The Centaur Weekly | AI acceleration, geopolitical risks, and economic disruptions are converging at the same time. This newsletter, curated by Cenk Sidar, breaks down the major news, analyzes why it actually matters, and highlights the risks and opportunities shaping power, markets, and technology.Commodities Surge as Dollar Confidence Frays
Commodities opened the week with a sharp bid. Silver jumped more than 13% in a single session to a record above $117 an ounce, its largest intraday move since the financial crisis. Gold followed, climbing as much as 2.5% to $5,111 an ounce and extending its year-to-date gain to roughly 18%. The catalyst was a rapid slide in the dollar, down about 2% in six sessions, amid speculation that Washington may tolerate yen strength and growing unease over Federal Reserve independence under erratic policy signaling. In dollar terms, gold has doubled in under 18 months and silver in less than four.
Commodities are rallying for many reasons, but one dominates: the debasement trade. This is not about short-term inflation hedging or cyclical demand. It is about confidence specifically, the erosion of confidence in U.S. assets as the neutral core of the global system.The dollar is no longer treated as politically insulated. Fiscal indiscipline, erratic policy signaling, and growing doubts about Federal Reserve independence have reframed the U.S. from anchor to variable.
For global institutions and sovereign actors, that changes the calculus. When the reserve currency itself becomes a source of volatility, diversification stops being optional.The response is predictable. Capital retreats toward the last asset that once underwrote monetary trust: gold. As gold reasserts itself as a monetary asset rather than a commodity, the effect bleeds into adjacent metals. Silver benefits not because of industrial demand alone, but because it retains residual monetary credibility. This is less a boom than a quiet vote of no confidence not against markets, but against the political foundations of the dollar system itself.
Trump Signals Tactical Shift on Minnesota Deportations After Fatal Raids
President Trump indicated a recalibration of his administration’s deportation crackdown in Minnesota following nationwide backlash over the killing of two U.S. citizens during federal immigration raids. He said border czar Tom Homan would be sent to Minneapolis to ease tensions, a move seen as a moderating counterweight to hardline officials such as Homeland Security Secretary Noem. Trump also spoke with Minnesota Governor Walz, signaling openness to independent investigations into the shootings and a reduced federal agent presence. The White House later suggested it could withdraw Customs and Border Protection personnel if state and local authorities adopted greater cooperation with federal enforcement, marking a notable shift from an earlier confrontational posture.
This is damage control, not leadership. The shift comes after the harm is done, not before. Federal agents were given broad authority and political cover, and once that signal was sent, escalation was inevitable.After the first killing, the warning signs were unmistakable. Instead of tightening oversight, the administration doubled down on enforcement rhetoric and protection for agents in the field. That environment produced the second death. Responsibility doesn’t rest with individual officers alone; it sits squarely with the White House that empowered them.
Pulling back now is still the right political move. Continuing to press this issue would only harden opposition and further polarize the country at a moment of visible public anger. But the retreat doesn’t rewrite what happened. It reflects a president responding to backlash, not one setting clear boundaries in advance. The damage was foreseeable and it was avoidable.
Fei-Fei Li’s World Labs Seeks New Funding at $5 Billion Valuation
World Labs, the AI startup founded by Fei-Fei Li, is in talks to raise several hundred million dollars at a valuation of roughly $5 billion, according to people familiar with the discussions. The round would mark a sharp step up from the company’s 2024 debut out of stealth, when it raised $230 million at a $1 billion valuation. Backers include Andreessen Horowitz, NEA, Radical Ventures and Nvidia’s venture arm. World Labs is developing what it calls “large world models” — AI systems designed to understand, navigate and reason about three-dimensional environments.
Investors are already gaming the next turn of the AI cycle. Large language models delivered the obvious wins in productivity, interfaces, and distribution, and now the question is what compounds from here. The growing view in Silicon Valley is that LLMs are approaching diminishing returns. Bigger models, more parameters, and higher training costs are all incremental. If you believe AGI is still on the table, you need a different axis of progress.
That’s where world models enter the picture. Fei-Fei Li and Yann LeCun are articulating what many technologists are privately coming to accept: intelligence that can’t understand the physical world is structurally limited. Text prediction alone doesn’t get you reasoning, agency, or real autonomy. World models aim to give AI a persistent, spatial understanding of reality, how objects interact, how environments change, and how actions create consequences. That’s the missing substrate if you want machines that can plan, adapt, and eventually act without constant human scaffolding.
The trade-off is time and capital. World models are orders of magnitude harder than LLMs. They require massive, multimodal datasets, real-world feedback loops, and tight integration with robotics and simulation. This isn’t a two-year hype cycle; it’s a decade-long build. Serious milestones are likely to arrive only in the 2030s. But that’s exactly why the smartest capital is leaning in early. If LLMs were the SaaS moment of AI, world models are the infrastructure bet fewer players, higher burn, and potentially asymmetric outcomes for those who get it right.
The UK is shedding more jobs than it is creating as a result of artificial intelligence adoption, and at a faster pace than its global peers, according to new research from Morgan Stanley. British firms reported an average net job loss of 8% over the past year due to AI — the worst outcome among a group that included companies in the U.S., Germany, Japan, and Australia. While UK businesses recorded an average productivity boost of 11.5% from AI, nearly identical to gains reported by U.S. firms, the employment outcomes diverged sharply.
🎥 Merryn Talks Money — Pippa Malmgren from Greenland
A sharp, on-the-ground conversation about what’s really at stake in the Arctic. This isn’t just about territory or rare earths, but space–earth data links, energy security, Arctic defense routes, and how Greenland sits at the center of a rapidly shifting geopolitical order. Watch or listen here: youtube.com/watch?v=EJwi_SJzHBo&list=PLe4PRejZgr0OV82z8tlMKm3p4GGelkTXr&index=2
🤖 Why Robotics Will Work
A clear-eyed take on the future of robotics: https://finaloffshoring.com/?utm_source=substack&utm_medium=email#why-robotics-will-work
✍️ Text Is King
A thoughtful essay on why text remains the dominant interface for thinking, reasoning, and coordination even in a world obsessed with video and multimodal AI. Read here: https://www.experimental-history.com/p/text-is-king?utm_source=substack&utm_medium=email
🎙️ 4x4
The latest episode of 4x4 takes a global view of power, technology, and geopolitics through a tightly curated discussion format. Sharp, unsentimental, and well-paced. English subtitles are available.
Watch here: https://www.youtube.com/watch?v=yNE2sjrt328