This is the investment framework I use to decide what I would actually own — written down as rules precise enough to be wrong, applied to the 200 largest companies in the S&P 500, and published with the hard cases and boundary calls, because a framework is only tested where it almost breaks.

What this is and isn't. Verdicts express the framework's logic at a single market snapshot (July 2026) and are not real-time trading calls or investment advice. Valuation language is directional — grounded in price-to-sales context read against margin structure, not per-name DCF models — and every price band below must be recomputed when prices or fundamentals move. Conduct items reference public settlements, court rulings, and reported records; sources are listed at the end of the piece.

The decision hierarchy

Every verdict comes from the same ordered sequence, and a name that fails an early stage cannot be rescued by a later one. Stage 0 is a conduct veto. Stage 1 is a pair of hard exclusions: a harm test and a defense threshold. Stage 2 is valuation, the primary screen: a wonderful business at a full price is a HOLD. Stage 3 is resonance. Among fairly priced businesses, preference goes to category-defining products built with trust and craft. Stage 4 is a set of tilts: founder conviction versus organizational capability, and stewardship provenance read against company lifecycle. Tilts break ties. They never rescue a bad price.

Stage 0 — the conduct veto, precisely bounded

One thing outranks price and quality: documented, active obstruction of a safety or investigative process: a company withholding or degrading the evidence a regulator or court needs to adjudicate harm. Tesla's record meets that bar and forces AVOID on its own, independent of valuation: delayed and incomplete crash reporting to NHTSA under the Standing General Order, now the subject of a federal audit query; a documented practice, unique among reporting automakers per a U.S. Senate letter to the regulator, of redacting crash details as confidential business information; and court-documented withholding of Autopilot crash data in a fatal-crash lawsuit Tesla lost at trial in 2025. Sources for each are listed at the end of this piece.

The veto is narrow, on three boundaries. Active, not historical: misconduct that has been surfaced, litigated, penalized, and remediated (Wells Fargo's fake accounts, J&J's talc record, Exxon's climate-science concealment, GM's concealed ignition-switch defect, 3M's decades of internal PFAS knowledge) is flagged prominently in the rationale rather than vetoed, because an investor can price a surfaced harm but cannot price one whose evidence is being hidden. The boundary's hardest test is Cummins, whose emissions defeat devices were engineered specifically to deceive a regulator's test; it escapes the veto only because the conduct was caught, penalized at near-record scale ($1.675B), and ended. A reader who thinks defeat devices should veto retroactively is disagreeing with a stated boundary. It is stated so it can be contested. Obstruction, not litigation: being sued, fined, or investigated does not fire the veto; nearly every mega-cap carries a litigation record, and a veto that fired on settlements would swallow the index and mean nothing. Evidence, not vibes: reputational distaste and polarizing founders do not qualify.

Stage 1 — the harm-exclusion test

A business is excluded on harm grounds only when it meets all three prongs: (a) certain, prevalent harm from intended use: the product seriously harms a large fraction of ordinary users used exactly as intended, at closed-science certainty; (b) involuntary or socialized harm: a material share falls on non-consenting bystanders or the public; and (c) concealment or obstruction conduct: the industry's posture toward the harm is hiding rather than disclosing. The prongs are conjunctive. That is what makes it a rule instead of a mood.

Tobacco meets all three: roughly 480,000 US deaths a year from intended use, 41,000+ non-smoker deaths from secondhand exposure with ~$600B in annual socialized costs, and decades of documented concealment. So Altria and Philip Morris are excluded at any price, smoke-free transition included. A hard exclusion that admits the best-executing tobacco company is not a hard exclusion. And tobacco's famously outsize returns cannot buy an exemption, because the rule exists exactly where the returns are real.

A test must also be shown failing to exclude, or it is a rubber stamp — Robinhood is the demonstration case. The allegation pattern rhymes with tobacco's: an engagement-engineered product (gamified trading) accused of harming its own users. But no prong binds in full. Prong (a) fails: the behavioral-finance evidence on gamified trading is contested and concentrated in susceptible subgroups, not a settled population-wide harm at closed-science certainty; most users trade without serious injury. Prong (b) fails: trading losses fall on the trader who chose to trade; there is no bystander equivalent of secondhand smoke. Prong (c) does not bind: Robinhood's conduct record (a then-record FINRA penalty over supervision failures and misleading communications, plus state actions over gamification) is caught-and-penalized under regulatory supervision, the opposite of decades of coordinated concealment. So: no exclusion, and the name is judged at Stage 2 like everything else, where it holds on valuation, with the flags named in its rationale and standing as sell triggers if the record worsens. A test that can only exclude is a rubber stamp; this one runs in both directions. Sugar fails prong (a): harm is dose-dependent, and there is a health-neutral serving of a Coca-Cola in a way there is no health-neutral cigarette. Fossil fuels flag but don't exclude: prong (b) binds and prong (c) binds historically, but fossil-fuel harm is aggregate and planetary rather than a high-certainty individual kill rate among users. That is a values line, stated so it can be argued with. The exclusion also binds at the level of primary economics: it removes manufacturers, not general retailers who incidentally stock cigarettes.

Stage 1 — the defense threshold

Weapons and defense are excluded where they are the company's primary business: RTX, Lockheed Martin, General Dynamics, Northrop Grumman, and Palantir, whose revenue is majority government work including targeting and battlefield AI; shipping software instead of missiles does not exempt a business whose primary economics are defense. Minority exposure (Boeing, GE Aerospace, Honeywell, Howmet, TransDigm) is flagged, not disqualifying, because at mega-cap scale trace defense exposure is nearly universal.

Stage 2 — valuation, the primary screen

Seven disciplines govern the screen. (1) Multiples are read against margin structure and durability, never as raw ratios. NVIDIA at ~25x sales on extraordinary margins is a coherent (if demanding) price, while pre-veto Tesla at ~16x on automaker margins was a price resting on narrative. (2) Cyclicals are never bought at cycle-high earnings; the current test cases are the AI-adjacent euphorics: Ciena at ~14.5x sales against a 2–3x historical norm, Lumentum at ~25x, Comfort Systems at ~6.4x sales for a mechanical contractor. Real demand, indefensible entries, all AVOIDs. (3) Cheap is not value: a low multiple attached to unresolved existential risk (Intel, Pfizer at the cliff) is a trap until the risk resolves. (4) A verdict is an allocation, not just a grade. Where one thesis has several expressions, the BUY concentrates in the premier expression, and a second earns its own BUY only by independently clearing every stage (Ross beside TJX; Progressive beside Chubb), never by resemblance. (5) Close multiples can carry opposite verdicts, and the pairs are named: Monolithic Power at ~24.7x holds while Lumentum at ~25x is avoided, because the deciding variable is the durability and quality of the earnings behind the multiple. (6) Every BUY carries a price ceiling above which it decays to HOLD, and every HOLD carries a conversion level — or, for event-driven names, a plain statement that no price alone converts it, because a cheaper price on an unproven turnaround is a deeper trap, not a bargain. (7) Selling has its own philosophy: for compounders the sell trigger is thesis-break, not price; for cyclicals, price is the sell; for flagged names, the conduct conditions double as sell triggers.

Stages 3 and 4 — resonance, founders, stewards

Among reasonably priced businesses, preference goes to category-defining products built with trust and craft: things I would use and evangelize without irony. The founder factor splits in two, and bluster earns nothing: conviction authenticity (do I believe the person) and organizational capability (does the machine ship regardless) are scored separately, which lets the framework say the precise thing about Tesla — high capability, zero conviction credit, conduct veto. Stewardship provenance is read against lifecycle: insider-stewards are gifts to compounding companies and liabilities to declining ones; outsider change-agents are the reverse (Niccol at Starbucks, Ortberg at Boeing: the correct provenance for a broken franchise).

I buy trusted, crafted, cash-generating businesses, ideally ones whose products I'd evangelize, only at prices that leave a margin of safety, under exclusions I can state as tests, with every known flag named, and with the conditions for changing my mind written down before anyone asks.

The verdicts

37Buy
146Hold
17Avoid

200 listings, 199 distinct businesses (both Alphabet share classes). First hundred: 27 / 61 / 12 · second hundred: 10 / 85 / 5.

The 37 BUYs

TickerCompanyThesis in one line
GOOGL / GOOGAlphabetThe definitional resonant default at a discount to mega-cap peers; antitrust ruling named, monitored, not vetoing. Verdict re-underwritten July 2026 for the AI-capex transition — see the live thesis note below.
MSFTMicrosoftThree interlocking enterprise franchises at elite margins; the AI hedge and the AI bet at once.
AMZNAmazonA logistics-and-compute utility priced on a retailer's income statement — the accounting shadow is the mispricing.
METAMetaConsumer-internet scale at the group's most modest multiple of cash generation. Disclosure: my employer — listed under the same rules as every other name, without further commentary.
BRK.BBerkshire HathawayThe equity that gets stronger in a crash; culture, not any individual, is the asset.
JPMJPMorgan ChaseScale leadership in everything it does; fortress capital as strategy, not slogan.
VVisaBasis points on global consumption at ~80% margins; the toll-bridge archetype.
MAMastercardThe duopoly's second seat, independently clearing on identical economics.
JNJJohnson & JohnsonAAA-grade diversified healthcare at a modest multiple; talc record named, not elided.
ABBVAbbVieReplaced a Humira-sized cliff in real time — demonstrated execution at an unheroic price.
MRKMerckPaid to underwrite a known cliff: the Keytruda fear is already in the price.
LINLindeThe most reliable compounding machine in industrials; take-or-pay oligopoly economics.
WFCWells FargoPost-turnaround catalyst (asset cap lifted); fake-accounts record named in full.
AXPAmerican ExpressA payments company that is also an aspirational brand; closed-loop economics.
BLKBlackRockToll collector on indexing itself; flows compound faster than fees compress.
TJXTJX CompaniesThe buying organization is the moat; counter-cyclical treasure-hunt economics.
TMOThermo FisherPicks-and-shovels of medicine at a cycle-trough-flavored price.
ISRGIntuitive SurgicalThe switching cost is surgeons' hands; the one name where quality earns a premium entry.
GLWCorningMaterials-science craft with an AI optical inflection the market long ignored.
ABTAbbottDiversified medtech tortoise with a genuine consumer product (Libre) as the engine.
UBERUberThe network-effects survivor of mobility, finally cash-generative — the most on-thesis name in public markets.
CBChubbUnderwriting craft institutionalized; the premier P&C expression.
SPGIS&P GlobalFour toll bridges in one company.
BKNGBooking HoldingsThe demand aggregator of global travel; the bear case has lost for fifteen years.
SYKStrykerMedtech's most consistent compounder; Mako builds an installed-base annuity.
PGRProgressiveA data company selling insurance; earns the second P&C BUY on its own edge.
DHRDanaherThe operating-system acquirer of life sciences at a post-hangover fair price.
INTUIntuitCategory-defining defaults priced for AI disruption; FTC flag named.
CMECME GroupA monopoly on derivatives liquidity that converts volatility into revenue.
ADBEAdobeThe clearest quality-at-a-discount case in software; the disruption thesis is the risk being paid for.
ICEIntercontinental ExchangeExchange, data, and mortgage rails — half the revenue recurs regardless of volumes.
MRSHMarsh & McLennanThe toll-bridge principle applied to risk itself.
ABNBAirbnbFounder-led, category-defining, and financially mature — clears every stage at once.
MCOMoody'sThe ratings duopoly's second seat with a growing analytics annuity.
CRHCRHWhat cities are made of, at a reasonable price — the built-environment core holding.
ROSTRoss StoresThe off-price thesis, independently cleared: same moat, distinct customer.

The 17 AVOIDs, by ground

TickerCompanyGround
TSLATeslaConduct veto — delayed and redacted crash reporting to NHTSA (under federal audit query) and court-documented data withholding in a fatal-crash suit. No price converts this.
MO / PMAltria · Philip MorrisHarm-exclusion test, all three prongs.
RTX / LMT / GD / NOCRTX · Lockheed · Gen. Dynamics · NorthropDefense primes — categorical exclusion.
PLTRPalantirDefense-primary economics, and the most extreme multiple in the two hundred (~62x sales) — two independent grounds.
INTCIntelValue trap: cheapness reflects an unproven foundry bet. Re-entry on customers committing, not on price.
BABoeingCash-burning turnaround already priced for recovery; re-entry on sustained free cash flow.
SNDK / STX / WDCSandisk · Seagate · Western DigitalStorage cyclicals capitalized at cycle-peak earnings.
APPAppLovinExtreme multiple atop named earnings-quality questions.
CIEN / LITE / FIXCiena · Lumentum · Comfort SystemsAI-adjacent cyclical euphoria — multiples several times historical norms.

The 146 HOLDs, including beloved franchises like Apple, Costco, Netflix, Nike, and Coca-Cola, are wonderful-or-fair businesses whose prices offer no margin of safety today. Each carries a written conversion level or event trigger in the full document.

What would change these verdicts

A framework that cannot say what would change its mind is a portfolio of opinions. Every flagged name's conduct conditions double as its sell triggers. The Tesla veto lifts only on documented restoration of regulator data access, and the valuation stage still applies after. Boeing revisits on sustained free cash flow; J&J flips on any evidence of current concealment; Intel, Pfizer, and the managed-care complex convert on events, not prices; and every premium-priced HOLD becomes a BUY at a price restoring margin of safety. The watchlist is the HOLD list.

Live thesis note — Alphabet and the capex question (July 2026)

The largest fact to change since these verdicts were written sits under the list's first name. In Q2 2026 Alphabet reported its first negative free-cash-flow quarter since listing in 2004, roughly −$5.9B, not from operational weakness (revenue grew 24%) but from $44.9B of quarterly capital expenditure, with full-year guidance raised to $195–205B. The company the BUY was originally written on, a software business with extraordinary cash conversion, is becoming an infrastructure owner at the margin: data centers, TPUs, power. That is a different kind of business, and it needs its own note rather than a silent carry-forward.

The verdict holds, with its terms restated. The question is no longer only "is Alphabet cheap relative to mega-cap peers" but "will roughly $200B a year of AI capital earn returns above its cost." My working split: ~60% the buildout earns attractive returns, ~25% it proves a moderate overbuild, ~15% it meaningfully destroys value. These are judgments, not statistics. The demand evidence currently favors the first branch: Cloud revenue up 82% with a $514B backlog, and management stating demand still outpaces capacity, which is what offensive spending looks like. But backlog is a claim on future revenue, not proof of future returns on capital.

The deciding variable is the auction, not model quality. Alphabet's moat was never Search itself. It was the efficiency of matching commercial intent to advertisers. If AI preserves or strengthens that matching, today's capex compounds on top of distribution no rival matches. If AI answers replace monetizable queries while monetizing worse, world-class models will not save the returns. Secondary conditions, each necessary: utilization staying high (idle accelerators earn nothing), AI pricing not collapsing under many-way competition, hardware not depreciating faster than it earns, and Cloud maturing into a true second profit engine, where the Q2 margin expansion is real evidence. The market sometimes prices management's flexibility at zero, but much of the announced spend is not yet irreversible. Future-year plans are largely deferrable, purchase commitments partially so, poured concrete not at all. Management keeps real option value to throttle if returns disappoint.

The decay triggers: evidence the spend has turned defensive (backlog stalling, pricing power fading, capacity ahead of demand); search monetization deteriorating without offsetting AI revenue; or utilization and pricing disclosures implying returns below the cost of capital. The monitoring list is short and specific: free cash flow and its trajectory against guidance, purchase commitments and construction-in-progress in the filings, Cloud growth and operating margin, search monetization per query, and any disclosed utilization or AI pricing data. If those turn, the BUY decays on its own stated terms.

Reminder: this is a framework document, not investment advice. Positions, prices, and multiples reference a July 2026 snapshot and change daily; conduct items summarize public records and reported events. Do your own work — that's rather the point of the piece.

Sources and records

The conduct claims above rest on public records. The load-bearing Tesla items are sourced first; the remainder are settled enforcement matters of public record, cited precisely enough to locate in seconds. Allegations still in litigation are identified as allegations.

Tesla / NHTSA

Alphabet capex note

Settled enforcement and litigation records (public)