Three Questions Your Brokerage Statement Can't Answer

Question 1: "How am I actually doing?"

Open a brokerage app and you will see a price chart. For a growth stock that pays no dividend, that chart is the whole story. For anything that distributes income — a covered-call ETF, a REIT, a dividend payer, most of what an income-focused portfolio actually holds — the price chart is close to the least informative view available.

The reason is structural. A fund that pays out most of what it earns is handing you the return as cash instead of accumulating it in the share price. Its price line can sit flat for a year, or drift gently downward, while the total return — price movement plus every distribution — climbs steadily the entire time. Nothing is wrong. That is the design. But if you judge that fund on the chart your broker shows you by default, you will conclude it is dead money and sell something that is doing exactly what it was built to do.

The distinction that matters:

The word "reinvested" is load-bearing. Adding distributions up as cash undercounts the result, because in reality each payment buys more shares that themselves go on to appreciate and pay. Measuring it properly means treating every distribution as a purchase at that day's closing price — which is worth roughly a percentage point a year on a high-distribution fund, compounding.

This is what the new Compare page is for. Put up to eight tickers on one chart over a window from one month to five years, and toggle between total return and share price alone. On a broad index fund the two lines sit nearly on top of each other. On an income fund the gap between them is the entire investment case. A table underneath breaks out how much of each return came from income, how many payments there were, and flags anything that has not existed for the whole window rather than quietly comparing four years against one.

You can also plot your own accounts as lines on the same chart — all of them blended, or one at a time. That turns "am I beating the index?" and "is my Roth doing better than my taxable account?" into questions with a visible answer rather than a feeling.

Question 2: "Am I actually diversified?"

Most people check diversification by reading names. Four different funds, four different tickers, four different factsheets describing four different strategies. It feels diversified.

Correlation is the test that settles it, and it frequently disagrees. Two covered-call ETFs writing options on the same underlying index will move together almost perfectly — a correlation near 1 — no matter how differently their strategies are worded. Different strike selection, different expiry ladders, different sponsors, same behaviour. If you hold four of them you do not have four positions. You have one position with four tickers, and in the drawdown where you were counting on diversification, all four will do the same thing at the same time.

Compare now includes a correlation matrix computed from actual daily returns, which is the closest thing to a real diversification check: it measures what your holdings did, not what they claim to do.

Alongside it is a drawdown view — the chart switched to show how far each holding sits below its own running peak, with a maximum-drawdown column in the table. This is the number a cumulative return line hides most completely. Two holdings can finish the year at an identical total return while one dipped 8% along the way and the other fell 34%. Those are not the same thing to own, and the difference only becomes visible when you plot it.

Total return tells you what you made. Drawdown tells you what you had to sit through to get it.

Question 3: "What should I have been told?"

The third question is the one nobody asks out loud, because the honest version is embarrassing: what happened this week that I would have wanted to know about, and didn't, because I wasn't looking?

Watching for things manually does not scale. You check for a few days after deciding something matters, then you stop, and the thing you were watching for happens six weeks later while you are not looking.

Triggers are conditions you write down once and stop having to remember. There are three kinds, and the difference between them is worth understanding before you build any.

Market triggers watch a ticker for a moving-average cross, a price level, a new 52-week high or low, or an RSI level. Portfolio triggers watch your own positions for a gain or loss threshold, an option expiring soon, or cash building up past a level you set. Both are arithmetic: exact, and they cost nothing to run, however many you have.

AI triggers take a condition described in plain English and judge it on a schedule. They are the right tool when a condition genuinely needs interpretation, and the wrong tool when it does not. The rule we would give anyone: if you can write your condition as a number, use one of the first two kinds. It is exact, it is free, and it cannot misread itself.

Triggers are checked every weekday at 12:20pm ET while the market is open. Each one goes quiet for a cooldown after firing, which is the detail that decides whether an alert system is usable at all. A condition like "the 50-day is above the 200-day" does not stop being true the day after it becomes true — it can hold for months. Without a cooldown you get the same email every morning until you filter the sender. With one, you hear about the change rather than the state.

When a trigger fires it tells you why, with the actual numbers, and that history stays on the page whether or not you opened the email.

The fourth thing: asking directly

Compare and Triggers each answer a specific shape of question. Some questions do not have a shape — "how exposed am I to a tech downturn?", "what is actually generating my income?", "which of my positions is doing the least work?"

Research takes those in plain English and returns a written report built from your real positions, weights, open options, closed trade history, cash and tax figures. Follow-up questions cost less than the one that opened the topic, and the whole conversation is saved, so your research history is something you can return to rather than something you lose when you close the tab.

One design decision is worth stating plainly, because it is the difference between an assistant that is useful and one that is quietly dangerous. Every number is calculated in code and handed to the assistant finished. Language models are unreliable at arithmetic and at direction comparisons — whether a figure is above or below a threshold — so anything that computes in prose will eventually state a confidently wrong number in a context where you might act on it. The assistant's job is explaining, weighing several considerations at once, and structuring the answer. The numbers are not its opinion.

It is also explicit about its limits: it will not forecast returns, project a future balance, or claim to see anything outside your connected brokerages. If you mention a tax rate or a holding in conversation that conflicts with your saved settings or your actual synced positions, it uses the saved data, says which it used, and notes the difference — rather than running the rest of the answer on a number typed once in passing.

All three are on every plan

Research, Compare and Triggers are included on every plan, free tier included. Plans differ in how many AI queries you get each month, not in which pages you can open.

Compare never uses a query — it is computation over price history. Market and Portfolio triggers never use one either, for the same reason. A Research question uses three queries and a follow-up uses one, and only AI triggers spend a query each time they run. Both pages show your remaining balance and what an action costs before you take it, so nothing quietly draws down an allowance you were not watching.