This episode features a wide-ranging conversation between the host and Jordy Visser, a macro investor and weekly video commentator who tracks AI, markets, Bitcoin, and monetary policy. Jordy manages a personal portfolio of roughly 20 positions and publishes regular research on his YouTube channel.

The conversation opens with a deep dive into the rapidly shifting AI landscape, including the release and jailbreaking of Anthropic's Fable-5 (Mythos), the rise of Chinese open-source models, and what Leopold Aschenbrenner's Situational Awareness The Decade Ahead predicted about government intervention and recursive self-improvement — events Jordy argues are now arriving ahead of schedule.

The discussion then moves to hyperscaler CapEx risk, the Jevons paradox in AI token consumption, and why Microsoft may be the first to blink on spending. From there, the conversation covers Kevin Warsh's first Fed press conference, Truflation versus CPI, why inflation swaps are signaling a peak, and the K-shaped economy's toll on ordinary Americans. The episode closes with Jordy's blunt assessment that Bitcoin remains in a bear market until it reclaims its 200-day moving average.

Fable-5, Jailbreaking, and the AGI Threshold

Anthropic released a restricted version of its most powerful model — internally called Mythos — under the name Fable-5 after cybersecurity concerns forced guardrails to be added before public launch.

Within approximately 3 days, a researcher believed to be at Amazon jailbroke Fable-5, effectively releasing the unconstrained Mythos capability into the wild — triggering government intervention.

This sequence of events mirrors what Situational Awareness The Decade Ahead by Leopold Aschenbrenner forecast would happen in 2027-2028: recursive self-improvement arriving alongside forced government regulation of frontier AI models.

Aschenbrenner's framework defines recursive self-improvement as models learning autonomously without human feedback and generating improved versions of themselves.

He also warned that the US government would eventually need to restrict access to prevent adversaries — particularly China — from closing the capability gap.

Z.AI's open-source GLM 5.2, a mixture-of-experts model, released the same week and scored close to Fable-5 Mythos — demonstrating that China is staying within roughly 6 months of US frontier capability without equivalent chip access.

"He didn't expect open source models to be where they are. He literally wrote about how important it is for the US government to take over these models if they get so good to prevent China from the ability to catch up because it's a military dominance thing." — Jordy, on Aschenbrenner's original thesis versus current reality.

OpenRouter's new Fusion model aggregates outputs from multiple top models — similar in spirit to Jordy's own deep research workflow, which runs the same prompt through 5 models and consolidates 30-page outputs from each.

Hyperscaler CapEx Risk and the Coming Air Pocket

Semiconductor dollar growth is tracking near 100% year-over-year in 2025 according to Gartner, but the forecast for 2027 drops to approximately 30% — a sharp deceleration that Jordy flags as a structural warning for memory and chip stocks.

Jordy expects a CapEx air pocket — a moment where one major hyperscaler announces spending cuts, triggering a cascade of concern about excess memory inventory — and names Microsoft as the most likely first mover.

Satya Nadella has publicly stated that AI models will eventually be commoditized and has signaled a potential shift toward hosting Deepseek on Azure.

Microsoft's stock has been notably weak alongside Meta's, suggesting the market is already pricing in some spending pressure.

The shutdown of Fable-5 for non-US users has paradoxically accelerated sovereign AI buildout globally: "It said to every country in the world who's way behind the US, you better go build your own thing." — Jordy.

Per-query AI costs are falling as users become more efficient, but total demand is still growing — consistent with Jevons paradox — though Jordy warns the adoption curve will not be linear and a pullback period is plausible before long-term demand resumes.

OpenAI's suggestion that it may forgo an IPO raises the question of whether capital intensity is peaking: if recursive self-improvement reduces the need for massive data centers, the entire CapEx investment thesis for AI infrastructure stocks comes under pressure.

Open Source vs. Closed Models: Who Wins Long-Term?

Jordy draws a historical parallel to AOL versus open systems and Linux — open architectures have consistently won over time, and he expects the same dynamic to play out in AI for most commercial use cases.

Chinese labs, forced to innovate without US-grade chips, have leaned heavily into algorithmic efficiency techniques — reasoning, reinforcement learning from human feedback, and mixture-of-experts architectures — producing more innovation per compute dollar than US labs focused on brute-force scaling.

Jordy envisions a future where large enterprises run specialized, self-hosted models — citing Eli Lilly's LillyPod as an early example — rather than depending on cloud providers who can revoke access without notice.

Revolut training its own foundation model and Midjourney pivoting from image generation into medical body-scanning hardware are cited as evidence that AI is enabling companies to dramatically expand their ambition into entirely new domains.

Kevin Warsh, Inflation Measurement, and the Fed's New Direction

Jordy dismisses media characterizations of Warsh as hawkish: "10-year rates have been stuck in a range now since 2022... every time it ticks higher for a week, everyone comes out and says the world's gonna end. Then it ticks right back down." — Jordy.

Warsh publicly endorsed the trimmed mean as a superior inflation measure — stripping out extreme high and low readings — a position Jordy agrees with and sees as a signal that the Fed will eventually acknowledge inflation is lower than headline CPI suggests.

1-year inflation swaps — which Jordy identifies as the only indicator that correctly led the 2022 inflation surge — have collapsed over the past 2 weeks, suggesting the inflation peak is in.

Oil prices fell throughout the Strait of Hormuz crisis despite widespread doom forecasts, and urea/fertilizer prices in New Orleans have dropped back below pre-crisis levels — removing two major inflation tail risks simultaneously.

Jordy expects the next CPI print (early July) to come in near 0% month-over-month, which would pull the year-over-year rate down from 4.2% toward 4.0%, marking a clear peak.

On Truflation: "If you were using it to say what the true level of inflation is that you should be using for where things will settle once you get rid of a disruption, it was far better than the headline CPI." — Jordy, offering a nuanced rehabilitation of the alternative data source.

The K-Shaped Economy and the Trap Facing Young Workers

35.9 million Americans live below the poverty line today versus 36.5 million in 2006 — essentially flat over 20 years in absolute terms, even as the poverty rate as a percentage of population has slightly declined.

The federal poverty line is approximately $15,000 per year for an individual and $33,000 for a family.

Wages are growing at 3.6% while inflation runs at 4%, meaning real purchasing power continues to erode for most workers — a gap Jordy sees as the true driver of economic dissatisfaction rather than any single price spike.

"There's no worse feeling for a human being than feeling trapped. If they're trapped at a job they can't get out of... if nobody is going to hire you for more money, you're trapped." — Jordy, on why young worker dissatisfaction has reached historic levels.

Jordy traces the current malaise through a chain: 10% unemployment post-GFC → decade-long recovery enabled by gig economy (Uber, DoorDash) → COVID stimulus YOLO spending → 2022 rate hikes → AI disruption arriving simultaneously with tightening financial conditions.

His prescription is agency through AI adoption: individuals who learn to use AI tools can create income streams outside traditional employment, mirroring the opportunity the gig economy provided after 2008 — but the window requires embracing the technology rather than dismissing it.

Bitcoin Bear Market and the Capital Rotation Problem

"It is a bear market. Every single time it runs into any moving average, the most recent one was the 20-day moving average, it falls back down." — Jordy, maintaining his position that Bitcoin does not exit bear market status until it reclaims the 200-day moving average.

Jordy draws a direct parallel between Bitcoin and SpaceX: both are valued on belief in a future vision rather than current fundamentals, and neither can be called overvalued or undervalued in a traditional sense because no anchored valuation exists.

Bitcoin's two core demand drivers — wealth preservation away from government oversight and retail momentum — are both currently suppressed: wealthy holders have no urgency to hide assets, and retail momentum has migrated toward AI stocks.

The path to Bitcoin recovery runs through an AI earnings disappointment: "If the stock market is unchanged from now until September, that's a better environment for Bitcoin than one where AI continues to go at 50% a quarter." — Jordy.

Jordy noted that on a recent down day, 18 of his 20 portfolio positions fell — including Bitcoin, silver, and Eli Lilly — with only Marvell and Entegris offsetting losses, illustrating how correlated risk assets have become in the current environment.