ProCap Insights Review: What Does $2,500/Year AI Stock Research Actually Look Like?
April 8, 2026 · 14 min read
ProCap Financial (Nasdaq: BRR) launched ProCap Insights on April 7, 2026: an AI-generated stock research service priced at $2,500/year. The Wall Street Journal covered the launch. Anthony Pompliano called it the future of financial research.
We reviewed the free reports available at launch, examined their disclosed methodology, and identified their data sources to help investors evaluate what they're getting, what questions to ask, and what “AI-generated research” can mean in practice.
What We Reviewed
ProCap published roughly 12 reports in its first two days. About seven were free. We read the free reports available at launch, including:
- "Companies Replacing Workers with AI Have Lagged the S&P 500 by a Landslide"
- "Insiders Are Dumping Tech Stocks and Buying Energy"
- "5 Stocks That Win From $166 Billion in Tariff Refunds"
- "3 Stocks That Win From Both Tariff Refunds and the Iran Oil Shock"
- "These 2 Stocks Just Struck Gold with Anthropic"
Each report runs approximately 3,000–4,000 words with a consistent structure: executive summary, data analysis, specific stock picks, counter-arguments, a catalyst timeline, and source citations. The format is professional and readable.
What ProCap Insights Does Well
The reports are better than most AI-generated financial content we've seen.
Quantitative grounding
Reports cite specific correlation coefficients (r=0.03 in the AI-replacement study), sample sizes (N=34 companies), and percentage-point deltas. The insider trading report includes specific buy/sell ratios and dollar amounts ($100M+ NVDA insider sales). This is real data, not hand-waving.
Counter-argument sections
Each report includes a multi-paragraph counter-argument section that steel-mans the opposing view. The insider trading report's discussion of 10b5-1 pre-scheduling plans and mixed academic evidence on insider signal validity is substantive.
Volume and speed
Six reports per day covering timely macro themes (tariff refunds, Iran oil shock, Fed regime change) demonstrates genuine capability in rapid thematic research. For investors who want breadth and speed, this delivers.
The Methodology Gap
ProCap's marketing describes their process as “AI agents that evaluate data from multiple perspectives and pressure-test conclusions through structured debate.” Their website adds: “Humans have limits. Our AI agents don't.”
That is the entirety of their methodology disclosure.
For a $2,500/year research product, investors might reasonably want to know more. Here's what's disclosed versus what isn't:
Methodology Transparency Checklist
Not disclosed. No mention of specific models, model versions, or whether multiple models are used.
Not disclosed. No details on agent roles, debate structure, convergence criteria, or how disagreements are resolved.
Partially disclosed. Report citations reference "FMP via OpenBB MCP" — Financial Modeling Prep data accessed through OpenBB's agent framework.
Not disclosed. Reports present specific tickers but don’t explain the screening or selection process.
Phil Rosen is described as the sole human overseeing output. No disclosed review process, compliance framework, or accuracy tracking.
No. The product launched April 7. No historical predictions, no backtesting results, no accuracy metrics.
ProCap says reports "may draw on trends derived from anonymized data" from Silvia. No specifics on what this means in practice.
Data Sources: What's Under the Hood
The most revealing detail in ProCap's reports is their source citations. Reports consistently cite “FMP via OpenBB MCP”, meaning they use the OpenBB platform with Financial Modeling Prep as their primary financial data API.
This is a perfectly reasonable tech stack for financial data retrieval. FMP provides fundamentals, price data, insider transactions, and SEC filing metadata. OpenBB's MCP integration allows AI agents to call these APIs as tools.
However, it also means the reports are built primarily on a single commercial data API. We didn't find evidence of primary source analysis: no direct SEC filing parsing, no earnings transcript analysis, no alternative data collection (patent filings, job postings, Google Trends, congressional trading disclosures). The data breadth appears limited to what FMP exposes through its API.
Report Quality: Competent Commentary, Familiar Format
The reports are well-written financial commentary with useful data tables and substantive counter-argument sections. The quality is comparable to what you'd find on Seeking Alpha or Yahoo Finance: an interesting thematic angle, specific tickers that fit the narrative, and a case for why the trade works. The AI layer makes this faster and more polished, but the underlying motion is the same: narrative first, stock picks second.
Several patterns are worth noting:
Headlines over-state the analysis. The AI-replacement report finds r=0.03 (essentially no correlation) but frames it as “lagged the S&P 500 by a landslide.” The data doesn't support that framing.
Thin sample sizes drive trading suggestions. One report suggests shorting companies announcing AI-driven layoffs based on a 10-company sample. That's not enough data to validate a trading strategy.
Shallow per-ticker analysis. Individual stock sections run 2–3 paragraphs. No discounted cash flow analysis, no segment breakdowns, no management assessment beyond the headline thesis. The analysis is thematic first, company second.
Trade ideas from a publicly traded company. ProCap Financial is a Nasdaq-listed bitcoin treasury company (BRR). Generating subscriber revenue improves their financial metrics. This doesn't mean the research is biased, but the incentive structure is worth noting, especially when reports include explicit trade suggestions (“stocks to buy,” “short these companies”).
The Corporate Context
ProCap Financial (BRR) is a bitcoin treasury company that went public via SPAC, holding ~5,400 BTC. The stock has declined roughly 85% from its peak, trading near $2. Pompliano said ProCap Insights was built in two weeks by one employee at a cost of a few thousand dollars.
The Silvia Acquisition: What the SEC Filings Show
The day before ProCap Insights launched, ProCap closed its acquisition of CFO Silvia, an AI personal finance dashboard. ProCap says Insights “may draw on anonymized data” from Silvia's user base. We reviewed the SEC filings around this deal.
Related-party transaction
Per the DEFM14A proxy statement, Pompliano's private company (Inflection Points, Inc.) owned 51% of CFO Silvia. Pompliano is CEO of both the acquirer (ProCap) and the majority seller. The deal was structured as an all-stock merger issuing up to 17.4 million shares (7.5M at closing, 900K in escrow, up to 9M in earnout shares triggered if BRR reaches $9.00).
What was acquired
CFO Silvia Inc. was incorporated September 19, 2025, approximately five months before the merger agreement. At the time of the deal, it had zero revenue, four full-time employees, and no disclosed pricing model. The proxy statement notes the combined company is expected to incur losses “for the foreseeable future.”
The product is a Next.js web application that aggregates financial accounts via Plaid (banks), SnapTrade (brokerages), and Coinbase OAuth (crypto), with manual entry for physical assets. It provides a net worth dashboard and an AI chat interface. The founder, Shain Noor (age 26), has a legitimate technical background (internships at Microsoft, Amazon Lab126, and NASA) and received a $5M cash signing bonus plus $700K base salary as CTO.
The “$30 billion in tracked assets”
ProCap cites $30B+ in assets tracked across ~12,000 users with an average net worth above $2.5M. This figure represents read-only account balances via Plaid plus self-reported asset values (users manually entering their home value, car value, etc.). It is not assets under management. For comparison, Mint “tracked” trillions using the same Plaid aggregation approach. The $30B figure has not been independently audited.
Opposition and governance concerns
ISS (Institutional Shareholder Services), the leading proxy advisory firm, independently recommended shareholders vote against the merger. Activist investor ATG Capital published two open letters detailing conflicts of interest, noting that the fairness opinion fee was 83% contingent on deal closing ($250K of $300K) and that no discounted cash flow analysis was performed because “adequate long-term financial projections” were unavailable.
ProCap held only one regular board meeting in all of fiscal 2025, with zero committee meetings. A Special Committee member resigned mid-process without explanation. ProCap's Chief Investment Officer resigned three days before the deal closed, with his non-compete fully waived.
The data question
Users signed up for a free personal finance app. Their data now potentially feeds a $2,500/year stock research product sold by a publicly traded bitcoin treasury company. The privacy policy was written for a standalone fintech app and states Silvia “may retain de-identified or aggregated data for research and analytics.” Whether this covers the current use case is an open question. With ~12,000 users, the statistical utility of “anonymized” aggregate data for institutional-grade research is also unclear.
What Transparent AI Research Looks Like
The difference between narrative-driven commentary and rigorous analysis is the framework. ProCap tells you “these 5 stocks benefit from tariff refunds” and lets you decide whether to act. We map the constellation of evidence across multiple lenses, show where our models agree and disagree, and publish falsifiable predictions that we score against reality. Methodology transparency should be the baseline for anyone charging for AI-generated analysis.
| Dimension | ProCap Insights | Runchey Research |
|---|---|---|
| Methodology | "AI agents debate" | 14 named lenses, published signal registry, evidence ladder |
| Analysis depth | 2-3 paragraphs per ticker | Multi-lens committee discourse per company |
| Data sources | FMP API via OpenBB | SEC filings, earnings transcripts, FRED, patents, job postings, congressional trading |
| Model transparency | Not disclosed | Named personas across Claude, Gemini; visible model agreement scores |
| Framework | Narrative → stock picks | Evidence map → calibrated forecasts |
| Track record | None (day 2) | Prediction markets with Brier score calibration |
| Coverage breadth | ~6 thematic reports/day | ~200 company analyses + sector deep-dives + macro conditional forecasts |
| Speed | Same-day macro reactions | Multi-day deep dives |
| Distribution | Pompliano brand, Phil Rosen (207K subscribers) | Organic |
| Price | $2,500/year | Free |
We publish our persona system, individual lenses, and assessment methodology because we believe the process is the product. If our methodology has weaknesses, we want readers to be able to identify them.
Questions to Ask Any AI Research Provider
Whether you're evaluating ProCap Insights or any other AI-driven research service, here are the questions that matter:
Which models generate the analysis, and how are they orchestrated?
A single model producing a report differs from multiple models debating a conclusion. One generates prose; the other stress-tests it.
Can you see the disagreements, or only the conclusion?
The most valuable signal in multi-model analysis is where models disagree. If you only see the polished output, you're missing the most important information.
What primary sources are analyzed: filings, transcripts, or just API data?
An AI summarizing FMP data points is different from an AI parsing a 10-K filing and identifying accounting irregularities.
Is there a track record with measurable accuracy?
Any system can produce confident-sounding analysis. Calibration scoring (like Brier scores) reveals whether confidence correlates with accuracy.
How is the editorial layer applied?
One human reviewer for six daily reports is a different quality bar than a multi-stage review process. Neither is inherently wrong, but investors should know which they're getting.
The Bottom Line
ProCap Insights is a real product producing readable financial commentary with genuine quantitative grounding. The counter-argument sections alone put it above most AI-generated financial content.
But at $2,500/year, investors deserve to know more than “AI agents debate.” They deserve to know which models, which data, which process, which checks. They deserve a track record before trusting trade suggestions. And they should understand the corporate context (a bitcoin treasury SPAC that acquired a zero-revenue AI startup the day before launch).
The quality spectrum across AI research products will be enormous. The differentiator will be which ones show their work.
We show ours. We think you should demand that of anyone asking for $2,500.