Release notes · current model NAIR v0.6

What changed, and why.

Every change to the NAIR model, the data, the API and the site. Model releases show exactly what they did to the published figures. Follow along by RSS or JSON, or in the weekly newsletter.

9 Nov 2026Data

Coverage 118 → 332: Lowe's Companies, Equinix, Phillips 66, McKesson and 210 more

214 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Lowe's Companies (LOW): Not testable, low confidence, AI spend $252M counted as unverified spend.
Added Equinix (EQIX): Not testable, low confidence, AI spend $41M counted as unverified spend.
Added Phillips 66 (PSX): Not testable, low confidence, AI spend $24M counted as unverified spend.
Added McKesson (MCK): Not testable, low confidence, AI spend $81M counted as unverified spend.
Added The Bank of New York Mellon (BNY): Not testable, low confidence, AI spend $98M counted as unverified spend.
Added The Southern Company (SO): Not testable, low confidence, AI spend $71M counted as unverified spend.
Added Vertiv Holdings Co (VRT): Not testable, low confidence, AI spend $88M counted as unverified spend.
Added Quanta Services (PWR): Not testable, low confidence, AI spend $22M counted as unverified spend.
Added CME Group (CME): Not testable, low confidence, AI spend $16M counted as unverified spend.
Added Airbnb (ABNB): Not testable, low confidence, AI spend $471M counted as unverified spend.
Added Trane Technologies (TT): Not testable, low confidence, AI spend $70M counted as unverified spend.
Added HCA Healthcare (HCA): Not testable, low confidence, AI spend $366M counted as unverified spend.
Added Adobe (ADBE): Partly visible, medium confidence, net AI return -$1.3B.
Added Constellation Energy (CEG): Not testable, low confidence, AI spend $62M counted as unverified spend.
Added General Dynamics (GD): Not testable, low confidence, AI spend $73M counted as unverified spend.
Added U.S. Bancorp (USB): Not testable, low confidence, AI spend $168M counted as unverified spend.
Added The PNC Financial Services Group (PNC): Not testable, low confidence, AI spend $138M counted as unverified spend.
Added Duke Energy (DUK): Not testable, low confidence, AI spend $67M counted as unverified spend.
Added Marriott International (MAR): Not testable, low confidence, AI spend $9M counted as unverified spend.
Added Howmet Aerospace (HWM): Not testable, low confidence, AI spend $7M counted as unverified spend.
Added Datadog (DDOG): Not testable, low confidence, AI spend $387M counted as unverified spend.
Added Johnson Controls International (JCI): Not testable, low confidence, AI spend $55M counted as unverified spend.
Added Monster Beverage (MNST): Not testable, low confidence, AI spend $17M counted as unverified spend.
Added 3M Company (MMM): Not testable, low confidence, AI spend $234M counted as unverified spend.
Added Emerson Electric Co. (EMR): Not testable, low confidence, AI spend $154M counted as unverified spend.
Added Intercontinental Exchange (ICE): Not testable, low confidence, AI spend $34M counted as unverified spend.
Added CSX (CSX): Not testable, low confidence, AI spend $63M counted as unverified spend.
Added Lumentum Holdings (LITE): Not testable, low confidence, AI spend $53M counted as unverified spend.
Added DoorDash (DASH): Not testable, low confidence, AI spend $286M counted as unverified spend.
Added Regeneron Pharmaceuticals (REGN): Not testable, low confidence, AI spend $878M counted as unverified spend.
Added Cadence Design Systems (CDNS): Not testable, low confidence, AI spend $442M counted as unverified spend.
Added WM (Waste Management, Inc.) (WM): Not testable, low confidence, AI spend $41M counted as unverified spend.
Added American Tower (AMT): Not testable, low confidence, AI spend $9M counted as unverified spend.
Added Bloom Energy (BE): Not testable, low confidence, AI spend $37M counted as unverified spend.
Added Hewlett Packard Enterprise Company (HPE): Partly visible, medium confidence, net AI return -$296M.
Added Moody's (MCO): Not testable, low confidence, AI spend $27M counted as unverified spend.
Added Marsh McLennan Companies, Inc. (branded Marsh from January 2026) (MRSH): Not testable, low confidence, AI spend $208M counted as unverified spend.
Added Comcast (CMCSA): Not testable, low confidence, AI spend $519M counted as unverified spend.
Added Cintas (CTAS): Not testable, low confidence, AI spend $31M counted as unverified spend.
Added Synopsys (SNPS): Not testable, low confidence, AI spend $620M counted as unverified spend.
Added Simon Property Group (SPG): Not testable, low confidence, AI spend $3M counted as unverified spend.
Added Mondelez International (MDLZ): Not testable, low confidence, AI spend $80M counted as unverified spend.
Added Ecolab (ECL): Not testable, low confidence, AI spend $40M counted as unverified spend.
Added Warner Bros. Discovery (WBD): Not testable, low confidence, AI spend $94M counted as unverified spend.
Added Illinois Tool Works (ITW): Not testable, low confidence, AI spend $45M counted as unverified spend.
Added Motorola Solutions (MSI): Not testable, low confidence, AI spend $242M counted as unverified spend.
Added General Motors Company (GM): Not testable, low confidence, AI spend $1.7B counted as unverified spend.
Added The Travelers Companies (TRV): Not testable, low confidence, AI spend $92M counted as unverified spend.
Added Ross Stores (ROST): Not testable, low confidence, AI spend $27M counted as unverified spend.
Added Apollo Global Management (APO): Not testable, low confidence, AI spend $60M counted as unverified spend.
Added EOG Resources (EOG): Not testable, low confidence, AI spend $12M counted as unverified spend.
Added Carvana Co. (CVNA): Not testable, low confidence, AI spend $35M counted as unverified spend.
Added Moderna (MRNA): Not testable, low confidence, AI spend $626M counted as unverified spend.
Added The Cigna Group (CI): Not testable, low confidence, AI spend $146M counted as unverified spend.
Added Northrop Grumman (NOC): Not testable, low confidence, AI spend $220M counted as unverified spend.
Added Cummins (CMI): Not testable, low confidence, AI spend $279M counted as unverified spend.
Added Norfolk Southern (NSC): Not testable, low confidence, AI spend $46M counted as unverified spend.
Added Digital Realty Trust (DLR): Not testable, low confidence, AI spend $6M counted as unverified spend.
Added FedEx (FDX): Not testable, low confidence, AI spend $478M counted as unverified spend.
Added O'Reilly Automotive (ORLY): Not testable, low confidence, AI spend $43M counted as unverified spend.
Added Kinder Morgan (KMI): Not testable, low confidence, AI spend $6M counted as unverified spend.
Added Hilton Worldwide Holdings (HLT): Not testable, low confidence, AI spend $4M counted as unverified spend.
Added Colgate-Palmolive Company (CL): Not testable, low confidence, AI spend $92M counted as unverified spend.
Added Monolithic Power Systems (MPWR): Not testable, low confidence, AI spend $57M counted as unverified spend.
Added American Electric Power Company (AEP): Not testable, low confidence, AI spend $67M counted as unverified spend.
Added United Rentals (URI): Not testable, low confidence, AI spend $17M counted as unverified spend.
Added Boston Scientific (BSX): Not testable, low confidence, AI spend $308M counted as unverified spend.
Added Air Products and Chemicals (APD): Not testable, low confidence, AI spend $14M counted as unverified spend.
Added Royal Caribbean Cruises (RCL): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Teradyne (TER): Not testable, low confidence, AI spend $76M counted as unverified spend.
Added TE Connectivity (TEL): Not testable, low confidence, AI spend $124M counted as unverified spend.
Added Targa Resources (TRGP): Not testable, low confidence, AI spend $3M counted as unverified spend.
Added TransDigm Group (TDG): Not testable, low confidence, AI spend $18M counted as unverified spend.
Added Aon (AON): Not testable, medium confidence, AI spend $284M counted as unverified spend.
Added Cencora (COR): Not testable, low confidence, AI spend $65M counted as unverified spend.
Added W. W. Grainger (GWW): Not testable, low confidence, AI spend $68M counted as unverified spend.
Added PACCAR (PCAR): Not testable, low confidence, AI spend $89M counted as unverified spend.
Added NXP Semiconductors (NXPI): Not testable, low confidence, AI spend $472M counted as unverified spend.
Added Keysight Technologies (KEYS): Not testable, low confidence, AI spend $151M counted as unverified spend.
Added Arthur J. Gallagher & Co. (AJG): Not testable, low confidence, AI spend $76M counted as unverified spend.
Added CRH (CRH): Not testable, low confidence, AI spend $62M counted as unverified spend.
Added The Allstate (ALL): Not testable, low confidence, AI spend $112M counted as unverified spend.
Added ONEOK (OKE): Not testable, low confidence, AI spend $19M counted as unverified spend.
Added Fastenal Company (FAST): Not testable, low confidence, AI spend $20M counted as unverified spend.
Added Comfort Systems USA (FIX): Not testable, low confidence, AI spend $7M counted as unverified spend.
Added Occidental Petroleum (OXY): Not testable, low confidence, AI spend $10M counted as unverified spend.
Added AMETEK (AME): Not testable, low confidence, AI spend $47M counted as unverified spend.
Added Nucor (NUE): Not testable, low confidence, AI spend $9M counted as unverified spend.
Added Delta Air Lines (DAL): Not testable, low confidence, AI spend $270M counted as unverified spend.
Added Garmin (GRMN): Not testable, low confidence, AI spend $169M counted as unverified spend.
Added NIKE (NKE): Not testable, low confidence, AI spend $121M counted as unverified spend.
Added Coinbase Global (COIN): Not testable, low confidence, AI spend $418M counted as unverified spend.
Added Nasdaq (NDAQ): Not testable, low confidence, AI spend $27M counted as unverified spend.
Added Ciena (CIEN): Not testable, low confidence, AI spend $170M counted as unverified spend.
Added Ford Motor Company (F): Not testable, low confidence, AI spend $1.4B counted as unverified spend.
Added Diamondback Energy (FANG): Not testable, low confidence, AI spend $3M counted as unverified spend.
Added Devon Energy (DVN): Not testable, low confidence, AI spend $7M counted as unverified spend.
Added Cardinal Health (CAH): Not testable, low confidence, AI spend $61M counted as unverified spend.
Added Edwards Lifesciences (EW): Not testable, low confidence, AI spend $162M counted as unverified spend.
Added Becton, Dickinson and Company (BDX): Not testable, low confidence, AI spend $253M counted as unverified spend.
Added State Street (STT): Not testable, low confidence, AI spend $102M counted as unverified spend.
Added eBay (EBAY): Not testable, low confidence, AI spend $328M counted as unverified spend.
Added Westinghouse Air Brake Technologies Corporation (Wabtec) (WAB): Not testable, low confidence, AI spend $45M counted as unverified spend.
Added Entergy (ETR): Not testable, low confidence, AI spend $31M counted as unverified spend.
Added Vistra (VST): Not testable, low confidence, AI spend $17M counted as unverified spend.
Added Agilent Technologies (A): Not testable, low confidence, AI spend $91M counted as unverified spend.
Added Block (XYZ): Not testable, low confidence, AI spend $727M counted as unverified spend.
Added Autodesk (ADSK): Not testable, low confidence, AI spend $411M counted as unverified spend.
Added Workday (WDAY): Partly visible, medium confidence, net AI return -$404M.
Added Carrier Global (CARR): Not testable, low confidence, AI spend $125M counted as unverified spend.
Added PayPal Holdings (PYPL): Not testable, low confidence, AI spend $621M counted as unverified spend.
Added Xcel Energy (XEL): Not testable, low confidence, AI spend $27M counted as unverified spend.
Added Humana (HUM): Not testable, low confidence, AI spend $154M counted as unverified spend.
Added IQVIA Holdings (IQV): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added Exelon (EXC): Not testable, low confidence, AI spend $52M counted as unverified spend.
Added Ferguson Enterprises (FERG): Not testable, low confidence, AI spend $64M counted as unverified spend.
Added Veeva Systems (VEEV): Not testable, low confidence, AI spend $192M counted as unverified spend.
Added Keurig Dr Pepper (KDP): Not testable, low confidence, AI spend $14M counted as unverified spend.
Added Waters (WAT): Not testable, low confidence, AI spend $39M counted as unverified spend.
Added Flex (FLEX): Not testable, low confidence, AI spend $11M counted as unverified spend.
Added Chipotle Mexican Grill (CMG): Not testable, low confidence, AI spend $7M counted as unverified spend.
Added IDEXX Laboratories (IDXX): Not testable, low confidence, AI spend $63M counted as unverified spend.
Added Microchip Technology (MCHP): Not testable, low confidence, AI spend $217M counted as unverified spend.
Added CBRE Group (CBRE): Not testable, low confidence, AI spend $55M counted as unverified spend.
Added Paychex (PAYX): Not testable, low confidence, AI spend $35M counted as unverified spend.
Added Archer-Daniels-Midland Company (ADM): Not testable, low confidence, AI spend $49M counted as unverified spend.
Added D.R. Horton (DHI): Not testable, low confidence, AI spend $37M counted as unverified spend.
Added MSCI (MSCI): Not testable, low confidence, AI spend $44M counted as unverified spend.
Added Live Nation Entertainment (LYV): Not testable, low confidence, AI spend $41M counted as unverified spend.
Added PG&E (PCG): Not testable, low confidence, AI spend $113M counted as unverified spend.
Added Take-Two Interactive Software (TTWO): Not testable, low confidence, AI spend $161M counted as unverified spend.
Added Yum! Brands (YUM): Not testable, low confidence, AI spend $19M counted as unverified spend.
Added NetApp (NTAP): Not testable, low confidence, AI spend $198M counted as unverified spend.
Added Sysco (SYY): Not testable, low confidence, AI spend $125M counted as unverified spend.
Added United Airlines Holdings (UAL): Not testable, low confidence, AI spend $297M counted as unverified spend.
Added Illumina (ILMN): Not testable, low confidence, AI spend $193M counted as unverified spend.
Added Axon Enterprise (AXON): Not testable, low confidence, AI spend $171M counted as unverified spend.
Added Everpure, Inc. (formerly Pure Storage) (P): Not testable, low confidence, AI spend $193M counted as unverified spend.
Added Old Dominion Freight Line (ODFL): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added Roper Technologies (ROP): Not testable, low confidence, AI spend $213M counted as unverified spend.
Added The Estée Lauder Companies (EL): Not testable, low confidence, AI spend $56M counted as unverified spend.
Added TKO Group Holdings (TKO): Not testable, low confidence, AI spend $11M counted as unverified spend.
Added Martin Marietta Materials (MLM): Not testable, low confidence, AI spend $3M counted as unverified spend.
Added Iron Mountain (IRM): Not testable, low confidence, AI spend $21M counted as unverified spend.
Added The Hartford Insurance Group (HIG): Not testable, low confidence, AI spend $70M counted as unverified spend.
Added The Kroger Co. (KR): Not testable, low confidence, AI spend $258M counted as unverified spend.
Added Public Service Enterprise Group (PEG): Not testable, low confidence, AI spend $28M counted as unverified spend.
Added Kenvue (KVUE): Not testable, low confidence, AI spend $57M counted as unverified spend.
Added DexCom (DXCM): Not testable, low confidence, AI spend $120M counted as unverified spend.
Added The Hershey Company (HSY): Not testable, low confidence, AI spend $12M counted as unverified spend.
Added WEC Energy Group (WEC): Not testable, low confidence, AI spend $18M counted as unverified spend.
Added Steel Dynamics (STLD): Not testable, low confidence, AI spend $6M counted as unverified spend.
Added Biogen (BIIB): Not testable, low confidence, AI spend $267M counted as unverified spend.
Added EMCOR Group (EME): Not testable, low confidence, AI spend $13M counted as unverified spend.
Added Kimberly-Clark (KMB): Not testable, low confidence, AI spend $49M counted as unverified spend.
Added ResMed (RMD): Not testable, low confidence, AI spend $95M counted as unverified spend.
Added Expedia Group (EXPE): Not testable, low confidence, AI spend $192M counted as unverified spend.
Added EQT (EQT): Not testable, low confidence, AI spend $3M counted as unverified spend.
Added Northern Trust (NTRS): Not testable, low confidence, AI spend $86M counted as unverified spend.
Added Vulcan Materials Company (VMC): Not testable, low confidence, AI spend $4M counted as unverified spend.
Added Crown Castle (CCI): Not testable, low confidence, AI spend $4M counted as unverified spend.
Added Centene (CNC): Not testable, low confidence, AI spend $129M counted as unverified spend.
Added Raymond James Financial (RJF): Not testable, low confidence, AI spend $85M counted as unverified spend.
Added Carnival (CCL): Not testable, low confidence, AI spend $34M counted as unverified spend.
Added Zoetis (ZTS): Not testable, low confidence, AI spend $140M counted as unverified spend.
Added Extra Space Storage (EXR): Not testable, low confidence, AI spend $9M counted as unverified spend.
Added Mettler Toledo (MTD): Not testable, low confidence, AI spend $40M counted as unverified spend.
Added GE HealthCare (GEHC): Not testable, low confidence, AI spend $315M counted as unverified spend.
Added Reddit (RDDT): Not testable, low confidence, AI spend $196M counted as unverified spend.
Added HP (HPQ): Not testable, low confidence, AI spend $320M counted as unverified spend.
Added Ingersoll Rand (IR): Not testable, low confidence, AI spend $17M counted as unverified spend.
Added ON Semiconductor (ON): Not visible, medium confidence, net AI return -$117M.
Added The Kraft Heinz Company (KHC): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Teledyne Technologies (TDY): Not testable, low confidence, AI spend $63M counted as unverified spend.
Added Ameren (AEE): Not testable, low confidence, AI spend $15M counted as unverified spend.
Added Willis Towers Watson Public Limited Company (WTW) (WTW): Not testable, low confidence, AI spend $70M counted as unverified spend.
Added Cboe Global Markets (CBOE): Not testable, low confidence, AI spend $6M counted as unverified spend.
Added Citizens Financial Group (CFG): Not testable, low confidence, AI spend $53M counted as unverified spend.
Added EchoStar (ECHO): Not testable, low confidence, AI spend $14M counted as unverified spend.
Added Williams-Sonoma (WSM): Not testable, low confidence, AI spend $22M counted as unverified spend.
Added Halliburton Company (HAL): Not testable, low confidence, AI spend $82M counted as unverified spend.
Added Dollar General (DG): Not testable, low confidence, AI spend $82M counted as unverified spend.
Added VeriSign (VRSN): Not testable, low confidence, AI spend $16M counted as unverified spend.
Added Super Micro Computer (SMCI): Paying off, medium confidence, net AI return $12.6B.
Added Atmos Energy (ATO): Not testable, low confidence, AI spend $7M counted as unverified spend.
Added Copart (CPRT): Not testable, low confidence, AI spend $6M counted as unverified spend.
Added American Water Works Company (AWK): Not testable, low confidence, AI spend $15M counted as unverified spend.
Added Cognizant Technology Solutions (CTSH): Not testable, low confidence, AI spend $333M counted as unverified spend.
Added West Pharmaceutical Services (WST): Not testable, low confidence, AI spend $11M counted as unverified spend.
Added DTE Energy Company (DTE): Not testable, low confidence, AI spend $18M counted as unverified spend.
Added Qnity Electronics (Q): Not testable, low confidence, AI spend $71M counted as unverified spend.
Added Quest Diagnostics (DGX): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added FirstEnergy (FE): Not testable, low confidence, AI spend $31M counted as unverified spend.
Added Incyte (INCY): Not testable, low confidence, AI spend $410M counted as unverified spend.
Added Corpay (CPAY): Not testable, low confidence, AI spend $18M counted as unverified spend.
Added Otis Worldwide (OTIS): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added Las Vegas Sands (LVS): Not testable, low confidence, AI spend $11M counted as unverified spend.
Added F5 (FFIV): Not testable, low confidence, AI spend $135M counted as unverified spend.
Added Labcorp Holdings (LH): Not testable, low confidence, AI spend $28M counted as unverified spend.
Added W. R. Berkley (WRB): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added Dover (DOV): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Xylem (XYL): Not testable, low confidence, AI spend $45M counted as unverified spend.
Added Eversource Energy (ES): Not testable, low confidence, AI spend $21M counted as unverified spend.
Added CenterPoint Energy (CNP): Not testable, low confidence, AI spend $23M counted as unverified spend.
Added PPL (PPL): Not testable, low confidence, AI spend $24M counted as unverified spend.
Added Texas Pacific Land (TPL): Not testable, low confidence, AI spend $1M counted as unverified spend.
Added Expeditors International of Washington (EXPD): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Fiserv (FISV): Not testable, low confidence, AI spend $297M counted as unverified spend.
Added Smurfit Westrock (SW): Not testable, low confidence, AI spend $38M counted as unverified spend.
Added Darden Restaurants (DRI): Not testable, low confidence, AI spend $5M counted as unverified spend.
Added Ulta Beauty (ULTA): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Veralto (VLTO): Not testable, low confidence, AI spend $53M counted as unverified spend.
Added Church & Dwight Co. (CHD): Not testable, low confidence, AI spend $29M counted as unverified spend.
Added Tapestry (TPR): Not testable, low confidence, AI spend $43M counted as unverified spend.
8 published companies moved because their peer groups gained members; see the table.

Why it moved the numbers. Coverage is expanding across the S&P 500. New companies can change peer medians, which moves the verified value of companies already published; every such move is listed here.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$84.2B	$102B	+$17.5B
Net return on tested spend	$23.5B	$37.7B	+$14.2B
Untested AI spend	$86B	$112B	+$26.2B

Paying off: 10 of 118 before, 12 of 332 after.

Company	Verified, before	Verified, after	Change	Net AI return
Citigroup C	$195M	$163M	$(32M)	$(455M) → $(487M)
Broadcom AVGO	$7.7B	$8.7B	+$997M	$5B → $6B
Advanced Micro Devices AMD	$2.4B	$2.7B	+$310M	$(54M) → $256M
Dell Technologies DELL	$7.5B	$9.4B	+$1.9B	$6.7B → $8.6B
Arista Networks ANET	$1.1B	$1.2B	+$135M	$742M → $877M
ServiceNow NOW	$600M	$614M	+$14M	$(288M) → $(274M)
Corning GLW	$1B	$1.2B	+$208M	$834M → $1B
AppLovin APP	$1.5B	$1.5B	+$63M	$1.4B → $1.5B

Every company whose verified value moved by more than $1M. All other companies were unchanged.

7 Oct 2026ModelNAIR v0.6

NAIR v0.6: an AI Disclosure Grade for every company

Every company now carries a Disclosure Grade (A to F) that says how clearly it reports AI value: how specific its claims are, whether any can be tested against its statements, where it said them, whether they held up, and how clearly AI is the driver. It is not a value measure. No verified value, net return, verdict or confidence changes.

A Disclosure Grade for every company with at least one logged claim, scored out of 100 on five parts: specificity (30: quantified claims, a dollar-denominated value claim, disclosed AI spend), testability (25: a claim tested against a statement line, and the share of claims Jev rates dollar-testable), sourcing (20: the company's own filings, releases and transcripts; third-party transcripts count half; news counts nothing), consistency (15: minus 5 per contradicted claim, minus 3 per dropped claim) and AI attribution (10: mean Jev probability that AI is the stated driver). A from 80, B from 65, C from 50, D from 35, otherwise F.
The leaderboard now opens on all 118 companies: the 25 tested companies first by net AI return, then the 93 Not testable companies by disclosure score. The tested-only view is one click away. A new Disclosure column is sortable; the company page shows the five parts; the compare table carries the grade.
Companies with fewer than three logged claims carry a dagger (†) because the grade rests on little. Companies with no logged claims are not graded.
API and MCP: every company carries disclosure (grade, score, claims, note, parts); list_companies adds disclosure_grade and disclosure_score; screen_companies accepts disclosure_grade and disclosure_score_gt; sector_benchmarks reports the grade mix. The database adds disclosure_grade and disclosure_score to scores.
Why a grade and not more tests: we sized the alternative, converting percentage claims with a named cost base into dollar figures. Of 137 percentage claims across 59 companies, none passes the attribution rule published in v0.5 with a statement line behind it, so it would have moved no company out of Not testable. Most large companies describe AI value in operational percentages with no line in the financial statements behind them; the grade makes that visible instead of hiding it behind a blank row.

Why it moved the numbers. Nothing moved: the grade is a new published metric and never enters verified value, net return, the verdict or confidence. Distribution on release: A 4, B 20, C 48, D 30, F 8, with 8 companies ungraded for lack of claims. Among the 93 Not testable companies: B 3, C 45, D 29, F 8. Examples: Cisco Systems (CSCO) A 84, Intel (INTC) A 82, Microsoft (MSFT) B 72, Salesforce (CRM) B 73, Alphabet (Class A) (GOOGL) C 58, Meta Platforms (META) C 52, Apple (AAPL) D 36, The Goldman Sachs Group (GS) F 32.

Effect on published figures, v0.5 → v0.6

Total	v0.5	v0.6	Change
Claimed value, tested claims	$149B	$149B	+$0
Verified value, 118 companies	$84.2B	$84.2B	+$0
Net return on tested spend (25 companies)	$23.5B	$23.5B	+$0
Untested AI spend (93 companies)	$86B	$86B	+$0

No company's verified value, net AI return, verdict or confidence changed. The grade is shown beside them.

For the 40 companies researched before the Harvester recorded source types (September 2026), the sourcing part is read from each claim's URL; a re-sourcing pass to primary sources is scheduled and will be logged as a data update.

7 Oct 2026Data

Coverage 109 → 118: S&P Global, Altria Group, Valero Energy, Robinhood Markets and 5 more

9 companies added and 8 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added S&P Global (SPGI): Partly visible, medium confidence, net AI return -$51M.
Added Altria Group (MO): Not testable, low confidence, AI spend $39M counted as unverified spend.
Added Valero Energy (VLO): Not testable, low confidence, AI spend $10M counted as unverified spend.
Added Robinhood Markets (HOOD): Partly visible, medium confidence, net AI return -$39M.
Added AppLovin (APP): Paying off, medium confidence, net AI return $1.4B.
Added Starbucks (SBUX): Not testable, low confidence, AI spend $26M counted as unverified spend.
Added Automatic Data Processing (ADP): Not testable, low confidence, AI spend $257M counted as unverified spend.
Added Freeport-McMoRan (FCX): Not testable, low confidence, AI spend $11M counted as unverified spend.
Added Stryker (SYK): Not testable, low confidence, AI spend $325M counted as unverified spend.
Updated Prologis (PLD) after a new QA review: Not testable → Not testable.
Updated Bristol-Myers Squibb Company (BMY) after a new QA review: Not testable → Not testable.
Updated Capital One Financial (COF) after a new QA review: Not testable → Not testable.
Updated Parker-Hannifin (PH) after a new QA review: Not testable → Not testable.
Updated Booking Holdings (BKNG) after a new QA review: Not testable → Not testable.
Updated Lockheed Martin (LMT) after a new QA review: Not testable → Not testable.
Updated The Progressive (PGR) after a new QA review: Not testable → Not testable.
Updated Marathon Petroleum (MPC) after a new QA review: Not testable → Not testable.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$82.6B	$84.2B	+$1.5B
Net return on tested spend	$22.2B	$23.5B	+$1.3B
Untested AI spend	$85.3B	$86B	+$668M

Paying off: 9 of 109 before, 10 of 118 after.

6 Oct 2026Data

Coverage 101 → 109: Prologis, Bristol-Myers Squibb Company, Capital One Financial, Parker-Hannifin and 4 more

8 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Prologis (PLD): Not testable, low confidence, AI spend $5M counted as unverified spend.
Added Bristol-Myers Squibb Company (BMY): Not testable, low confidence, AI spend $2.0B counted as unverified spend.
Added Capital One Financial (COF): Not testable, low confidence, AI spend $305M counted as unverified spend.
Added Parker-Hannifin (PH): Not testable, low confidence, AI spend $40M counted as unverified spend.
Added Booking Holdings (BKNG): Not testable, low confidence, AI spend $52M counted as unverified spend.
Added Lockheed Martin (LMT): Not testable, low confidence, AI spend $500M counted as unverified spend.
Added The Progressive (PGR): Not testable, low confidence, AI spend $113M counted as unverified spend.
Added Marathon Petroleum (MPC): Not testable, low confidence, AI spend $33M counted as unverified spend.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$82.6B	$82.6B	+$0
Net return on tested spend	$22.2B	$22.2B	+$0
Untested AI spend	$82.3B	$85.3B	+$3B

Paying off: 9 of 101 before, 9 of 109 after.

5 Oct 2026Data

Coverage 92 → 101: McDonald's, NextEra Energy, Uber Technologies, ServiceNow and 5 more

9 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added McDonald's (MCD): Not testable, low confidence, AI spend $30M counted as unverified spend.
Added NextEra Energy (NEE): Not testable, low confidence, AI spend $54M counted as unverified spend.
Added Uber Technologies (UBER): Not testable, low confidence, AI spend $680M counted as unverified spend.
Added ServiceNow (NOW): Partly visible, medium confidence, net AI return -$288M.
Added Corning (GLW): Paying off, low confidence, net AI return $834M.
Added Newmont (NEM): Not testable, low confidence, AI spend $33M counted as unverified spend.
Added Vertex Pharmaceuticals (VRTX): Not testable, low confidence, AI spend $586M counted as unverified spend.
Added Chubb Limited (CB): Not testable, low confidence, AI spend $45M counted as unverified spend.
Added Fortinet (FTNT): Not testable, low confidence, AI spend $204M counted as unverified spend.
4 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$81.9B	$82.6B	+$747M
Net return on tested spend	$22.5B	$22.2B	$(307M)
Untested AI spend	$80.6B	$82.3B	+$1.6B

Paying off: 8 of 92 before, 9 of 101 after.

Company	Verified, before	Verified, after	Change	Net AI return
Broadcom AVGO	$8B	$7.7B	$(245M)	$5.2B → $5B
Advanced Micro Devices AMD	$2.4B	$2.4B	$(39M)	$(15M) → $(54M)
Dell Technologies DELL	$8B	$7.5B	$(535M)	$7.2B → $6.7B
Arista Networks ANET	$1.1B	$1.1B	$(33M)	$776M → $742M

Every company whose verified value moved by more than $1M. All other companies were unchanged.

4 Oct 2026Data

Coverage 84 → 92: Union Pacific, The Boeing Company, Interactive Brokers Group, Danaher and 4 more

8 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Union Pacific (UNP): Not testable, low confidence, AI spend $89M counted as unverified spend.
Added The Boeing Company (BA): Not testable, low confidence, AI spend $542M counted as unverified spend.
Added Interactive Brokers Group (IBKR): Not testable, low confidence, AI spend $10M counted as unverified spend.
Added Danaher (DHR): Not testable, low confidence, AI spend $240M counted as unverified spend.
Added ConocoPhillips (COP): Not testable, low confidence, AI spend $12M counted as unverified spend.
Added Blackstone (BX): Not testable, low confidence, AI spend $52M counted as unverified spend.
Added Intuitive Surgical (ISRG): Not testable, low confidence, AI spend $262M counted as unverified spend.
Added The TJX Companies (TJX): Not testable, low confidence, AI spend $86M counted as unverified spend.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$81.9B	$81.9B	+$0
Net return on tested spend	$22.5B	$22.5B	+$0
Untested AI spend	$79.3B	$80.6B	+$1.3B

Paying off: 8 of 84 before, 8 of 92 after.

3 Oct 2026Data

Coverage 76 → 84: KLA, Abbott Laboratories, PepsiCo, The Charles Schwab and 4 more

8 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added KLA (KLAC): Not testable, low confidence, AI spend $306M counted as unverified spend.
Added Abbott Laboratories (ABT): Not testable, low confidence, AI spend $588M counted as unverified spend.
Added PepsiCo (PEP): Not testable, low confidence, AI spend $210M counted as unverified spend.
Added The Charles Schwab (SCHW): Not testable, low confidence, AI spend $125M counted as unverified spend.
Added BlackRock (BLK): Not testable, low confidence, AI spend $168M counted as unverified spend.
Added T-Mobile US (TMUS): Not testable, low confidence, AI spend $352M counted as unverified spend.
Added Eaton (ETN): Not testable, low confidence, AI spend $159M counted as unverified spend.
Added Western Digital (WDC): Not testable, low confidence, AI spend $232M counted as unverified spend.
4 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$82.8B	$81.9B	$(892M)
Net return on tested spend	$23.4B	$22.5B	$(892M)
Untested AI spend	$77.2B	$79.3B	+$2.1B

Paying off: 8 of 76 before, 8 of 84 after.

Company	Verified, before	Verified, after	Change	Net AI return
Broadcom AVGO	$8.3B	$8B	$(325M)	$5.5B → $5.2B
Advanced Micro Devices AMD	$2.3B	$2.4B	+$79M	$(94M) → $(15M)
Dell Technologies DELL	$8.6B	$8B	$(602M)	$7.8B → $7.2B
Arista Networks ANET	$1.1B	$1.1B	$(44M)	$820M → $776M

Every company whose verified value moved by more than $1M. All other companies were unchanged.

2 Oct 2026Data

Coverage 67 → 76: Marvell Technology, Amgen, Linde, Qualcomm and 5 more

9 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Marvell Technology (MRVL): Partly visible, medium confidence, net AI return -$395M.
Added Amgen (AMGN): Not testable, low confidence, AI spend $1.5B counted as unverified spend.
Added Linde (LIN): Not testable, low confidence, AI spend $44M counted as unverified spend.
Added Qualcomm (QCOM): Not testable, low confidence, AI spend $2.3B counted as unverified spend.
Added Seagate Technology Holdings (STX): Partly visible, medium confidence, net AI return -$148M.
Added American Express Company (AXP): Not testable, low confidence, AI spend $223M counted as unverified spend.
Added Amphenol (APH): Not testable, low confidence, AI spend $97M counted as unverified spend.
Added Gilead Sciences (GILD): Not testable, low confidence, AI spend $1.2B counted as unverified spend.
Added Analog Devices (ADI): Not testable, low confidence, AI spend $353M counted as unverified spend.
5 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$87.4B	$82.8B	$(4.7B)
Net return on tested spend	$28.9B	$23.4B	$(5.4B)
Untested AI spend	$71.6B	$77.2B	+$5.6B

Paying off: 8 of 67 before, 8 of 76 after.

Company	Verified, before	Verified, after	Change	Net AI return
Microsoft MSFT	$23B	$18.5B	$(4.5B)	$12.3B → $7.8B
Broadcom AVGO	$8.5B	$8.3B	$(188M)	$5.7B → $5.5B
Advanced Micro Devices AMD	$2.2B	$2.3B	+$176M	$(270M) → $(94M)
Dell Technologies DELL	$9B	$8.6B	$(349M)	$8.2B → $7.8B
Arista Networks ANET	$1.2B	$1.1B	$(26M)	$845M → $820M

Every company whose verified value moved by more than $1M. All other companies were unchanged.

2 Oct 2026ModelNAIR v0.5

NAIR v0.5: Jev AI attribution, and no net return for untested companies

Whether a claim credits AI is now decided by the Jev decision model instead of a keyword list, and the 0.40 attribution cut published in v0.4 is now applied. Companies with no testable claim no longer get a net AI return; their AI spend is reported on its own as untested spend.

AI attribution is scored by Jev (model jev-1.13.0, pinned so results are reproducible). For each claim it estimates the probability that AI is the main stated driver of the result, as opposed to one lever among several. It replaces the v0.4 keyword rule. On claims our analysts labelled, Jev separates AI-driven claims from the rest with an AUC of 0.88, against 0.53 for the keyword rule.
The thresholds published in v0.4 are now enforced: below 0.40 a claim is excluded and counts as neither claimed nor verified value; 0.40 to 0.70 still counts and is flagged for analyst review. v0.4 published the exclusion but did not apply it.
Not testable companies (no AI-attributed dollar claim that maps to a reported line) no longer receive a net AI return. In v0.4 their net was minus their full AI spend, which treated “cannot test yet” as “returned nothing”.
The headline now shows net return on tested spend and untested AI spend side by side, with a combined floor (tested net minus all untested spend) beneath for comparison with v0.4.
API and MCP: net_ai_return, ai_roi and margin_impact_bps are null for Not testable companies; new unverified_spend (base, low, high) on companies and in benchmarks; each verification test reports attribution_backend and attribution_review.

Why it moved the numbers. Five claims credit their results mainly to restructuring, cost programs or several levers at once, so they no longer count. IBM, Caterpillar, CVS Health and UnitedHealth Group become Not testable; Accenture loses its restructuring-savings test, so its verified value falls from $550M to $497M. Separately, charging the full AI spend of every untested company against the total mixed missing evidence with negative evidence; that spend is now reported on its own.

Effect on published figures, v0.4 → v0.5

Total	v0.4	v0.5	Change
Claimed value, tested claims	$151B	$143B	$(8.6B)
Verified value, 67 companies	$87.8B	$87.4B	$(364M)
Net return on tested spend (22 → 18 companies)	$23.7B	$28.9B	+$5.1B
Untested AI spend (45 → 49 companies)	$66.1B	$71.6B	+$5.5B
Net AI return, all 67 companies (v0.5: combined floor)	$(42.4B)	$(42.7B)	$(364M)

Paying off: 8 of 67 in v0.4; 8 of 18 tested companies in v0.5.

Flagged for analyst review (attribution 0.40 to 0.70), still counted: JPMorgan Chase, Pfizer, Palo Alto Networks, Arista Networks.

Company	Verified, v0.4	Verified, v0.5	Change	Net AI return
IBM IBM	$312M	$0	$(312M)	$(2.2B) → Not testable
Accenture ACN	$550M	$497M	$(53M)	$(450M) → $(503M)
Caterpillar CAT	$0	$0	+$0	$(477M) → Not testable
CVS Health CVS	$0	$0	+$0	$(1B) → Not testable
UnitedHealth Group UNH	$0	$0	+$0	$(1.5B) → Not testable

Every company with a claim excluded by the attribution cut. Verified value elsewhere is unchanged; the other Not testable companies change only in that their net AI return is no longer shown.

1 Oct 2026Data

Coverage 57 → 67: Applied Materials, Philip Morris International, Sandisk, RTX and 6 more

10 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Applied Materials (AMAT): Not testable, low confidence, net AI return -$892M.
Added Philip Morris International (PM): Not testable, low confidence, net AI return -$151M.
Added Sandisk (SNDK): Not testable, low confidence, net AI return -$266M.
Added RTX (RTX): Not testable, low confidence, net AI return -$561M.
Added Arista Networks (ANET): Paying off, medium confidence, net AI return $845M.
Added CrowdStrike Holdings (CRWD): Not testable, low confidence, net AI return -$346M.
Added GE Vernova (GEV): Not testable, low confidence, net AI return -$239M.
Added Wells Fargo & Company (WFC): Not testable, low confidence, net AI return -$548M.
Added Texas Instruments (TXN): Not testable, low confidence, net AI return -$312M.
Added Thermo Fisher Scientific (TMO): Not testable, low confidence, net AI return -$279M.
4 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$89.3B	$87.8B	$(1.5B)
Net AI return, all companies	$(36.9B)	$(42.4B)	$(5.4B)

Paying off: 7 of 57 before, 8 of 67 after.

Company	Verified, before	Verified, after	Change	Net AI return
Microsoft MSFT	$25.8B	$23B	$(2.8B)	$15.1B → $12.3B
Broadcom AVGO	$8.5B	$8.5B	+$7.4M	$5.7B → $5.7B
Advanced Micro Devices AMD	$2.1B	$2.2B	+$97M	$(367M) → $(270M)
Dell Technologies DELL	$8.9B	$9B	+$14M	$8.2B → $8.2B

Every company whose verified value moved by more than $1M. All other companies were unchanged.

30 Sep 2026Data

Coverage 49 → 57: Visa, Chevron, Costco Wholesale, Lam Research and 4 more

8 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Visa (V): Not testable, low confidence, net AI return -$227M.
Added Chevron (CVX): Not testable, low confidence, net AI return -$85M.
Added Costco Wholesale (COST): Not testable, low confidence, net AI return -$250M.
Added Lam Research (LRCX): Not testable, low confidence, net AI return -$475M.
Added Merck & Co. (MRK): Not testable, low confidence, net AI return -$3.2B.
Added Dell Technologies (DELL): Paying off, low confidence, net AI return $8.2B.
Added Palo Alto Networks (PANW): Partly visible, medium confidence, net AI return -$588M.
Added Netflix (NFLX): Not testable, low confidence, net AI return -$678M.
3 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$88.8B	$89.3B	+$541M
Net AI return, all companies	$(31.2B)	$(36.9B)	$(5.8B)

Paying off: 6 of 49 before, 7 of 57 after.

Company	Verified, before	Verified, after	Change	Net AI return
Microsoft MSFT	$33.6B	$25.8B	$(7.8B)	$22.9B → $15.1B
Broadcom AVGO	$9.3B	$8.5B	$(813M)	$6.5B → $5.7B
Advanced Micro Devices AMD	$1.9B	$2.1B	+$122M	$(489M) → $(367M)

Every company whose verified value moved by more than $1M. All other companies were unchanged.

29 Sep 2026Data

Coverage 40 → 49: Eli Lilly and Company, Advanced Micro Devices, Exxon Mobil, Intel and 5 more

9 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added Eli Lilly and Company (LLY): Not testable, low confidence, net AI return -$2.7B.
Added Advanced Micro Devices (AMD): Partly visible, medium confidence, net AI return -$489M.
Added Exxon Mobil (XOM): Not testable, low confidence, net AI return -$180M.
Added Intel (INTC): Not visible, medium confidence, net AI return -$2.8B.
Added Mastercard (MA): Not testable, low confidence, net AI return -$122M.
Added AbbVie (ABBV): Not testable, low confidence, net AI return -$1.4B.
Added Oracle (ORCL): Not testable, low confidence, net AI return -$3.1B.
Added Palantir Technologies (PLTR): Not testable, low confidence, net AI return -$167M.
Added Cisco Systems (CSCO): Paying off, medium confidence, net AI return $87M.
2 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$81.4B	$88.8B	+$7.4B
Net AI return, all companies	$(23.9B)	$(31.2B)	$(7.3B)

Paying off: 5 of 40 before, 6 of 49 after.

Company	Verified, before	Verified, after	Change	Net AI return
Microsoft MSFT	$30.9B	$33.6B	+$2.7B	$20.2B → $22.9B
Broadcom AVGO	$8.5B	$9.3B	+$813M	$5.7B → $6.5B

Every company whose verified value moved by more than $1M. All other companies were unchanged.

28 Sep 2026Data

Coverage 30 → 40: NVIDIA, Apple, Alphabet, Microsoft and 6 more

10 companies added and 0 updated. Each passed an independent QA review of its sources, test numbers and AI investment before publication.

Added NVIDIA (NVDA): Paying off, medium confidence, net AI return $21.0B.
Added Apple (AAPL): Not testable, low confidence, net AI return -$6.9B.
Added Alphabet (GOOGL): Not testable, low confidence, net AI return -$18.3B.
Added Microsoft (MSFT): Paying off, medium confidence, net AI return $20.2B.
Added Amazon (AMZN): Partly visible, medium confidence, net AI return -$9.4B.
Added Meta Platforms (META): Not testable, low confidence, net AI return -$14.3B.
Added Broadcom (AVGO): Paying off, medium confidence, net AI return $5.7B.
Added Tesla (TSLA): Not testable, low confidence, net AI return -$1.6B.
Added Micron Technology (MU): Not testable, low confidence, net AI return -$1.1B.
Added Berkshire Hathaway (BRK.B): Not testable, low confidence, net AI return -$297M.
1 published companies moved because their peer groups gained members; see the table.

Effect on published figures

Total	Before	After	Change
Verified value, all companies	$2.7B	$81.4B	+$78.7B
Net AI return, all companies	$(18.4B)	$(23.9B)	$(5.5B)

Paying off: 2 of 30 before, 5 of 40 after.

Company	Verified, before	Verified, after	Change	Net AI return
Salesforce CRM	$800M	$400M	$(400M)	$(1.9B) → $(2.3B)

Every company whose verified value moved by more than $1M. All other companies were unchanged.

25 Sep 2026Site

New brand and home page

A new identity for the site, research and email, and a home page that leads with the gap between claimed and verified value.

New mark (Paradox: half water drop, half diamond, after Adam Smith’s paradox of value) with Solow as the alternate mark.
New palette and type across the site, API docs and research.
Home page shows claimed versus verified value for all covered companies.
Sample research library redesigned: tear sheet, Quick Take, State of AI Value summary, sector deck and key-metrics workbook.
Sign-in and confirmation emails now come from The AI Value Report at aivaluereport.com.

25 Sep 2026Platform

Public launch

The leaderboard, API, MCP server and free accounts go live on NAIR v0.4.

Leaderboard of 30 public companies with verdict, confidence and full range; company pages and side-by-side compare.
REST API (v1) and an MCP server with seven tools, one credit pool. Reference at /docs, OpenAPI 3.1 at /openapi.json.
Free accounts: the full evidence behind 10 companies a month, the weekly newsletter and the sample research library.

25 Sep 2026ModelNAIR v0.4

NAIR v0.4: peer baseline, method ensemble and a separate confidence score

A claim now has to beat the company’s peers as well as its own trend. The published verified value is the average of the two, and confidence is scored separately from the verdict.

Added a peer difference-in-differences baseline: the claimed line must improve more than the median of the company’s peer group over the same years. Six peer groups: Financials, Health care, Industrials, Technology & services, Consumer & retail, Telecom.
Verified value is now the mean of the own-trend (τ = 0.5) and peer baselines. The range spans all four baselines (own trend, peer, generous τ = 0, strict τ = 1) combined with the low and high investment estimates.
Added an AI-attribution probability for every claim: 0.70 or above counts in full, 0.40 to 0.70 counts and is flagged for analyst review, below 0.40 is excluded. An attribution-weighted verified value is published alongside the base figure.
Added a productivity signal: revenue per employee against the peer median. Context only, never added to verified value; gaps over 40% are flagged as not comparable.
Confidence is now a 0–5 score shown as High, Medium or Low, separate from the verdict. It replaces the A–C grades used in v0.3.

Why it moved the numbers. Own-trend testing alone credits a company for a sector-wide tailwind. Where the two central baselines disagree the average sits between them: Accenture and IBM roughly halved because their peers improved as much, while Citigroup and Morgan Stanley gained because they beat their peers even though they did not beat their own trend.

Effect on published figures, v0.3 → v0.4

Total	v0.3	v0.4	Change
Verified value, 30 companies	$3.3B	$2.7B	$(643M)
Net AI return, 30 companies	$(17.8B)	$(18.4B)	$(643M)

Paying off: still 2 of 30. Accenture moved to Partly visible; Klarna moved to Paying off.

Company	Verified, v0.3	Verified, v0.4	Change	Net AI return
Accenture ACN	$1.1B	$550M	$(550M)	$99M → $(450M)
IBM IBM	$623M	$312M	$(312M)	$(1.9B) → $(2.2B)
Klarna KLAR	$52M	$61M	+$9.1M	$(3.2M) → $5.9M
Morgan Stanley MS	$0	$14M	+$14M	$(450M) → $(436M)
Citigroup C	$0	$195M	+$195M	$(650M) → $(455M)

Every company whose verified value moved by more than $1M. All other companies were unchanged.

How releases work

Model releases change how figures are computed. Each one says what changed, why, and what it did to the published numbers, including every company whose verified value moved.
Data updates add filings, claims or corrections without changing the method. They are logged here when they move a verdict.
Every API and MCP response carries method_version and as_of, so any figure can be traced to the release that produced it.

On the roadmap

Quarterly lines with Bayesian structural time-series counterfactuals (CausalImpact).
Calibrated 80% and 95% intervals (conformal prediction) in place of the min–max range.

See the method for how each piece works today.