Four analysts, one argument, and the receipts

What does a research report contain?

A 0–100 score and letter grade across six dimensions, a plain-English synthesis, a bull-versus-bear debate with a judge, validated alternatives, and how the holding fits your actual portfolio.

• ETFs are scored on cost, what you own, diversification, methodology, liquidity and fit
• Stocks on valuation, quality, growth, financial health, liquidity and fit
• Four personas: the Indexer, the Value Analyst, the Risk Manager, the Growth Scout

How do you stop the AI making things up?

By not letting it own the numbers. Every dimension starts from a deterministic score computed straight from market data, and the panel’s verdict is clamped to within a band of it — so a confident, wrong model cannot drag an objective measure to an extreme.

• Scores are the median of the panel, with disagreement shown, not hidden
• Dimensions the panel cannot verify score neutral and are excluded — a data gap is not a defect
• Suggested alternatives are checked against live market data; invented tickers are dropped

Whose AI key does it use?

Yours. Add a key from OpenAI, Anthropic or Google, pick the model, and you pay that provider directly at their prices — we add no markup and never resell inference. The page estimates what a run will cost before you start it.

Does it tell me what to buy?

No. It scores what you asked about, shows how it reached the score, and puts it next to what you already own. These are educational analysis, not individualised financial advice.