Strategy and Risk say one number per fact, the Inbox's Quant cards read in words, the symbol page reads one truth

What changed for people using Bellwether

• Strategy and Risk say one number per fact. The Autonomy setting reads "Semi-automatic up to $300" and cannot be read as its opposite; the size-tier card shows today's equity, the same figure as the header, names the band and the days left before anything loosens. A limit the tier holds below your stored value reads "In force 6 a day".
• One name per desk session, counts that add up. The morning plan, check-ins, research brief, end-of-day check and daily review carry the same names on Your desk, the Schedule and Usage; a session that stopped early says so everywhere and a paused slot reads "paused". The Schedule's summary is counted from its rows and names the Portfolio Manager's share.
• Quant recommendation cards read like trade cards. Each card leads with one sentence of what would change and what the Quant expects, shows labeled rows in your limit names ("Largest order, $500 to $250, tighter"), and keeps the statistics behind "Show the numbers". Accepted or auto-applied cards say so in the past tense; "Needs your code" says which code.
• The symbol page reads from one source. Your position block lists the same protective stop as the strip above it and the same open orders as the Portfolio; with live prices off the page shows the last close, dated, with no intraday change, and a missing price is words, not "$0.00". Each algorithm has one label, defined on hover.
• A symbol page that fits a phone. The actions fit in two rows with the chart inside the first screen, the chart has a price scale that names its dotted reference line, peer prices say which close they are, a failing tradability screen leads with what it found, and the nightly screen's numbers sit behind a disclosure with plain labels.

What you may notice

• Tab counts on Strategy and Risk say what they count ("Guidance 2 to acknowledge", "Recommendations 8 open", "Limits 11 rules") and no tab or badge carries a hover title.
• The Quant's attribution on the PM's page is one sentence ("… has done best so far, +10.9% over 9 trades; … worst, −3.7% over 4 trades") with the per-algorithm table behind a disclosure.
• The tell-desk card in Consult says "the words your desk will read" rather than claiming they are exactly what you typed.
• The go-live check "Guidance notes acknowledged" reports the acknowledged share of the desk's last five sessions and passes only at 100%.
• Request cards from Consult say "If you accept" as labeled before-and-after rows; the Critic's finding card and the Requests strip say what their links open.
• On the symbol page the sentence about the PM and the signals never asserts an agreement where the PM has made no call; a name the desk cannot trade leads with "Not tradable here"; guidance rows read Enforced, Instructed or Watching with their status in words.
• A stock split older than your lot is no longer listed as applied to it, "sector" means one level everywhere, a retired kind of recommendation is marked "Retired, decline only", countdowns read "3h 57m left", and Admin › Usage names a Consult session "Consult question".

Known gaps and what’s next

• The Algorithms page still shows the older short algorithm names in two places; it adopts the symbol page's label table when the Algorithms tabs work lands.
• A few words on these screens (the tier badge's glue, "operator pin", ISO dates in two captions, the "↗" on session links) are left for the wording pass of the same audit.
• Decided rows in the Inbox and the order sentences in Activity converge on one order grammar with the Monitor pass; a sale's "since you sold" figure waits for its data to be defined per side.
• The help article on Consult still says the tell-desk card carries your words "exactly as you typed them"; the console no longer claims it and the article follows in the help wording pass.