ANS-100/P v2.1 — Ratification & Freeze Record

30 Aug 2026 · Batch-1 amendments A1–A10 ratified by author · v2.1 FROZEN as the anchor set for the blind second ratings

1. Ratification record

All ten Batch-1 amendments ratified 30 Aug 2026: A1 ☑ A2 ☑ A3 ☑ A4 ☑ A5 ☑ A6 ☑ A7 ☑ A8 ☑ A9 ☑ A10 ☑.

v2.1 = Index_Mapping_and_Anchors_v2_ACTUAL + Batch-1 amendment text, specifically: the A1 Q25 attribution test, A2 Q30 segment/fixed-price rules, A3 Q18 gradation, A4 Q24 ≥5%-of-TTM-revenue bright line, A5 analyst-evidence rule, A6 N→P reclassification rule (new §2a), A7 countable level-3 test for P-questions, A8 boundary-marker publication rule, A9 pack-completeness floor (step 0), A10 disclosure-bias statement (methodology page).

Freeze condition: no further anchor amendments until both blind second ratings are complete. Divergences found by rater 2 are logged for Batch 2 — they do not modify v2.1 mid-test. The inter-rater test must measure the anchors, not a moving target.

2. Amendment note — append to ANS100P_Infosys_Calibration_File

Amended under Batch 1 (v2.1), 30 Aug 2026. Q18 re-scored 2 → 3 under A3 (audited ISO/IEC 42001 management system covering own operations, plus open-sourced Responsible AI Toolkit; the certification artifact also supports Q22 per the A3 multi-question rule — noted). D4: 58% → 67% (weighted 8.75 → 10.00). Q25:2, Q30:2 confirmed by rule (A1, A2). Q24:3 confirmed under the A4 bright line. All level-3 P-question scores pass the A7 countable test; criterion (c) now independently satisfied by the Everest Group Marketing Transformation PEAK Matrix 2026 Leader placement (added to pack under A9 spot-check — the prior third-party citation was outside the 18-month window). Pack A9-complete.

3. Amendment note — append to ANS100P_TCS_CompanyFile_R1

Amended under Batch 1 (v2.1), 30 Aug 2026. No score changes. Q25:2 pinned by A1 (headcount-driven RPE without workflow attribution = 2 regardless of magnitude). Q30:2 pinned by A2 (segment-level disclosure; segment weight undisclosed). Q18:2 pinned by A3 (productised framework = level 2 by definition; annual-report review under A9 confirms no enforced own-posture disclosure — this is now a pack-complete conclusion and TCS's clearest right-of-reply upside, worth +1.25). Q24:4 pinned by A4 (HyperVault committed amounts ≈ 20%+ of TTM revenue, well above the 5% line; annual report confirms subsidiary consolidation). All 3s pass the A7 countable test (four client functions; Avasant third-party grade). Honest range tightens to ~62–67. Publishes with the A8 boundary marker pending blind-rating confirmation. Pack A9-complete (FY26 Integrated Annual Report added; Q2 FY26 release added; Q4 FY26 transcript to archive).

4. Amendment note — append to ANS100P_HCLTech_CompanyFile_R1

Amended under Batch 1 (v2.1) and the A9 completeness pull, 30 Aug 2026. Three changes: Q18 2 → 3 under A3 (Schellman-audited ISO/IEC 42001 with published scope across own delivery model; artifact also supports Q22 per the A3 rule — noted); Q34 admitted at 3 under A6 (CEO's public disclosure of the modelled 2–3% annual services deflation; N→P reclassification for this company only, P-cap applied); Q21 2 → 3 under the A9 pull (FY26 Integrated Annual Report: 144,223 employees trained in AI, proficiency progression, 715 Blackbelts, AI Institute six-level pathways, Everest talent-readiness Leader). Q14 confirmed at 2 on a complete pack — flag withdrawn, absence genuine, targeted right-of-reply item. Q27:2 pinned by A5 (single root source, showcase deployments). Q24:3 pinned by A4 (₹3,500 Cr ≈ 2.9% of TTM revenue, below the 5% line). All 3s pass the A7 countable test. D4 50% → 58% · D5 69% → 75% · D7 70% → 70.8%. Honest range tightens to ~61–67. Publishes with the A8 boundary marker. Pack A9-complete; LUMI, Talent Navigator metrics, and Responsible AI Transparency Report logged as unscored pack additions for rater 2.

5. Frozen state of edition one, 30 Aug 2026

The #2/#3 gap of 0.07 points is below instrument resolution; published order within band is alphabetical (A8), so the tie does not affect edition one's format. It affects only the internal narrative, and the blind ratings plus right-of-reply will settle it — TCS's upside is an attributable confirmation of the 80% outcome-based figure and any own-posture responsible-AI disclosure; HCLTech's is publishing its eval practice against Q14.

6. Rater-2 handoff protocol (step 5 of the sequence)

1. Rater 2 receives, per company: the evidence pack only — evidence items with dates and sources, the v2.1 anchor set, and the observability map. Not the company files, not the totals, not the comparison notes (Pull Memo §7 control, adopted).
2. Rater 2 scores all three companies against v2.1; the 8-point tolerance per company decides whether the anchors hold.
3. Practical note on who rater 2 can be: rater 1 (this assistant, this conversation) is contaminated for all three companies. A fresh, context-free session given only the blinded packs can serve as a mechanical second rater, but a same-model rater may share rater 1's judgment tendencies — a human second rater is the stronger inter-rater test, and the stronger claim on the methodology page. Ideal: both.
4. Log divergences by question; Batch 2 drafts only after both ratings land.

7. Open items after this freeze (in order)

1. Universe sign-off: rulings 4 and 5 in the Edition-One Universe draft (operating-control exclusion; private cap), freeze date, listing verification snapshot.
2. Blind second ratings (three companies) → anchors hold/fail decision.
3. Companies #3 onward under frozen v2.1 (Wipro next; the IT slice completes the Exposure-ceiling test at full width).
4. Right-of-reply notices drafted from the company files' flagged items once the blind ratings land.