Skip to main content

Pre-mint producer sweep: the $10–$100 equities under the price-derived grid

Status: MEASURED 2026-08-25, against the live kind rig (:yu17-markwait2) and the operative scripts/ tree, per format-8-price-derived-grid-design.md Β§5 ("Pre-mint sweep") and prove-cluster-engine-change Β§1b. This is a report, not fixes; the fixes belong to the mint chip. Written by the proof-set chip alongside the Β§5 red halves.

The question. The seven equities the map moves from tick 1000 to tick 100; DB, UBS, BAC, C, FNF, FIS, MS; have their half-band tighten from Β±$65.54 to Β±$6.5536 (65 536 Γ— 100 Px). Any /orders producer under scripts/ that rests one of them further than the new band from the live feed reference goes red at the mint. Live references at measurement (publisher, 00:24Z):

tickerliveopen/close (committed)new band around live
DB16.61216.80 / 17.05[10.06, 23.17]
UBS29.78229.40 / 29.90[23.23, 36.34]
BAC42.95841.10 / 41.70[36.40, 49.51]
C60.38459.20 / 58.60[53.83, 66.94]
FNF49.50049.50 / 49.95[42.95, 56.05]
FIS75.97674.70 / 75.20[69.42, 82.53]
MS90.62390.50 / 91.10[84.07, 97.18]

Hits​

1. scripts/proofs/yu03-risk-proof.sh:90; BAC rested at a literal 40 (CONDITIONAL, the only hit)​

{"security":"BAC","side":"Buy","quantity":11,"limitPrice":40}, rested three times per run (lines 112, 116 expect NEW; line 114 expects RESTRICTED), seeded at BAC:40 (line 104).

  • Arithmetic: |40 βˆ’ 42.958 live| = 2.96 < 6.55; inside the new band today. The break arrives when the live reference leaves [33.45, 46.55]; against the committed close 41.70 the headroom is $4.85 of publisher walk.
  • Why it is conditional and not clean: the script's own comment (lines 99–102) says the seed pins the reference at the traded price. Since the feed adapter went live (2026-08-24) that is no longer true; the adapter re-sequences the publisher's BAC within one flush, so the proof's correctness silently depends on the publisher's walk staying within Β±$6.55 of the literal 40. Under the current Β±$65.54 band that coupling is invisible; the mint makes it a ~$5-headroom bet.
  • For the mint chip: order (and seed) BAC at the live publisher price the way seed-proof-fixtures.sh does (live_px), or move the restricted-security arm to a β‰₯$100 ticker whose band the mint does not change. Also update the stale lines 99–102 while there (message/mechanism rule: the comment asserts a pinning the adapter has revoked).

yu03's other arms are unaffected: IBM@190/200 (Ξ” ≀ 15.4, band unchanged at Β±$65.54) and the IBM@400 PRICE_COLLAR arm (Ξ” 215, refused on both builds).

The seeder's hold() crossings; claim VERIFIED, not inherited​

scripts/yu15/seed-proof-fixtures.sh:231–233 crosses NVDA/AAPL/IBM with both legs at live_px <ticker> (read from the same publisher /prices payload the adapter sequences): distance from reference β‰ˆ 0, safe under any band; and all three are β‰₯$100 instruments whose grid the mint does not move anyway. Two caveats recorded, neither a mint blocker:

  • live_px falls back to a literal 200 for a ticker absent from the feed (line ~229 awk END {if (!f) print "200"}). All three holds would then be refused; loudly, since 2026-08-22 hold() treats a refused leg as fatal. Pre-existing, unchanged by the mint.
  • The seeder's FIRST pass (line 99) ticks all 20 fixture tickers flat at PRICE:-200; for DB that is 12Γ— the live price; before the same script re-seeds the whole quoted universe at live prices and the adapter resumes. Post-mint wrinkle, benign: a book created inside that seconds-wide window on a $10–$100 ticker would derive tick 1000 from the transient 200 reference; no producer orders in the window (the holds run after the live re-seed), and Β§2.3's empty-book re-derivation heals it at the next empty admission. Noted so the mint chip does not rediscover it as a mystery tickDrift.

Producers checked and clean (the population, per vacuous-pass-audit rule 2)​

Every /orders producer under scripts/ was enumerated (41 files) and each was checked for resting prices on the seven tickers. None besides yu03 names them with a literal price:

  • scripts/sim/session.mjs; quotes RELATIVE to the prevailing price (its own stated rule 4: "a hardcoded limit is how a script rejects itself"), probes each anchor AT the published price, rounds to cents; and cents sit on every producible map tick (each divides 10 000 Px). Clean.
  • scripts/proofs/; yu05-settlement (IBM 190/200), yu10-fix-session (IBM @ live), yu13-, yu15-, yu16-* (minted RM…/DUP…/STP…/BND… throwaways, bonds, or IBM/AAPL), seed-option-chain (options at live premium Β± cents), yu12-gke-* / yu13-gke-* (other rig, and IBM/minted tickers). None rest the seven.
  • scripts/bench/; loaders default to JPM/GS/COF/DFS (live 195.7/397.4/153.3/126.7; all β‰₯$100, band unchanged by the mint) with env-overridable tickers; TICKERS=DB etc. would inherit the tightened band, an operator note, not a code hit.
  • scripts/yu15/demo-otel-traffic.sh; AAPL only (band unchanged).

Out-of-scope observations (pre-existing under ADR-066, not mint deltas; left alone)​

  • max-load.mjs defaults LIMIT=1_000_000 on β‰₯$100 tickers: already refused PRICE_COLLAR by today's engine band on the cluster tier (the bench predates ADR-066 and its comments describe risk-gate 4xxs as the expected wall).
  • demo-otel-traffic.sh seeds its universe flat at 150 and orders AAPL@150 against a live ~238: Ξ” 88 > today's Β±$65.54; the same adapter-overrides-the-seed mechanism as the yu03 hit, already live at the current band. Filed here as context for whoever next runs that demo; not a format-8 regression.