What's SETRA?

SETRA (Scientific & Economic Trend Research Agent) is an AI-powered system that monitors the intersection of academic research and financial markets, detecting emerging technologies with real market validation.

System Architecture

Collection

arXiv papers, FRED economic data, ETF prices, USPTO patents

Enrichment

Semantic Scholar citations & author data, OpenAlex FWCI & topic mapping

Analysis

Attention Index scoring, field momentum, macro anomaly detection, LLM narrative generation

Output

Dashboard, AI chat (GAME SDK), automated X thread posts

Data Collection

arXiv Papers

Daily automated collection from arXiv API across all categories. Each paper includes title, abstract, authors, categories, publication date, and PDF link. Up to 2,000 papers per query with rate-limited API calls.

Key tracked categories:

Artificial IntelligenceMachine LearningComputational Linguistics / NLPRoboticsCryptography & SecurityComputer VisionSoftware EngineeringDistributed ComputingInformation RetrievalStatistics / MLMedical ImagingQuantitative BiologyQuantum Physics

Semantic Scholar Enrichment

Papers are enriched via S2 batch API (400 papers/batch) with:

  • Citation count, influential citation count, reference count
  • TLDR (AI-generated one-sentence summary)
  • Author h-index and affiliations
  • Citation history (year-by-year breakdown for surge detection)
  • DOI (enables OpenAlex lookup)

OpenAlex Enrichment

DOI-based enrichment from OpenAlex (480M+ works corpus) providing:

  • FWCI (Field-Weighted Citation Impact) — normalized citation score
  • Citation normalized percentile within field
  • Topic hierarchy (domain → field → subfield → topic)
  • Yearly citation trajectory for trend analysis

Financial & Patent Data

FRED Economic Data (12 series)Fed & ECB balance sheets, M2, Fed Funds rate, 10Y/2Y yields, CPI, HY spread, R&D spending, EUR/USD. 90-day lookback per collection cycle.
ETF / Market Data (28 tickers)Theme ETFs: BATT, LIT, TAN, ICLN, SOXX, SMH, AIQ, BOTZ, ROBO, QTUM, ARKG, XBI, XLB, URA, ARKW. Plus VTI, AGG, VNQ, GSG, GLD and regime indicators (VIX, DXY, Copper).
USPTO Patents7-day rolling window of patent filings via PatentsView API, filtered by CPC codes mapped to research categories.

Attention Index (0-115)

A multi-signal score measuring how much academic and market attention a paper is receiving. The score has two parts: a Base Score (0-100) from academic signals, and a Financial Boost (0-15) from market signals. Each academic component is percentile-ranked within the paper's arXiv category.

Base Score (0-100)

30%
Citation Surge

Recent citation momentum vs. historical baseline. Detects papers gaining attention rapidly via Semantic Scholar weekly diffs and citation history.

25%
Field Momentum

Category-level publication growth blending OpenAlex global corpus data (70%) and local DB growth (30%). Hot field threshold: FMS > 50.

20%
Citation Scale

Citation impact using OpenAlex FWCI (Field-Weighted Citation Impact) and citation percentile when available, with category percentile fallback.

15%
Author Impact

Maximum h-index among the paper's authors (via Semantic Scholar), percentile-ranked within category.

10%
Influential Citations

Number of highly-cited papers that cite this paper, indicating recognition by key researchers in the field.

Financial Boost (0-15)

+10
ETF Anomaly

When a category-linked theme ETF shows abnormal returns (z-score > 2.0), the paper gets a boost proportional to the anomaly strength.

+5
Patent Velocity

YoY growth of patent filings in related technology areas (USPTO PatentsView API). 50%+ YoY growth yields the full boost.

Field Momentum Score (FMS) is computed per arXiv category by blending OpenAlex global publication growth (70% weight) with local database publication growth (30% weight). Categories with FMS > 50 are classified as "Hot Fields."

Macro Analysis

Daily macro snapshots aggregate financial data from FRED (12 series) and market sources (28 tickers). SETRA analyzes how macro conditions affect technology sectors, R&D funding, and research activity.

Central Bank Liquidity

  • FRB + ECB balance sheets (USD trillions)
  • M2 Money Supply (weekly)
  • Reverse Repo operations (daily)
  • MoM and YoY % changes

Interest Rates & Regime

  • Fed Funds Rate
  • 10Y / 2Y Treasury yields & yield spread
  • CPI and real rate (10Y - CPI)
  • Rate regime classification (negative / low / moderate / high real)
  • High Yield credit spread (bps)

Market Regime Indicators

  • VIX (volatility index) with regime labels
  • DXY (US Dollar index) with 1M change
  • Copper price (economic activity proxy)

Asset Class Performance

  • Equity (VTI), Bonds (AGG), Real Estate (VNQ)
  • Commodities (GSG), Gold (GLD)
  • YTD and 1-month returns

Innovation Metrics

  • US private R&D spending (FRED quarterly)
  • Patent velocity (YoY growth by CPC code)

Anomaly Detection

SETRA's 3-layer anomaly system detects: (1) single indicator anomalies (rapid liquidity changes, rate moves), (2) cross-asset patterns (liquidity-rate divergences, equity-bond disconnects), and (3) tech-impact analysis explaining how each anomaly may affect technology sectors and R&D funding.

Automated X Posts

SETRA generates and publishes paper highlight and macro analysis threads to X using Claude (Sonnet) for content generation and Virtuals Protocol GAME SDK for publishing.

Paper Thread (4 tweets)

  1. Hook — bold opening that stops the scroll
  2. Data + Explanation — Attention Index, citation surge, what the paper does
  3. Authors + Forward look — provokes engagement
  4. PDF link + hashtags (links in thread avoid reach penalty)

Macro Thread (4 tweets)

  1. Hook — lead with the anomaly or inflection point
  2. Numbers + Regime — key changes in liquidity, rates, markets
  3. So What — connect to tech & innovation funding
  4. Hashtags

Each tweet is capped at 270 characters. Human-like delays (5-15s) between thread tweets. Performance feedback loop: engagement metrics from past posts are injected into prompts to improve future content.

Chat Capabilities

SETRA's chat interface is powered by Virtuals Protocol GAME SDK. The AI agent can query the paper database, analyze macro conditions, and provide research insights in real-time via 7 specialized functions.

search_papers

Full-text search on title and abstract with attention threshold filtering

get_paper_detail

Full paper details including attention score breakdown and narrative context

get_trending_papers

Top papers ranked by Attention Index

get_macro_snapshot

Latest macro snapshot with all financial indicators

get_macro_anomalies

Recent anomalies with severity level and tech impact analysis

get_field_stats

Field momentum scores — filter by hot fields only or see all categories

get_recent_posts

SETRA's recent X posts about papers and macro insights

Example questions: "What are the hottest AI papers this week?", "Show me papers with citation surges", "What's the current macro environment?", "Are there any anomalies in the market?"

Tech Stack

BackendPython / FastAPI
DatabasePostgreSQL
FrontendNext.js 15 / React 19
AI ChatGAME SDK v0.1.5
Post LLMClaude Sonnet
X Publishingvirtuals-tweepy
ChartsRecharts
ORMSQLAlchemy 2.0