AlphaGenome Research Workspace

Tier Map is a review-only view of candidate Tier 2 Mega themes grouped under Tier 1 families. Each row keeps the old demo grammar: a 120-month timeline, deterministic phase ribbon, emerging companies, and drill-down into the underlying narratives. The investability gate shows whether a group has a clear tradeable company basket and enough history for later price backtesting; it does not claim alpha.

Investable Themes is a shortlist from the full Theme Map. It answers: which themes should a researcher inspect first because attention is moving and evidence is available. It is not the full theme catalog, not a forecast, and not a buy list. Use Theme Map to browse the complete theme group → theme → Narrative hierarchy.

Why only a subset? The full hierarchy can contain many Tier 1 theme groups, Tier 2 themes, and narratives. This tab only shows sector research priorities that pass the current shortlist gates; event catalysts are counted separately and are not mixed into sector exposure.

Why the first card is first: the default order uses the research-actionability band first, then a history-adjusted attention score. The adjustment reduces over-ranking of themes with short observation histories. Lifecycle and filing labels describe evidence, not expected return.

How to read a card: Why now = attention movement; Evidence = narratives, articles, and filing exposure status; Exposure = whether current ticker-resolved leaders exist; Next step = inspect the card or open Theme Map. Missing filing exposure is shown as unavailable, never as zero.

Investment Concepts separates the observed narrative from an LLM-generated research frame. Only current-tree concepts that pass exact-roster provenance, company-evidence, and grounding gates appear. They are not recommendations, forecasts, or return claims.

Research Relevance uses a bounded DeepSeek agent to interpret the question, run local Narrative searches, and review the retrieved evidence. There is no regex fallback: if DeepSeek is unavailable or violates the response contract, the run fails visibly. Frozen fund-profile scores, when applicable, remain separately identified from the live agent run and from article sentiment, expected return, and recommendations.

Detail Workbench is the deep view over the same 290 narratives shown in Opportunities, but framed at the narrative lifecycle level (not period-relative). Use this tab when you need fields that Opportunities does not expose: Backtest signal (70% normalized sentiment z + 30% normalized news-volume z), POC alpha score, 3M vs SPY / QQQ, sentiment z, full investability gate breakdown, and the trend bucket (Rising / Cooling / Persistent / etc.) computed over the entire history, not just the selected period. Same labels & companies you see in Opportunities; different framing.

Each row is a review narrative from the new contract layer. Bars show monthly article volume for 2016-01 → 2025-12. Period is analyst-selected start/end month. New, Rising, and Re-emerging are recomputed inside that selected period, not stored as permanent narrative labels. Sorts are server-side over all matching narratives: latest volume z-score, eligible 3-month volume-z lift, POC alpha score, SPY/QQQ excess, volume, or confidence. Sentiment is not yet connected, so score displays are volume-driven. Trend and company identity support multi-select. Private companies are discovery candidates, not backtest targets.

Trend buckets are computed locally from cutoff presence and article-volume movement. They are not generated as free-form model scores.

Company Discovery is period-relative in this review layer. Click any company to inspect the narratives where it appears.