Alerts that say what happened and what to do, with the day named

What changed for people using Bellwether

• Alerts say what happened and what to do, in a sentence. Every alert on the Overview and in Activity now reads as a short title that names what happened (the check that did not run, the tickers involved, the position that differs from the broker), one plain line under it that says what to do or what happens next, and a fold for the specifics. The long unformatted paragraphs are gone, and so are the internal tool and field names that had leaked into them.
• A refused protective stop that later lands says so. When the broker refuses a protective stop because another sell order is holding the shares, and the retry a minute later is accepted, the Overview row now leads with "protective stop is back in place" and a "Restored 15:05" pill, with the original refusal kept behind "What happened at 14:30". A refusal no longer sits on the list for hours reading as if it were still going on.
• The desk's own reports go to the right reader. When the desk notices that two of the platform's own readings disagree, or that one of its tools failed, that report now goes to the platform team's log instead of the owner's Overview, titled in plain words with the desk's note folded underneath. What stays on the Overview is what the owner can act on: "The PM cannot follow the mandate as set" for the names involved, with a link that opens Guidance.
• A missed check names the check, the cause and the day. "The 10:30 intraday check did not run" now says why in one sentence (the previous check was still running, the day's AI budget held it back, or it started once and failed) and ends with "The next check is at 12:30 CT; nothing for you to do." A slot left over from a previous day is closed out quietly instead of appearing as a fresh alert, and lateness is never shown as a negative number.
• A broker connectivity blip is not an incident. One failed pass of the trading service caused by a momentary broker or network error is now logged as a warning; only five in a row, or a full minute without a good pass, is raised as critical. The service already retried and recovered in every such case; the alert's weight now says so.

What you may notice

• Alerts from earlier days carry the day: "Tue 14:30 CT" for the past week, "Sep 9" for older ones; today's rows keep the bare time.
• Tickers in an alert's title and line wear the same chip they wear everywhere else, and a repeated alert says "twice" or "3 times".
• The line you are meant to read is set in the brighter secondary text rather than the muted one, and nothing on a resolved row is struck through.
• A position that differs from the broker's records reads as one sentence ("0 on the books but 1.26 at the broker, and no order of ours explains it"), with the questions worth asking behind Details.
• When the broker refuses to move a stop, the row says where the stop stays and, behind Details, where it was being moved to; while the move is still being retried a small "Retrying every 60 s" pill sits next to it.
• When the platform changes how it reads a finished bar, the desk is told for a day and a half that an earlier reading of the same bar may have shown a different number and that this is a restatement, not a data problem, so it does not report one.
• The four alert rows from September 15 that prompted this work are re-worded and re-routed when read; nothing in the stored history was rewritten.

Known gaps / coming next

• Activity keeps the bare time in its leading column because its rows already sit under day headings; its expanded view will adopt the same folded note the Overview shows in a later release.
• More pairs of "problem" and "back to normal" alerts (the trading service stalling and recovering, the live price stream dropping and returning) will fold the same way once each pair carries the key that ties its two rows together.
• A single missed check stays informational even when the one before it was also missed; the platform team is paged on repeats separately.