Trend Phase Chart
Weighted volume per theme over time. Bands: emerging · scaling · mature · declining.
Hype vs Substantiation
Same themes scored two ways. Gap between panels = hype index for that theme.
All Themes
| Theme | State | Volume | Velocity | Diversity | Hype risk | Confidence |
|---|---|---|---|---|---|---|
| No data — pipeline has not run yet. | ||||||
Active Themes
No themes yet
Themes appear after the trend pipeline runs.
Source-Class Coverage Heatmap
Rows = themes. Columns = source classes. High media, low primary = hype risk.
No heatmap data yet.
Source-Class Coverage
Items processed this week by source class. A balanced intake reduces single-source distortion.
No source data yet
Run the pipeline to populate source coverage.
All Sources
Full source list with item counts, active days, and top themes.
Cross-Class Confirmation
A claim only reaches high-confidence when supported by ≥ 2 distinct source classes. Items from a single class are flagged as unverified.
Cross-class confirmation breakdown will appear here once data is available.
Key Insights
Derived from trend state, hype risk, and cross-source diversity. Signals backed by ≥ 2 source classes are marked confirmed.
No insights yet
Insights appear once the pipeline has run and produced trend data.
Pipeline Log
Last build run: what was loaded, how much was AI-enriched, per-file breakdown.
No log data yet.
All Processed Items
Every item loaded from data/processed/. Filter by source or theme.
Loading items…
About this system
Source triangulation: Items are classified into five source classes (primary, operator, practitioner, media, market) and scored on five credibility axes: proximity to the claim, incentive alignment, reproducibility, adoption signal, and time decay.
Hype detection: Items where language intensity is high but evidence density is low are flagged. They are not removed — they are labelled explicitly, because hype is itself a signal about what narratives are being amplified.
Trend states: Each theme is classified as emerging, scaling, mature, or declining based on volume velocity, cross-source diversity, and adoption proxies. A media-only spike does not qualify as a trend.
Sources: YouTube (practitioner), Hacker News (practitioner), Substack newsletters (media), RSS feeds (varies). arXiv, Hugging Face, and operator changelogs are added incrementally.