ANS-100/P — Batch-2 Amendments + E12 Ratification Log
Drafted 30 Aug 2026 · Rater 1 · Ratification log at §1; threshold review at §2 · v2.1 frozen; items ratified as logged at §1
1. E12 — ratified 30 Aug 2026 (pharma A9 floor extension)
Rule as ratified: for pharmaceutical companies, the A9 pack floor gains a mandatory publications-and-registry pull before scoring — company-affiliated peer-reviewed publications, clinical-trial registry entries, and CDSCO/USFDA submission disclosures. Company-affiliated publications count at third-party grade for A7(c). Providers (Apollo, Max) are exempt; their evidence lives in investor and operational channels.
Executed for Sun Pharma this date. Result: clean negative. No Sun Pharma-affiliated AI publications were located, and Sun Pharma does not appear among companies with named clinical-stage AI-discovered or AI-enabled candidates — a list that runs to Insilico, Recursion, Schrödinger, XtalPi, Absci and others. Its five clinical-stage molecules (Ilumya, Fibromun, Nidlegy, GL0034, MM-II) are in-licensed or conventionally derived. The E10 gap I flagged was real and worth closing; for this company it did not change the picture, which is itself the useful finding. Cipla, Dr. Reddy's and Divi's now carry the pull as a floor item.
2. Threshold review — the ceilings have never fired, and the arithmetic explains why
Trigger: both economic ceilings have landed at exactly 25.0% four times (Zerodha, Bajaj Auto, UltraTech, Sun Pharma) without firing, since the caps require below 25%.
Finding 1 — 25% is not a threshold; it is the arithmetic floor. A dimension where every scored question receives 1 computes to exactly 1/4 = 25%. Because /P scoring has in practice used 1 as the minimum ("stated intent only"), rather than 0 ("no public evidence of activity"), no dimension can fall below 25% and no economic ceiling can ever fire. The four occurrences are not coincidence; they are the floor showing through.
Finding 2 — even if they fired, they would not bite. The Outcome ceiling caps a total at 55 and the Exposure ceiling at 50. Every company whose D6 or D7 approached 25% scored between 25 and 30 overall. A cap at 50–55 is irrelevant to a company scoring 28. The ceilings can only constrain a company with strong D1–D5 and genuinely absent D6/D7 — the "much announced, nothing landed" profile the anchors doc predicted. Across 36 companies, that profile has not appeared: leadership and platform disclosure and economic disclosure have moved together, not apart.
Finding 3 — a rater-1 scoring habit is implicated, and I should name it. The anchor set makes 0 available for "no public evidence of activity in this dimension." I have used 1 as a practical floor throughout, on the reasoning that absence should be flagged rather than punished. That is defensible under the methodology's absence-of-evidence rule, but it is a systematic choice that was never ratified, and it is the direct cause of Finding 1. It should be decided explicitly rather than left as a habit.
Proposed amendments (ratify individually)
B2-1 — 0-vs-1 discipline. Score 0 where a pack that meets the A9 floor contains no evidence of activity of any kind on a question, and reserve 1 for evidenced intent, announcement, or a single unsubstantiated claim. Files must state which questions received 0 and why. Effect: dimensions can fall below 25%; the ceilings become operative; scores for the four floor companies fall by roughly 2–5 points each; no band changes are expected (all four are already AI Assisted). Ratify: ☐
B2-2 — cap levels. If B2-1 is ratified, revisit the cap levels: an Outcome cap at 55 and Exposure cap at 50 are set well above where sub-25% dimensions actually occur. Proposed: Outcome cap 45, Exposure cap 40, so that a triggered ceiling changes an outcome rather than decorating it. Ratify: ☐
B2-3 — or, the honest alternative. If B2-1 and B2-2 are declined, publish the finding: state on the methodology page that the economic ceilings are designed for the Full instrument and are non-operative in /P, because public-evidence scoring has a floor that the ceiling thresholds sit below. This is the Exposure-ceiling precedent from the IT slice — a pre-registered mechanism that did not fire, disclosed rather than quietly retained. Ratify: ☐
B2-4 — D4 advisory ceiling (logged from manufacturing). Q6 never exceeded 2 in eight manufacturing files because D4 caps advisory and predictive systems at 2. Tata Steel's 860 production models and JSW's 2,900-asset platform both hit that cap. Proposed for consideration: allow level 3 where predictive systems operate at disclosed scale in delivery-critical processes even without actuation authority, on the reasoning that process industries deliberately keep humans in the loop. Ratify: ☐
Recommendation
B2-1 plus B2-3. Tighten the 0-vs-1 discipline and publish the ceiling finding — because even with B2-1 the ceilings will rarely bite, and the honest disclosure is worth more than a mechanism that fires only in theory. B2-2 is a judgment call best made after edition one's full distribution is visible. Applying B2-1 retroactively means re-scoring four to six files; applying it from edition two means disclosing that edition one used a 1-floor. Either is defensible; the former is cleaner and the re-score cost is small.
3. Sequencing note
Batch-2 does not unfreeze v2.1 for the remaining companies. If B2-1 is ratified, it applies from the ratification date forward, with the four to six affected files re-scored and carrying an amendment note — the Batch-1 mechanism exactly. Remaining sectors (healthcare 5, retail 6, telecom/energy 3) continue under v2.1 plus whatever Batch-2 items are ratified.
Published 9 September 2026 under CC BY-SA 4.0. Status changed from "proposed" to "ratified" on publication per the v2.1 Ratification & Freeze Record of 30 August 2026. No text is changed.