BlackbookBlackbook
Sign up→
Blackbook Intelligence Layer · For AI labs & data partners

Train financial AI on
consented human reasoning
not scraped clicks.

Blackbook turns public financial filings and expert user behavior into AI-ready financial reasoning datasets — vertical-specific labels, theses, and feedback from PMs, analysts, quants, allocators, and serious independents.

Talk to partnerships →See the consumer product

Three data products

Pick a layer or license the full stack.

$25k – $500k
Filing-to-Thesis Dataset

Labeled SEC, 13F, and STOCK Act events paired with human-written theses, confidence scores, time horizons, and outcome tracking.

Best for: Fine-tune financial LLMs on real expert reasoning.
$2.5k – $25k / mo
Signal Graph API

Real-time anonymized, aggregated behavior + sentiment per ticker, fund, sector, and politician. What serious investors actually pay attention to.

Best for: Power financial agents, copilots, and research tools.
Custom
Financial Agent Eval Benchmark

Held-out tasks for testing whether AI models can read filings, spot crowded trades, and explain unusual disclosures the way a PM would.

Best for: Benchmark and rank financial reasoning models.

A sample reasoning graph

Every market event in Blackbook produces a structured reasoning graph: what serious investors voted, why, with how much confidence, over what horizon, and what actually happened next. Cleaner than scraped sentiment. Safer than raw user data.

  • · Anonymized — no identities ship with datasets
  • · Tier-weighted — verified contributors signal stronger than retail noise
  • · Outcome-anchored — labels tied to subsequent price moves
  • · Schema-stable — versioned for fine-tuning + eval pipelines
{
  "subject_type": "fund_move",
  "ticker": "AMD",
  "event": "multi-fund accumulation, 3 PMs Q3",
  "human_sentiment": "bullish",
  "expert_confidence_avg": 76,
  "dominant_reasoning": [
    "AI capex cycle",
    "fund convergence",
    "valuation reset"
  ],
  "counterarguments": [
    "crowded trade",
    "margin pressure"
  ],
  "horizon": "3m",
  "30_day_outcome_pct": 4.8,
  "vote_count": 142,
  "contributor_tier_mix": { "verified": 18, "contributor": 124 }
}

What we collect — and what we never will

Licensed
Public filings, cleaned + enriched
SEC EDGAR · 13F · STOCK Act · derivatives disclosures
Licensed
User labels, votes, theses, annotations
Anonymized · consented · contributor-tier rewarded
Never sold
Raw PII, private messages, emails
Never licensed. Never sold.

Users opt in explicitly. Verified contributors and analysts earn status, credits, and future revenue share when their labels ship inside licensed datasets.

Building a financial AI agent?

We license raw datasets, real-time APIs, and custom eval benchmarks. Tell us what you're training and what shape of signal would move the needle.

partners@blckbook.ai →