Exits, targets and stops

Every position the desk opens is opened with an exit already written: a stop, the price that says the idea is wrong, and usually a target, the price that says it worked. Those two numbers are not a hope. They are the plan the Portfolio Manager has to manage the position against, the level the Executor watches on every check, and the level a second stop rests under at your broker. This page says who acts on them, when you are asked and when you are not.

Where you see them

Open a holding on Portfolio: Stop, Target and Sells if sit under the thesis, and the row's Above stop column carries the distance

Who acts on a stop

The Executor, on every check, in any autonomy mode, without asking you

Behind it

A stop order resting at your broker a little lower, so the position is still covered if Bellwether is offline

Who changes them

The PM, and only in one direction: a stop is tightened, never loosened

What you do

Nothing, normally. A protective exit is a notice, not a question

The plan is written at entry

When the PM proposes a buy it writes the stop with it, and the approval card shows it before you say yes (reading a card). Once the position exists, that stop is armed: the Executor holds the level and checks the price against it every time it runs. The target is the price the plan sells into, whole or in part, and a position can also carry a time limit ("out by Friday") or a rule to close on a weak daily close instead of an intraday dip.

The Portfolio row says which of these exist for each name in one line each: the one stop the Executor will act on with the distance to it, the target with the PM's own qualifier ("$425.00, trim a third"), and the condition the PM wrote that would kill the idea. A name whose thesis names a line the PM has not armed yet reads "not armed", and one with nothing reads "none".

The desk manages the plan mid-session

On the mid-session check of a book that holds positions, the desk works the plan rather than re-arguing it:

• A target covered at today's prices is sold in that session, in whole or in the fraction the plan named. Leaving it for the end-of-day check is not an option, because the end-of-day check parks orders for the next open and cannot sell today.
• A stop is never lowered or widened. It moves up only where the written plan said it would, for example to the entry price once the position is a certain distance ahead.
• Selling part of a position that has hit neither stop nor target needs a named reason: the thesis broken by a filing or a print, the market turning risk-off in that session, an event inside the idea's horizon, or one of the book's own limits, such as a name that has grown past its concentration limit or a halt. "Locking in a little" is not a reason the desk accepts from itself.
• Every held name gets one line in the session's handoff, even when nothing happened ("NVDA stop 106, target 127, neither touched, no action"), so a name nobody mentions counts as unmanaged.

This applies to the scheduled mid-session check. The morning plan, the end-of-day check and sessions the desk is woken for are unchanged. You can read the lines it wrote on Desk under "What it decided and why".

An exit that acts is a notice, not a question

When a stop or a target is touched, the Executor sells. It does not queue the sale for your approval, whatever your autonomy setting says about new buys, because the protection you already approved is what is being carried out. So the sale shows up as something that happened:

• On Inbox › Activity as its own sentence in the Executor's voice: "Protective exit selling 2 MSFT because its stop was hit".
• On Portfolio the row's Above stop cell reads selling once the exit is working, with the stop's price under it (or "target reached" when it was the target). A price that has reached an armed stop before the exit shows reads "at or through" in place of a distance. On the Overview's positions list the same name reads "price at or below it".
• On the Overview it is an Alerts row saying that a protective stop or target fired and the position is selling. In the Inbox it never appears under Waiting and is never counted in the number beside Inbox; it is a Trades row on the Activity tab. It carries no Approve and no countdown. The one exception is the next section, and it is the only protective exit that ever wears them.

The one exit that waits for you

There is a single case where a protective exit genuinely needs a person: a position that was not opened through Bellwether, such as one that was already in the account when you connected it. The Executor has no entry of its own behind that holding, so the sale falls over your approval line like any other order, and it waits.

That card is drawn as an exception rather than a routine ask. It leads with what happened and why you of all people are being asked, says what happens if nobody answers, and only then shows the figures:

In the Inbox the ask is a row pinned at the top of Waiting ("Bellwether needs your OK to sell 2 shares because …"), and opening the row shows the card quoted here.

It keeps Approve and Reject because the Executor is really waiting for the answer. Approving a sale never asks for your authenticator code, however large it is: the code guards putting more money at risk and loosening your protection, never using it.

The broker stop behind the stop

Bellwether's own stop only works while Bellwether is running. So for every armed stop, the Executor also rests a stop order at your broker a little under the same level, moves it whenever the stop moves, and cancels it just before selling the name itself, so nothing is ever sold twice. If the broker's copy fills first, Bellwether adopts that fill as the exit.

This is on by default on every book whose broker can hold a stop order, and it is the reason a holding reads one stop figure and then one sentence, never two competing numbers:

If your broker refuses the resting order, the row says so and Bellwether's own stop keeps watching; you also get an alert, which clears itself when a later attempt is accepted (alerts). The gap between the two levels and the kind of order that rests there are set for the whole platform, not per book, so there is nothing for you to tune here.

The broker's copy also fires on the first trade at its own level, where Bellwether's stop waits for a second print to confirm. The small gap between the two keeps an ordinary touch of your stop from setting the broker's copy off, but one stray print that reaches the lower level can still sell a position Bellwether itself would have held.

What it looks like afterwards

A sale through a resting broker stop can fill in pieces. The alert folds the pieces into one row and says the whole thing once, with each fill behind Details:

Portfolio's closed trades name the reason the position ended (stop hit, target reached, trailing stop hit, thesis invalidated, closed by the PM, sold in steps by the PM, rebalanced by rule, you sold it, or closed by Flatten all), and Performance counts it from there.