Fewer false alarms about positions, cheaper desk sessions, steadier live prices

What changed for people using Bellwether

• Fewer false "positions differ from the broker" alerts. While the desk's own order is filling, the books and the broker differ for a moment by exactly the shares just filled; that no longer rings your phone. An alert arrives only when nothing of ours explains the difference, and it names any open orders that do not account for it.
• Desk sessions cost less for the same decisions. The part of the desk's instructions that is identical for every book is now sent once and reused across books and check-ins instead of being paid for each time. Nothing about what the desk reads, decides or proposes changed, and no model was downgraded.
• Admin › Usage says where the money goes. A new Efficiency section lists each kind of session (research brief, morning plan, intraday check, Critic review, Consult, the shared market brief) with its count, spend and cost per session, plus one sentence on how many check-ins were skipped on purpose and why.
• The go-live scorecard judges AI cost sensibly. Cost per filled order punished a desk that correctly sat out. The scorecard now reads cost per decision against a ceiling, and yearly running cost as a share of the book against an aim; each says "Not measured" until there is enough to judge.
• Live prices stay on the stream. Every five minutes the Executor checks whether a starter buy would pass right now. That check no longer asks the data vendor for quotes on every watchlist name of every book; it prices only what the book can act on, so the vendor's allowance goes to the names you hold.

What you may notice

• Once skipping unchanged check-ins is switched on for a book, Overview's status line can read "last looked 10:34 CT, checked again 10:45 CT: nothing changed, so that look was skipped, next 11:00 CT". It is off for every book today.
• The go-live page's rows are sentences with a status word (Green, Amber, Red, Not measured, For information) and figures like "$0.61 each; the ceiling is $0.80"; the thresholds list reads in words.
• Admin › Usage's per-book table is headed "By book"; usage counts say their units once under the table; an empty cell reads "n/a".
• A "positions differ from the broker" alert that does arrive may add "(still differing, reported earlier: XOM)" rather than repeating numbers it already sent.
• The help article on paper versus live money describes the scorecard's four states and the two AI-cost rows; its screenshot was re-shot.

Known gaps and what’s next

• A daily AI budget that grows with the number of books, money kept aside for end-of-day checks, and an Admin panel that says when a desk is paused for budget arrive next.
• Three cost savers are built and off, each with a dated test that decides whether it stays: one shared market brief a day, skipping a quiet check-in, and reading several names' signals at once.
• For a few days the live-price stream carries more watchlist symbols than it needs; the extras are dropped after a check at the open.