| 1 | # AI-ML Continuity — Multi-Week Narrative |
| 2 | |
| 3 | ## Active Trend Arcs |
| 4 | |
| 5 | ### Agent Skills Economy (W21→W23+) |
| 6 | W21 surfaced early hints (chrisbanes/skills, vercel-labs/zero). W22 confirmed the |
| 7 | category with multiple independent skill-packaging repos alongside 140k+ ⭐ trending |
| 8 | anchors. W23 crossed a cultural boundary — first major Chinese-platform design skills |
| 9 | (Xiaohongshu, WeChat) at 1k-2k+ stars. The pattern: clustered movement across |
| 10 | independent teams is a stronger signal than any single repo launch. Geographic and |
| 11 | linguistic expansion now constitutes its own sub-trend distinct from English-developer |
| 12 | workflow skills. |
| 13 | |
| 14 | ### Coordinated Manipulation Patterns (W22→W23+) |
| 15 | W22 saw the most concentrated coordinated star-farming campaign on record — game-cheat, |
| 16 | software-unlock, and AI-branded repos landing at 421-429 stars with zero forks in |
| 17 | minutes. W23 shifted the vector: fork-inflation via polymarket bot repos with fork |
| 18 | counts 10-20x their star count and keyword-stuffed descriptions. Same coordinated |
| 19 | intent, rotating metric vectors. Noise classification must track both creation-timestamp |
| 20 | clustering and fork/star ratio anomalies. |
| 21 | |
| 22 | ### Press-Developer Convergence (W22→W23+) |
| 23 | W22: TechCrunch inflated-AI-ARR coverage ran alongside GitHub star inflation — a |
| 24 | meta-convergence of metrics manipulation narratives. W23: Copilot token-billing backlash |
| 25 | (TechCrunch, 2026-05-30) produced same-week developer response in self-hosted AI |
| 26 | workspace repos — strongest press-developer convergence observed. Platform |
| 27 | billing/policy changes are now a confirmed leading indicator for self-hosting repo |
| 28 | spikes. |
| 29 | |
| 30 | ## Structural Insights |
| 31 | |
| 32 | - **Analysis contract**: The analyzer output is a superset of published frontmatter plus |
| 33 | quality gate — one artifact serves editorial and pipeline needs. |
| 34 | - **Press as correlation, not content**: External coverage is a signal layer explaining |
| 35 | the gap between narrative and developer traction, not material to repackage. |
| 36 | - **Multi-source press strategy**: Compact correlation artifact (≤8k tokens) with ranked |
| 37 | items, strong/weak correlation tiers, divergences, and full citations. Raw dumps are |
| 38 | prohibited. |
| 39 | - **Prediction tracking**: `predictions` frontmatter (`{repo, direction, confidence, claim_type}`) |
| 40 | feeds hindsight scorecards for reskill loops. `claim_type` is one of `signal|noise|gap`. |
| 41 | |
| 42 | ## Forward Indicators to Watch |
| 43 | |
| 44 | - Skills economy: enterprise adoption, monetization signals, framework-level integrations |
| 45 | - Manipulation: new metric vectors beyond stars/forks (watch issues, discussions, sponsor badges) |
| 46 | - Self-hosting: whether billing-backlash repos sustain past the news cycle |
| 47 | - Map/reduce analysis: dry-run path (`analysis_path=map-reduce-dry-run`) under validation |