What's this text not telling you?
That answer in front of you is about to become a decision. Paste it. See which numbers are in a source, and which you would be guessing on.
I would overweight NVDA here.
NVIDIA reported FY2024 revenue of $60.9 billion, up 126% from $26.97 billion. Net income was $29.76 billion. Microsoft, Amazon, Google, and Meta together committed roughly $200 billion to AI data-center buildout. For FY2025 NVIDIA guided revenue to $130 billion.
Don't size the position off $200B or the $130B guide.
Don't act on these yet
- $200 billion Hedged
- $130 billion A forecast, not a fact
Checked
- $60.9 billion FY2024 revenue — 10-K
- $29.76 billion FY2024 net income — 10-K
- What could go wrong with $200 billion?
- What is unknown about $200 billion?
Try it on this recap, or paste your own.
Computational measurements first. LLM grounding when APIs cannot resolve.
Framing fingerprint
Five analytical dimensions, measured by regex and word lists.
- Coverage: which of five analytical categories the text addresses
- Temporal orientation: past, present, or future framing in percentages
- Voice: prescriptive, promotional, analytical, descriptive, advisory
- Epistemic basis: sourced vs unsupported numerical claims
- Claim density: hedged, unhedged, predictions, layered structure
Source network
Authoritative data sources queried per claim, in priority order.
- SEC EDGAR: US public company filings
- World Bank: country statistics, development indicators
- FRED: US macroeconomic time series
- REST Countries: demographics, geography
- CoinGecko: crypto prices and market caps
- Alpha Vantage: equity fundamentals
- Wolfram Alpha: physical and computational facts
- Wikipedia: encyclopedic reference
- Brave Search: fallback for everything else
Privacy by construction
Not a promise. The schema cannot represent your text.
- Never stored: your text, your source material, your topic string
- Never trained on: nothing you paste reaches a training pipeline
- Never logged: no IP address, no email, no fingerprint
- What is recorded: counts and categorical summaries only, contributing to the Frame Check Corpus