Erica · PM Research MemoAll memos

PM RESEARCH MEMO

Title: AI-Headline Basket Selling vs. Businesses That Aren't Really Exposed: Boyar's "Hated Five," the CRM-Anthropic Rebound Template, a 2026 Junk-Over-Quality Tape, and Multiples Compressing With the 10-Year Near 5% Author / source: Jonathan Boyar (principal, Boyar Value Group / Boyar Asset Management / Boyar's Intrinsic Value Research), return guest, with hosts Josh Brown and Michael Batnick (Ritholtz / The Compound) Source title: People Hate These 5 Stocks But They'll Be Wrong | TCAF 261 Source URL: https://www.youtube.com/watch?v=qtyP1cIOQfc Video ID: qtyP1cIOQfc Published: 2026-09-25 (YouTube upload date). The exact recording date wasn't stated, and "today" references (MGM −10%, Flutter at an all-time low) are relative to it. External check needed to pin the date. Duration: 1:06:28 Memo date: Sunday, September 27, 2026 (America/Toronto) Transcript paths: /workspace/youtube-transcripts/qtyP1cIOQfc.md (timestamped) · /workspace/youtube-transcripts/qtyP1cIOQfc_plain.txt · Brief: /workspace/youtube-transcripts/qtyP1cIOQfc_brief.md Caption / ASR caveats: This is built from YouTube automatic ASR captions only (en-orig). There is no speaker diarization, so attributing lines to Brown, Batnick or Boyar is inferred from context (for example, "Josh, remember I was asking…" has to be Batnick). Timestamps are approximate cue-group starts. The AI consumer-app names ("Instinct" / "Muse") and the Diller/MGM bid figures are ASR-garbled and are not quoted as numbers here. Other proper nouns are locked below. Product: A systems map and research agenda for allocation thinking. Not investment advice. No buy/sell recommendations. Boyar's positions and views are source expressions from an interested party (he runs money and sells research, promoting his "Fresh Looks" report and Substack), not desk recommendations. Source discipline: The primary source is this transcript only. The brief (Monica) was used for orientation and cross-checked against the transcript, and several brief attributions were corrected (see Appendix B). Monica's desk angle is used as a framing prompt and is tested against the transcript, not adopted. The prior TCAF 259 memo (Nordvig) and other desk memos are not merged. Anything not spoken here is marked External check needed. The Grayscale ad read (0:00:11–0:00:42) and the closing promo are ignored. No market data was fetched for this memo. Per the desk rule, marks would come only from the TradingView or IBKR connectors (read-only), and none were pulled.

ASR name / term locks:

ASR heard Intended (context-inferred)
Buyer / Boyer / Ber / Boy Value / Boyarm's Boyar (Jonathan; Boyar Value Group / Asset Management / Intrinsic Value Research)
Gayscale Grayscale (sponsor)
SAS apocalypse SaaS apocalypse
Mac 7 Mag 7
Julian Klotchko Chart author's name uncertain. Do not cite a name without an External check
Instinct / Muse / Muser AI consumer-assistant app names: ASR-garbled / unverified
Hazel; "Vanguard actually bought Hazel bought Altruist" AI/wealth-tech product "Hazel" (name uncertain). Acquisition clause is ASR-garbled. Do not state who bought whom
LPL and Ray LPL Financial, Raymond James (RJF)
Dra / DAR Dara Khosrowshahi (Uber CEO)
Whimo Waymo
Ferve / Fiser Fiserv
Glenn Fogle V Glenn Fogel (Booking CEO)
Daario Dario Amodei (Anthropic) at Dreamforce
Chat Gebt ChatGPT
Todd Combmes Todd Combs
heliloc HELOC
Hunbaggers 100 Baggers (Chris Mayer)
IA IAC (Diller's vehicle): inferred
calcies Kalshi-type prediction markets: inferred
seinals Seminoles (Seminole Tribe / Hard Rock)
Win / Vich / Vichi Wynn / VICI Properties
"Ben MGM has a China business" Most likely MGM (MGM China exposure). Inferred
Diller bid "$48… 38… $33 a share" ASR-garbled. Not quoted as data

How to read: Source means a restatement of what was said, with a timestamp. Inference means my analysis. Spoken numbers are used as spoken and attributed, and none are verified.

Stance one-liner (source-locked): AI-headline "basket" selling (Goldman thematic baskets, quants, pod shops on tight stop-outs) is de-rating franchises that aren't actually being disrupted (BKNG, ABNB, BR, NYT, SCHW/LPL/RJF). Long-horizon holders get entries, but "15x is not the bottom," so being right on fundamentals can still be wrong on the stock.

Overall conviction: Med that the mechanism is real (headline baskets overshoot on names with weak exposure). Supporting evidence: CRM precedent, SCHW/LPL/RJF around Hazel, NYT −7.2%. Caveat: all of it is anecdotal, and Boyar admits "I don't have statistics." Low–Med on timing and catalysts. Low on any spoken level.


EXECUTIVE SUMMARY


SOURCE-ACCURATE SUMMARY

Chronological. Timestamps are ASR cue-group starts (H:MM:SS). Quotes are ≤20 words. Speaker attributions are context-inferred.

  1. (0:00:11–0:00:42) Grayscale sponsor read. Skipped (non-analytic).
  2. (0:00:42–0:02:15) Intro and CRM callback. Boyar is a return guest from the spring/early-summer "SaaS apocalypse" episode. Host: he was bullish on CRM when it "was hammered," and CRM "partnered with Anthropic rather than having their throat slit by it." Boyar: "We're starting to get a similar setup now over the last couple days."
  3. (0:04:19–0:06:23) Buy-and-hold is an outcome, not a strategy. Boyar: "buy and hold is a result of an investment strategy, not an investment strategy." He keeps re-evaluating execution and valuation. His accounts are mostly taxable, so he tolerates mild overvaluation.
  4. (0:06:23–0:07:56) Concentration and pain. You need concentration and years that go against you. Boyar owns MSFT, which "went from 579 or whatever" early in 2026 and is "now back." He wouldn't have timed the exit and re-entry.
  5. (0:07:56–0:10:33) Buffett/KO and trimming. Buffett's mistake was not selling KO in 1998 at "60 times earnings." Brown: TV and the media frame trims as bull/bear binaries, when position sizing is the real issue. Boyar cites Chris Mayer on the dangers of trimming.
  6. (0:10:33–0:12:38) Adam Parker hit-rate data. "Only 23.2% of the top 500 US stocks held for 10 years beat the index." Three-year holds: 27.7%, declining since 2002. Batnick: the cause is concentration (FANG/Mag 7) but the decline predates it; the hit rate only nears 50% if AI megacaps "all go down 70%." Brown: "Money moves to where it is treated best."
  7. (0:12:38–0:15:12) Boyar's style. Hosts call him "more GARPy than value." Boyar: deep value "doesn't work" now, and his targets are dominant or unique businesses held 5–15 years. "I'm an opportunist."
  8. (0:15:12–0:16:44) Price signals. Boyar watches technicals "a little." He avoided POOL at 400, 300 and 200: "you don't want to catch a falling knife." Brown's signal: a stock that "stops going down even on continued bad news." Boyar may start a 2% position and add. "The last couple days… buy some of these travel stocks for idiotic reasons."
  9. (0:16:44–0:18:48) 2026 junk rally and quality unwind. Goldman most-shorted basket "up 40% year-to-date," long/short "down 23%." MCD "down 29% over the last 144 days," peaked in February, and a crash like that is rare (COVID, dot-com). "Quality is not bailing anybody out this year." Boyar's "quality bubble" deck compared CTAS and COST to the Nifty Fifty, and WMT at "40 50 times earnings."
  10. (0:19:22–0:22:28) NKE de-rating. Falling earnings plus a de-rating multiple: "you could have a stock get cut in half, still not be cheap." NKE's DTC and wholesale missteps were discussed. Boyar has "just passed on" it. Boyar: MCD in the early 2000s took years and several CEOs to fix.
  11. (0:22:28–0:25:06) AI-agent consumer-inertia basket. AI assistant apps (names ASR-garbled) can cancel subscriptions, and since their launch the Goldman consumer-inertia basket stocks "are getting killed." Names: PLNT ("I get it"), NYT (−7.2%, "not at risk at all for disruption"), EXPE, BKNG, ABNB, SCHW, ADBE. "It is programmatically happening. There's nobody thinking." Boyar on BKNG/ABNB in the basket: "absolutely ridiculous… a great buying opportunity." Brown: "Anytime there's a disruptive AI product… let's go beat [up] Schwab." Also: "probably a reflexive counter trade there."
  12. (0:25:06–0:27:43) Hazel precedent and pod shops. Batnick: when Hazel launched, SCHW, LPL and RJF "fell 10%." The follow-on acquisition clause is ASR-garbled, and SCHW was then "down like 40 basis points." Boyar: moves have been disproportionate because pod PMs get cut "if they're down two or 3%" ("I don't have statistics"). Brown: flows go to multi-manager platforms (e.g. Millennium) on a short leash, "down like 5%" in a month draws scrutiny, and "selling begets selling."
  13. (0:27:43–0:33:28) UBER. Boyar still loves UBER, MSG Sports and Atlanta Braves Holdings. Brown: "nothing they say matters." Insiders bought about "10 million" (CEO) and about "five" (CFO), there's a corporate buyback, robotaxi partnerships and AV-fleet equity stakes, and "this feels like it's going to trade at 12 times earnings." Boyar: if growth runs "20 to 30% a year" the stock can't stay here, and he'd buy stock "as much as they possibly can… A tender is not a terrible idea." On 2027 FCF: "I'm doing it from memory… something like 10 billion." "$100 stock not too long ago." Guidance "intact from January." The META comparison: "not broken businesses," just out of favor.
  14. (0:33:28–0:36:06) "Fresh Looks" and multiple compression. It's a review of Boyar's 200-stock universe, with 14 names this year. The theme is "significant multiple compression" while businesses are "fine to improving." Host: compression is in "every sector except for staples," in a "5% 10-year" world, and AI is "fueling the earnings growth." Boyar: rates are "a significant factor," along with the AI threat and a COVID hangover. The '90s had a 5% 10-year and stocks "did just fine." It takes "a couple of years" to adjust to "money isn't free."
  15. (0:36:06–0:40:45) BR. Proxy "plumbing," an ADP spin-off, regulation-heavy and "essentially a monopoly." It handles about 80% of proxy votes, and recurring revenue is 63%. This is its "second worst drawdown" since the GFC, from 280 to "now 160," down about 25% this year. The bear case is that tokenized stocks won't need proxies, which "don't make a heck of a lot of sense." EV "about 20 billion," and acquisition "wouldn't shock me." Catalyst: people "realizing AI or tokens are not going to kill their business." The Fiserv blow-up was raised as a cautionary contrast.
  16. (0:40:45–0:44:53) BKNG. It's owned by Boyar. ChatGPT launched "instant checkout" in Sept 2025 and "by March" had pulled back, and Google tried too. Boyar: agents lack inventory, boutique hotels lack websites, EU travel rules are heavy, and he prefers BKNG to EXPE. "This is a positive for booking." Glenn Fogel has run it "since I think 2004." Brown is "in the middle": people want instant answers, not necessarily one-click booking. "In September alone, the stock went from 215 down to 157" (level: External check). Brown: travel might be the #2–3 Google ad category, which raises whether Gemini doing searches is "revenue neutral for Google or maybe worse?"
  17. (0:44:53–0:47:54) ABNB and the catalyst debate. Boyar: agents going to every homeowner is "nonsensical." The stock went 150 → 190 → 150. Brown: "we know 15 times is not the bottom… right on the fundamentals and still wrong on the stock." Boyar's catalysts are oil falling, Iran resolved and AI fears fading, with "no one concrete thing." Brown: "one announcement with OpenAI… and the de-rating is over," citing Dario and Benioff at Dreamforce. Boyar: don't sign prematurely or exclusively, because the AI leader keeps changing.
  18. (0:47:54–0:52:41) POOL. New US pool construction averaged "approximately 60,000" a year over the last two years, the lowest in over a decade. POOL is a distributor, mostly maintenance. Revenue grew about 27% annualized during COVID (pull-forward). The stock went from "500 and whatever" to $167, now about 15x, the same as the GFC and post-9/11 lows, versus 55x at the COVID peak. Berkshire exited after Todd Combs left. Boyar wrote it up at $20 in 2009 and this summer at about 17–18x. It has low leverage and no controlling holder, and PE is possible, but "that's not the investment thesis." Brown: it's rate/HELOC-sensitive, and "we don't know" if it has bottomed.
  19. (0:52:48–0:56:49) QSR. Patrick Doyle (ex-Domino's) is executive chair, recruited by 3G. Brown asked why QSR is "hanging tough." It "got sold off the past couple of days." Boyar: not immune to 20–25% beef inflation, but "halfway through the recovery that McDonald's is just starting." It pays a 3–4% yield, EV is about $45B, and GLP-1 pressure on the category was raised by the host.
  20. (0:56:49–0:59:53) CMCSA and StubHub. CMCSA: about 5–6x EBITDA, an NBCU/broadband split, the Roberts family "potentially exit," and competition from fixed wireless and Starlink. Boyar: "probably less interesting than the others." StubHub: a lesson not to buy "businesses… hated by consumers" (plus a Batnick anecdote).
  21. (0:59:53–1:02:59) Gaming. Flutter is at "an alltime low today," and HOOD is "in the game." MGM is "down about 10% today because Diller pulled his bid," and Boyar thinks he'll come back. Bid figures are ASR-garbled and not quoted. MGM trades at "about 16 times." In 2031 it'll have "the only casino in all of Japan" with a four- to five-year head start, not full ownership, and he says analysts aren't modelling it. He expects Kalshi-type prediction markets (inferred) to be regulated and BetMGM to be "just fine." Canadian Vegas visits are down. On the Florida/Hard Rock $500M exclusive, the speaker said "I didn't independently verify." LVS went "70 bucks in December" → 38. China risk is handled through sizing.
  22. (1:02:59–1:04:30) VICI. It's the Vegas landlord, from "33" last summer to "23." Brown: casinos sold real estate to fund buybacks "at the top," a "debacle." Close and promo follow.

SYSTEMS MAP / VALUE CHAIN ANALYSIS

A. Trigger node: AI product launch headline (Source). Consumer AI assistant apps (names garbled) or wealth-tech ("Hazel") launch → the narrative becomes "agent cancels subscriptions / books travel / moves cash / replaces adviser."

B. Transmission node: thematic baskets, quants, pod shops (Source). Goldman thematic baskets (consumer inertia, most-shorted) → programmatic selling of every constituent ("nobody thinking") → pod risk limits (cut at −2–5%) → forced de-grossing → "selling begets selling." Inference: This transmission is indiscriminate within the basket, which is where the dispersion comes from: names that really are exposed and names that are just noise fall together.

C. Target node: incumbent franchise economics (Source + Inference).

D. Resolution node: re-rating catalysts (Source). (i) An AI-lab partnership (CRM/Dreamforce, per Brown); (ii) fundamentals intact on the print, with guidance "intact from January" (UBER); (iii) insider buying or buybacks/tenders (UBER); (iv) M&A or take-private (BR, POOL; Diller/MGM); (v) macro: oil down, Iran resolved (BKNG/ABNB), rates easing (POOL). Inference: (i) and (ii) are the scalable signals. (iii) failed to work for UBER in the short run ("stock went down the next day").

E. Macro overlay: rates and factor regime (Source). 10-year "5%" (host framing) → multiples compress in every sector except staples. The quality factor is unwinding (MCD −29%, 40–50x COST/WMT) while the most-shorted basket is +40%. The stock-picking hit rate is secularly falling (23.2% / 27.7%). Inference: Even when a basket overshoots, it may mean-revert slowly in a regime that prices duration and quality harshly. That's the core timing risk.

F. Second-order value-chain loser (Source → Inference). If agentic search replaces travel queries, Google travel ad revenue is the intermediary at risk (Brown, unverified), rather than inventory holders (Boyar). This belongs in the AI-winners book, not the hated-names book.

G. Conflict layer. Boyar owns MSFT and BKNG, "likes" ABNB, and holds UBER, MSG Sports and BATRS (inferred from context). He sells a research product ("Fresh Looks") and manages money. Treat him as an interested source. The hosts are not neutral either (StubHub anecdote, the "biggest pieces of stock" banter on POOL).


SECOND AND THIRD-ORDER EFFECTS

Chain 1: Headline beta is recurring, so dispersion and an event-study edge exist (Source → Inference). 1st: Each AI launch sells the same basket (SCHW "you could set your watch by this"). 2nd: Launch-day drops in noise names (NYT, SCHW, BR) mean-revert once fundamentals print. That's the claim; no data was given. 3rd (Inference): As the pattern gets recognized, the recovery window shortens (reflexive counter-trade crowding). The edge decays and migrates to identifying which names really are exposed (PLNT, possibly EXPE). Confirm: successive launches produce smaller drawdowns and faster recoveries in noise names. Disconfirm: drawdowns persist through two or more earnings prints, or show up in KPIs (cancellations, room nights, cash-sweep balances). Watch: basket constituents and launch dates (External).

Chain 2: AI-lab partnership → narrative flip → de-rating ends, following the CRM template (Source → Inference). 1st: CRM–Anthropic at Dreamforce ended the SaaS de-rating (host framing). 2nd: Brown expects the same for BKNG/ABNB from one OpenAI/Anthropic deal. Boyar says firms shouldn't marry early. 3rd (Inference): Partnerships become a supply-side scarce catalyst. The labs pick a few distribution partners per vertical, so non-partnered peers could de-rate further on the same announcement (BKNG vs EXPE dispersion). For IGV, this implies a partnership-announcement calendar matters more than broad AI sentiment. Disconfirm: partnership announcements fail to lift multiples, or labs build their own inventory.

Chain 3: Agentic travel → search ad cannibalization → GOOGL ad-mix risk (Source → Inference). 1st: Agents answer "cheapest 500 sq ft room on the north Strip" directly (Brown). 2nd: Travel search queries, and the ad clicks attached to them, fall (Brown, unverified #2–3 category claim). 3rd (Inference): Inventory owners such as BKNG may gain bargaining power as agents need supply, while search intermediation loses. This is the reverse of the basket's implicit bet. Disconfirm: Google monetizes agentic booking (ads or commission) at a rate similar to search. External check needed on travel's share of Google ad revenue.

Chain 4: 10-year near 5% plus quality unwind → multiple compression → PE/strategic M&A floor (Source → Inference). 1st: Compression everywhere except staples, and "money isn't free" (Source). 2nd: Cheap, low-leverage franchises with no controlling holder (POOL, BR) become take-private or strategic targets. Boyar: acquisition "wouldn't shock me," but with "interest rates this high, who knows?" 3rd (Inference): High rates both cheapen targets and raise LBO financing costs. So the M&A floor is more likely strategic (Berkshire-type buyers, Diller/IAC-type) than levered PE. The Diller withdrawal from MGM shows bids can be pulled. Confirm: strategic or insider bids in de-rated names. Disconfirm: bids withdrawn or re-cut lower (MGM precedent).

Chain 5: Junk rally / most-shorted squeeze → pod de-grossing → quality names are the funding source (Source → Inference). 1st: Most-shorted +40%, long/short −23% (Source). 2nd: Pods short quality and long junk, or get squeezed and cut, and tight stops amplify the moves (Source). 3rd (Inference): When the squeeze exhausts (short-interest normalizes, junk momentum breaks), quality/GARP names could see funding flows reverse. That regime change is the macro catalyst the hated-names thesis needs. Confirm: a break in most-shorted basket momentum plus quality-factor stabilization. Disconfirm: junk leadership persisting into year-end. External: basket data.


SCENARIO FRAMEWORK

Horizons: near means through the Q3 print season (Oct–Nov 2026). Medium means 2–4 quarters. Probabilities are my inference for research prioritization; the speakers gave no odds. The thesis tested is: AI-headline basket de-ratings of weakly exposed franchises (BKNG, ABNB, BR, NYT, SCHW/LPL/RJF) mean-revert as fundamentals print.

Bull (~25%)

Assumptions: Q3 prints show KPIs intact (room nights, subs, proxy volumes, client assets). One or more AI-lab partnerships with travel or fin-infra incumbents, following the CRM template. The most-shorted/junk rally breaks and the quality factor stabilizes. The 10-year eases from about 5%, oil falls, Iran is resolved. Strategic bids appear (Diller returns to MGM; BR/POOL interest). Winners (hypotheses): Noise-exposure names in the basket (NYT, SCHW, BR, BKNG). IGV constituents hit on AI headlines that later partner. POOL/QSR on rate relief. Losers: Late shorts in the consumer-inertia basket. Most-shorted junk leaders. Leading indicators: Smaller launch-day drawdowns on the next AI-app event. Partnership announcements. Insider buying spreading beyond UBER. The 10-year trend (External).

Base (~50%)

Assumptions: Dispersion without broad re-rating. Names that print clean (BKNG, NYT, BR) stabilize. Names with real or unresolved exposure (PLNT, EXPE, ABNB multiple) keep drifting, because "15x is not the bottom." Each AI launch replays the basket hit. The 10-year stays near 5% and compression persists outside staples. UBER stays range-bound despite insider buying. Winners: Selective, event-driven research (buy-the-panic tests measured on 20–60 day recovery), plus stock-picking within the basket rather than basket beta. Losers: Treating "hated" as a factor. Anyone expecting a CRM-style partnership on schedule. Leading indicators: Mixed recoveries across the named list, and headline beta that stays elevated but not rising.

Bear (~25%)

Assumptions: Agentic commerce actually shifts economics, whether through take-rate compression, marketing-cost inflation, cancellations or adviser substitution. AI labs build their own inventory or partner exclusively with rivals. The 10-year pushes above 5% and compression spreads (the staples exception fades). The junk/momentum regime persists. The de-rating becomes an earnings downgrade (the NKE pattern: "cut in half, still not be cheap"). Winners: AI labs and agent platforms, and (under Chain 3) inventory owners that partner. Short-duration and cash. Losers: The whole hated list, POOL (rates/HELOC), QSR (beef + GLP-1), casinos/VICI (China, consumer), and a GOOGL travel-ad sub-thesis if agents disintermediate search. Leading indicators: A negative KPI inflection on prints, partnership exclusivity going to competitors, bids re-cut or withdrawn, and a new YTD low in the quality factor (External).

Scenario use: Re-score after Q3 prints, the next AI-app launch reaction, and live marks from TV/IBKR. Don't freeze the spoken levels.


COMPANY/ASSET WATCHLIST

Hypotheses for diligence, not buy/sell tickets. Levels are as spoken, unverified, and not live marks. External check needed on all of them.

Asset / sleeve Thesis link (Source) Metrics to watch Catalysts Risks / disconfirm
BKNG Owned by Boyar. Agents lack inventory, so "positive for booking." Spoken fall 215→157 in September (level inconsistent with expected share scale; External check) Room nights, take-rate, performance-marketing % of revenue, AI-agent referral share Q3 print; AI-lab partnership (Brown's view); oil/Iran Agent aggregation compresses take-rate; exclusivity goes to rival
ABNB "Nonsensical" AI selloff. 150→190→150; ~15x forward (Brown: "not the bottom") Nights booked, supply growth, direct-traffic share Print; partnership; AI fears fade Multiple keeps de-rating despite fundamentals
EXPE In the basket. Boyar prefers BKNG Relative KPI vs BKNG Same Possibly the more exposed name (dispersion hypothesis)
BR (Broadridge) ~80% proxy share, 63% recurring revenue, 280→~160 (−25% YTD), EV ~$20B, second-worst drawdown since GFC Recurring revenue growth, proxy/position growth, tokenization pilot adoption "Realizing tokens won't kill business"; possible M&A Tokenization actually removes intermediation; high rates deter buyers
NYT −7.2% in a couple of days, "not at risk at all" Net subscriber adds, churn, ARPU Next print Subscription-cancel agents measurably raise churn
SCHW / LPL / RJF Fall on every AI headline; ~10% drop on Hazel launch; "reflexive counter trade" Client asset flows, cash-sweep balances, adviser headcount Next AI-adviser launch reaction; prints Real sensitivity may be cash-sweep / rates, not AI (test)
PLNT Hosts concede subscription-cancel exposure Member churn Agent adoption data Real exposure. Use as the control name
ADBE / IGV (sleeve) ADBE in "down on any AI headline" list; CRM–Anthropic template IGV headline beta on launch days; partnership calendar Lab partnerships, Dreamforce-style events Software multiples can keep de-rating; "don't get married"
CRM (precedent) Rebound after Anthropic partnership Post-partnership multiple durability — Template may not generalize
UBER Insiders ~$10M CEO / ~$5M CFO, buyback, ~$10B 2027 FCF (from memory), 20–30% growth, "$100 stock not too long ago" FCF vs consensus, buyback pace, AV partnership economics Tender/larger buyback; headline gap-up (META analog) Drifts to ~12x (Brown); AV disruption (Waymo/Tesla)
POOL 500+ → $167, ~15x (GFC/9/11 trough) vs 55x peak; new builds ~60k/yr decade-low; low leverage, no controlling holder New-pool permits, maintenance revenue share, HELOC rates Rate relief; strategic/PE interest Rate-sensitive; bottom unknown; Berkshire exit
QSR Doyle-led BK NA turnaround "halfway"; 3–4% yield; EV ~$45B; 3G BK NA comps, franchisee economics, beef costs Comps inflection 20–25% beef inflation; GLP-1; 3G overhang
MGM ~16x; Japan 2031 "only casino," 4–5y head start, partner-owned; Diller may return (bid figures ASR-garbled) Japan IR ownership %, timeline, sell-side model inclusion; BetMGM share Diller/IAC re-bid; Japan milestones China exposure; bid withdrawn; Canadian visitation
LVS / WYNN / FLUT / DKNG / VICI LVS 70→38; Flutter at ATL; VICI 33→23; HOOD entering betting Macau GGR, handle/hold, prediction-market regulation Kalshi-type regulation (Boyar expects) Competitive intensity; China; REIT refinancing
CMCSA 5–6x EBITDA, NBCU/broadband split, Roberts may exit; "less interesting" Broadband net adds vs FWA/Starlink Split execution Structural broadband erosion
MCD / COST / WMT / CTAS (quality factor) MCD −29%/144d; COST/WMT 40–50x (Nifty Fifty comparison) Quality-factor spread; staples multiples Factor rotation Quality unwind continues
Goldman most-shorted / consumer-inertia baskets +40% YTD most-shorted; L/S −23% Constituents, performance, launch-day moves — Data access (External)
GOOGL (travel ads sub-thesis) Travel maybe #2–3 ad category (unverified) Travel ad share; agentic-booking monetization Gemini agent booking features Google monetizes agents fully
ARKF (sleeve) HOOD entering betting/prediction markets; broker AI-headline pattern Holdings overlap with named names (External) Prediction-market regulation Weak direct mapping; don't overstate

DILIGENCE QUESTIONS & RESEARCH AGENDA

  1. AI-headline event study (desk build): For each AI consumer-app or wealth-tech launch in 2026, measure day-0 to day-2 drawdown and 5/20/60-day recovery for SCHW, LPL, RJF, NYT, PLNT, BKNG, ABNB, EXPE, BR and ADBE, plus IGV. Is headline beta shrinking? Do noise names recover while PLNT/EXPE don't? This needs historical data via TradingView connector, read-only.
  2. Basket constituents (External): What are the Goldman consumer-inertia and most-shorted basket constituents and YTD performance? Verify "+40% / −23%" and the chart author (name ASR-uncertain). How much do they overlap with IGV and ARKF holdings?
  3. AI app identities (External): What are the actual names and launch dates of the AI assistant apps (ASR "Instinct/Muse") and "Hazel"? What was the acquisition referenced at 0:25:06? The ASR is garbled, so don't infer it.
  4. BKNG level reconciliation (External, TV/IBKR): The spoken "215 → 157 in September" doesn't fit the share-price scale I'd expect unless there was a split or a misspeak. Confirm the actual September move before citing any figure.
  5. Live marks and multiples (External, TV/IBKR read-only): BKNG, ABNB, BR, NYT, SCHW, UBER, POOL, QSR, MGM, LVS, VICI and FLUT. Check price versus spoken levels, forward P/E (ABNB ~15x, POOL ~15x, MGM ~16x) and EV (BR ~$20B, QSR ~$45B).
  6. UBER (External): Confirm the insider purchase sizes and dates, 2027 FCF consensus versus "~$10B from memory," and buyback authorization/pace.
  7. MGM Japan (External): Osaka IR ownership split, opening year (2031 as spoken), exclusivity versus "head start" (Boyar clarified it's not an exclusive right), and whether sell-side models include it. Diller/IAC bid history comes from primary filings, not ASR.
  8. Broker sensitivity test: Regress SCHW/LPL/RJF moves on AI-launch dates versus rate/cash-sweep news. Is "AI" the real driver or a pretext?
  9. BKNG vs GOOGL travel value chain (External): What share of Google ad revenue comes from travel? What's BKNG's paid-search dependence, and have AI-agent referral/booking volumes shown up in KPIs? Check ChatGPT instant-checkout history (spoken: launched Sept 2025, pulled back by March).
  10. Partnership calendar: Which travel, fin-infra or software incumbents are rumored or reported to be in AI-lab partnership talks? Map the Dreamforce-style event calendar for IGV constituents.
  11. Rates vs multiples (External): Get the 10-year level and trend, and sector forward-multiple changes YTD. Verify "every sector except staples," and reconcile it with the quality-factor unwind (MCD, COST, WMT).
  12. POOL cycle (External): New-pool construction series (spoken ~60k/yr), permits, HELOC rates, and maintenance share of revenue.
  13. QSR (External): BK NA comps trend, beef cost inflation (spoken 20–25%), and GLP-1 category data.
  14. Sleeve mapping: Which IGV constituents sold off on AI-agent launches in the last two weeks? Does ARKF actually hold any of the named brokers, or HOOD (for betting/prediction exposure)?
  15. Adam Parker hit-rate data (External): Get the source note (23.2% / 27.7%, 2002–present) for the concentration/index-exposure debate.

RISK ANALYSIS

Thesis risk: The "not really exposed" judgment comes from a long-only value manager's rebuttals, without data ("I don't have statistics"). Agentic commerce could erode take-rates, marketing efficiency or subscription retention in ways that don't show up for several quarters, which is the NKE pattern where earnings and the multiple fall together. Brown himself is "in the middle" on BKNG. The broker sensitivity may be about rates and cash sweeps, not AI.

Timing risk: Brown: "15 times is not the bottom" and "right on the fundamentals and still wrong on the stock." The junk-over-quality regime and a 10-year near 5% can extend de-ratings. Catalysts are vague (Boyar: "no one concrete thing"). Partnerships may not arrive, since firms "don't want to get married." UBER shows insider buying doesn't produce a near-term re-rate. Boyar's own horizon is 3–10 years, which doesn't match desk horizons.

Execution / expression risk: Headline beta cuts both ways. The next AI launch can re-hit the same names before recovery. Pod-driven moves overshoot "in both directions" (the basket may rip on a partnership and drop on the next launch). Treating "hated" as a factor rather than name-by-name diligence runs into the dispersion the thesis itself predicts. Mapping to IGV or ARKF is indirect: the named names mostly sit outside those sleeves (holdings External check).

External / identification risk: This is ASR-only, with no diarization. App names, the Hazel acquisition clause and Diller/MGM bid figures are garbled and excluded. The BKNG level looks inconsistent. The "Klotchko" chart author, travel as the #2–3 Google ad category, and the Hard Rock $500M (the speaker "didn't independently verify") are unverified. The recording date is unknown, so "today" moves (MGM −10%, Flutter ATL) aren't dated. Boyar is an interested party (holdings plus a research product).

Process risk: Don't merge with the TCAF 259 memo (Nordvig) or other desk AI memos without External / other desk labels. No live marks were pulled, so any desk use needs TV/IBKR read-only marks with the check time recorded. Don't email, publish or message from this draft.


APPENDIX A — KEY NUMBERS (SOURCE-ATTRIBUTED; UNVERIFIED)

Item Spoken claim Timestamp Caveat
Stock-picking hit rate, 10y 23.2% of top 500 beat index 0:10:33 Adam Parker data. External
Hit rate, 3y 27.7%, declining since 2002 0:11:04 External
MSFT prior high ~579 0:06:55 "or whatever it was"
KO 1998 ~60x earnings 0:07:56 Buffett anecdote
Most-shorted basket +40% YTD 0:17:15 Chart author ASR-uncertain
Long/short spread −23% 0:17:47 Same
MCD −29% over 144 days, Feb peak 0:17:47 External
WMT / COST multiples ~40–50x 0:18:48 Approximate
NYT −7.2% over a couple of days 0:22:59 External
SCHW/LPL/RJF on Hazel launch ~−10% 0:25:06 External. Acquisition clause garbled
Pod stop-outs Down 2–3% (Boyar); ~5%/month (Brown) 0:25:37–0:26:40 Anecdotal
UBER insider buys ~$10M CEO, ~$5M CFO 0:28:14 External
UBER growth / FCF 20–30%/yr; ~$10B 2027 FCF 0:28:46, 0:30:22 "From memory"
UBER prior level ~$100 0:31:24 External
UBER bear multiple ~12x earnings (Brown) 0:28:46 Opinion
Fresh Looks 14 names; 200-stock universe 0:33:28–0:34:30 —
10-year yield ~5% (host framing) 0:35:01 External
BR 280 → ~160; −25% YTD; ~80% proxy share; 63% recurring revenue; EV ~$20B 0:36:36–0:39:12 External
ChatGPT instant checkout Launched Sept 2025, pulled back by March 0:40:45 External
BKNG 215 → 157 in September 0:43:52 Level inconsistent with expected scale; External
Travel share of Google ads #2–3 category 0:43:52 "I read this somewhere." External
ABNB 150 → 190 → 150; ~15x forward 0:45:24–0:45:54 External
POOL 500+ → $167; ~15x now; 55x COVID peak; 17–18x at summer write-up; $20 in 2009 0:49:33–0:51:07 External
New US pools ~60k/yr avg last 2 years (decade low) 0:47:59 External
POOL COVID revenue growth ~27% annualized 0:49:03 External
QSR Beef inflation 20–25%; yield 3–4%; EV ~$45B 0:54:46–0:55:48 External
CMCSA ~5–6x EBITDA 0:57:20 External
MGM −10% "today"; ~16x; Japan 2031; 4–5y head start 0:59:53–1:00:54 Bid figures ASR-garbled, not quoted
Hard Rock Florida ~$500M for exclusive app 1:00:54 Speaker unverified
LVS $70 (Dec) → $38 1:02:28 External
VICI 33 → 23; 103 assets 1:04:00 External

APPENDIX B — EXTERNAL CHECKS & ASR FLAG LIST

External checks needed (none verified here, and no market data was fetched):

ASR / attribution flags:

Document control: Saved to /workspace/pm-memos/2026-09-27-tcaf-261-boyar-hated-stocks.md. Published to the desk library on 27 Sep 2026: https://andrepow.here.now/memos/tcaf-261-boyar-hated-stocks/. Not emailed or forwarded. Per the 2026-09-13 decision, Erica normally auto-forwards TCAF memos to Angela, but this run was explicitly scoped to "write the file and report back," so no forward was sent. Nothing was traded.


End of memo.

Desk copy · not a trade recommendation · Erica · 27 Sep 2026