Conviction custodians

A conviction custodian is your recorded reason for holding a position. Every name in your book is anchored to one of three things:

• a source — you hold it because a manager you follow holds it,
• your own thesis — you hold it because you decided to, on your own reasoning,
• the system — it's there mechanically (an index fund, a placeholder, a sweep).

That's the whole idea. It sounds administrative until you try to answer the question without it.

The problem it solves

Portfolios accumulate. A position enters for a good reason and then the reason quietly leaves: the manager who convinced you sold out, the thesis you had in 2023 stopped being true, or you honestly can't remember why it's there. The position stays, because nothing ever prompts you to revisit it.

Anchoring makes the reason explicit at the moment you hold it, so later — when you're deciding what to sell to fund something new — you're choosing between stated reasons rather than between tickers.

Orphans: positions with no reason left

An orphan is a position you hold that no source's conviction backs anymore. The clearest case: a manager you copy exits a name, you don't see the disclosure, and you keep holding a position that its originator abandoned.

Two things about how orphans are reported:

• The report never acts. It flags; you rule. It will not sell, will not auto-anchor a position to a source it guessed at, and will not decide something is a mistake.
• Right after you import a book, everything is an orphan — nothing is anchored yet. That's a setup step, not an alarm, and the app says so rather than showing you a red wall of nothing.

You can also declare what's out of scope: cash is ballast, not a conviction, and some holdings are deliberately off-strategy. Scope is set in your construction rules, and excluded holdings are named and counted rather than silently dropped — you can always see what wasn't examined.

Copy fidelity: are you actually following them?

If you set out to follow a manager, copy fidelity measures how well you actually are: how much of your book overlaps theirs, and which positions account for the biggest gaps.

It's the difference between believing you're running someone's strategy and knowing it. Fidelity is measured against what you could realistically have bought — your exclusions and untradeable names are applied first — because an infidelity you're structurally incapable of correcting isn't a measurement, it's noise.

Anchoring a whole book at once

Anchoring 80 positions by hand would be exactly the kind of machinery this product is supposed to spare you. So there are two paths, and both stay: click a dropdown on any row to anchor one position in context, or ask your connected AI assistant to propose anchors for the whole book with its reasoning, review the proposal in conversation, and let it apply the batch once you approve.

The AI executes an anchoring you state. It doesn't decide what you believe.