voluntary private audit work, not public surprise scorecards.
• Cross-sector pattern intelligence should remain source-backed and conservative until more sectors accumulate.

Source Packet

This brief uses accepted Diversum artifacts only and leaves scores, corpus membership, role assignments, centroid construction, calibration interpretation, and methodology unchanged. Found a stale repo, wrong source trace, or boundary classification issue? Send a correction to hello@diversum.dev with the repo name, displayed value, and replacement source.