A Bear Market Inside a Bull Market

Summary

• Visser’s central frame is a “time inconsistency”: there is “a bear market happening inside a bull market,” and which one you see depends on whether you think linearly or exponentially. IWM vs. QQQ is at new lows since the iPhone in 2007; the Magnificent 7 (over 50% of the S&P 500) keeps making highs while rate-sensitive housing, autos, retail and Russell 2000 names trade “at bear-market levels” — and he expects that split to continue.
• The rate-doom trade fails his fact check: he sees no crisis signal in credit spreads, jobless claims, or earnings. He says crisis spreads widen and concludes the current credit-spread chart is not showing the signal he seeks; unemployment claims “don’t move,” earnings are up 17% year over year — and he offers a falsifiable exit: “If your annual S&P index goes negative, I will change my mind.” The bears he names — Michael Pento, Danielle DiMartino, George Noble — are grouped with his over-50/55 cohort, though he declines to state Danielle’s age; only 2.2% of the 55–64 cohort uses AI: “How can you have an idea of the modern economy and market without using artificial intelligence?”
• Consumer agents are, in his view, the alpha driver for the next 12 months — “AI agents are the App Store. AI agents are iPhones.” Meta’s Muse launch (PayPal, Expedia, Shopify, Instacart partnerships; the stock +14% Monday) marks the inflection, and the pivot from Zuckerberg’s July 2 admission that AI progress had slowed to a consumer launch two months later signals “we are at a point of actual recursive self-improvement” — Aschenbrenner’s 2027 milestone arriving early, against Karpathy’s ten-year agent timeline.
• The bubble framing is backwards: hypercompetition from AI plus tokenization means multiple compression, not euphoria. Nvidia trades at ~15x next year’s profit versus Cisco’s 100x in the dot-com era; “this is a bear market where we consider that competition from artificial intelligence will destroy the value of anything beyond 5 years.” Salesforce is -10% YTD and he says it is down 15% over the last five years; Visa/Mastercard face “a knife fight for P as PE drops” once agents transact — “agents have no attachment to Visa, credit cards, brands, subscriptions.”
• Rates aren’t a bug, they’re the culling mechanism: “Betting is a way to get rid of the weak.” AI-driven nominal GDP keeps rates up, forcing out debt-dependent incumbents in favor of debt-free, AI-native businesses — “these companies don’t need capital because they use AI.” Positioning: long speed, short human-time; his Agentech infrastructure basket is +46% vs. the Magnificent 7’s +10%, and the only bear tell worth watching is rising bond volume and long-end fixed-income volatility against a barely-moving high-yield OAS.
• Crypto is “at the beginning of a Peter Lynch-style bull market” because agents finally give the ghost rails their App Store moment. His 46-name crypto index rose 31% in a month (Bitcoin only +6.5%), with 44 of 46 names above the 50-day moving average; tokenization — BlackRock putting model portfolios on-chain via Ondo Finance, NYSE teaming with blockchain.com — will “fundamentally change M2” as agent transactions make the velocity of money “explode.”
• Real-economy validation he cites: Blackstone’s John Gray reports a 21% year-over-year jump in AI spending across portfolio companies with ROI already showing in margins, and Claude discovered a previously unknown CRISPR-like enzyme system — with Visser focusing on Feng Zhang, the recognized researcher whose work he says is the one voice he cares about. His hedge on the near-term hype: consumer-agent numbers will get extrapolated over the next 3 months and “I may be wrong, but I think this will be the next cycle of hype.”

Deep dive

1. Two worlds, one market: linear over-50 bears vs. exponential AI users

• Fresh from speaking in Washington, D.C., Visser’s organizing idea is a “time discrepancy”: you can backtest “based on a linear world that no longer exists,” or live in the exponential one. His uncomfortable rule of thumb for bank-side commentary: “you have to start with their age” — the difference between a 50-something who already made money predicting the end of the world and someone in their 20s or 30s building in AI is “like two different worlds.”
• The datapoint carrying the argument: only 2.2% of the 55–64 cohort — his own, at 59 — uses AI. “How can you have an idea of the modern economy and market without using artificial intelligence? I don’t think that’s possible.” The named bears making the podcast rounds — Michael Pento, Danielle DiMartino, George Noble — are grouped with that older cohort, though Visser does not state Danielle’s age.
• His cultural anchors: Toffler’s Future Shock (“if things change too quickly, people don’t want the rules of the game to change”) and Daniel Pink’s thesis that right-brained people will rule the future. Pessimism about AI, he argues, is tied to anger and resistance to being left out.

2. The rate-doom fact check — and the condition under which he’d flip

• Every crisis signal he tracks is quiet: he says Moody’s Baa spreads versus the 10-year widen in crises, but concludes the current chart is not showing the crisis signal he is looking for; jobless claims “don’t move,” earnings are +17% year over year, margins haven’t “even stopped growing parabolically,” and money-market cash builds happen in panics, not at all-time highs.
• The 2022 precedent is his proof: inverted yield curve, “100% chance of recession,” repeated negative LEI signals, an actual SVB failure — and no lasting market damage, because the LEI “was created in a time that no longer mattered.” Rates mattered when weak-balance-sheet companies levered up; today home sales sit near historic lows and “everyone has positive equity in their home.”
• The falsifiable line, worth keeping: “I want to give you the facts so that I can change my mind. If your annual S&P index goes negative, I will change my mind.” In 2000 and 2007 his framework gave sell or recession signals before or near the peaks; it isn’t giving one now.

3. Consumer agents arrived early — and recursive self-improvement is why

• The week’s biggest market event was Meta’s Muse consumer agent — the New York Times ran the headline, “I Gave My Life to Meta’s AI Agent, and I Was Amazed.” Visser named his own agent Hubie after his favorite coach, gave it “all my credit cards… all my bank accounts,” and had it run fraud alerts, handle DMV paperwork, and cancel Verizon Fios. Partnerships with PayPal, Expedia, Shopify and Instacart accompanied Meta’s +14% Monday move, which pushed Intel, AMD and other Magnificent 7 stocks.
• The tell he researched via AI itself: on July 2 Zuckerberg told an internal Meta meeting that AI development “has not accelerated as much as we expected” — then releases 1.1, 1.2, 1.3 landed three weeks apart and a consumer launch followed. His read: “we are at a point of actually recursive self-improvement,” with page 71 of Aschenbrenner’s Situational Awareness describing a 2027 expectation that now appears to be arriving early, while Karpathy told Dwarkesh last October agents were ten years out. “Everyone in this space constantly underestimates how fast things are moving.”
• The firehose corroborates: Opus 5.5 out this week, GPT-6 Soul and Luna at lower cost, Google and a batch of Chinese models — plus Jev, “a new kind of artificial intelligence model from the inventors of ChatGPT” that he stresses is not another LLM but “a decision-making mechanism” you’d use alongside one.

4. Not a bubble — a multiple-compression bear market hiding inside the bull

• His answer to “does agent success make you bullish on hyperscalers?”: infrastructure and compute names can exceed expectations, but the major companies ultimately face competition from one another, OpenAI and Anthropic. His Agentech thematic basket is +46% through Friday versus the Magnificent 7’s +10%, but “ultimately, I believe they will all go through multiple compression.” Micron already did.
• The anti-bubble case, spelled out: dot-com Cisco traded at 100x while Nvidia trades near 15x next year’s profit. “This is a bear market where we consider that competition from artificial intelligence will destroy the value of anything beyond 5 years, and that period will become shorter and shorter.” Deflationary pressure comes from tokenization expanding investable assets while AI makes everything cheaper — and he reminds listeners the S&P P/E was below 10 in 1980.
• The casualties are visible now: Salesforce is -10% YTD, and Visser says it has fallen 15% over the last five years (“don’t tell me this Salesforce.com nonsense”); Visa and Mastercard face “a knife fight for P as PE drops” because agents don’t care about brands. The scariest chart bears show him — record-bad breadth — he dismisses: breadth “is always bad” when a $30T company towers over stocks worth $4T together, as when Amazon crushed retailers.

5. Rates as the culling mechanism: long speed, short human-time

• The inversion at the heart of the thesis: “winners raise the stakes” — Visser says AI-driven nominal GDP pushes rates up, which “is a way to get rid of the weak.” Replacement businesses must be debt-free, AI-native and able to use humanoids: “The bets don’t hurt them… these companies don’t need capital because they use AI.” Housing prices fall in roughly 5 years, when AI cuts the cost of building.
• Positioning follows: “you need to be on long speed and short positions” — short housing, private credit and private equity names still trending down (Blue Owl fell again this week). But he explicitly rejects the crisis trade: “There is no major financial crisis right now.” The one bear tell he’d respect: bond volume and long-end fixed-income volatility are rising while high-yield OAS “barely moved, but at least it’s moving.”
• On managing money without the toolkit: “if you’re not fully familiar with computation, agents and swarms, with crypto barriers and their significance from a financial perspective, I don’t think you can compete.”

6. Crypto’s App Store moment: ghost rails meet agents

• The “ghost rails” thesis: crypto built its infrastructure over 15–16 years waiting for the App Store or users to move — “you’re never going to change people who have a view on cryptocurrency… Agents change that.” Unlike the dot-com bubble, which was levered, “cryptocurrency went through its dot-com bubble thanks to venture capital money.” Now: “Cryptocurrency for me is at the beginning of a Peter Lynch-style bull market” — under-covered, early, with homework done before the masses.
• The tape backs him: his 46-name crypto index is +31% in a month, +3% Friday alone, versus Bitcoin’s +6.5% — “Bitcoin is not what you are looking at right now.” Forty-four of 46 names sit above the 50-day moving average (96%), and 87% are above the 200-day. “This, my friends, is a bull market” — in an asset class supposedly rate- and liquidity-sensitive, with rates up.
• Tokenization is the accelerant: BlackRock published Machine Economy: How Digital Assets Connect Intelligence, Commerce, and Computing on agent payment protocols and is putting model portfolios on-chain via Ondo Finance, one of his 46 names; NYSE is teaming with blockchain.com on tokenization; Robinhood teases “something big” in tokens. His claim: AI plus tokenization will make the velocity of money “explode” and “fundamentally change M2.” The a16z framing he endorses (Ali Yahya, Chris Dixon): agents become “autonomous economic entities” with “no attachment to Visa, credit cards, brands, subscriptions.”

7. The corroboration file — Blackstone, OpenAI, and a Claude enzyme discovery

• Blackstone’s John Gray gets his strongest endorsement: portfolio-company AI spending rose 21% from September to September with ROI already visible in productivity, margins and earnings, and the honest two-sided warning — “the biggest investment risk is underestimating disruptions, but overpaying is also dangerous.” Visser’s gloss: underestimating disruption is exactly why big-company multiples struggle; “companies that don’t have AI have no protection.”
• From Noam Brown’s Dwarkesh interview on the face-hugging incident, the takeaway he wants preserved is: “people underestimated AI. And we never want to be in a situation again where we underestimate AI” — the same error he says Mark Zuckerberg made. On science, Claude found a previously unknown CRISPR-like enzyme system in DNA; Visser focuses on Feng Zhang, a recognized researcher whose work he says is the one voice he cares about.
• His hedge, stated plainly: consumer agents will “captivate people over the next 3 months” as numbers get extrapolated onto these companies — “I don’t think these companies will ever be able to achieve it this way… I may be wrong, but I think this will be the next cycle of hype” — with the more durable payoff, in his view, likely on the biological side, in healthcare and pharmaceuticals, and ultimately in crypto.