Joe Weisenthal on Mark Carney's Davos declaration that the old world order is dead and China's export trap

Summary

Mark Carney's declaration at Davos that the old world order is dead stands out as the geopolitical signal of the forum. Carney, the former governor of both the Bank of England and the Bank of Canada, told attendees there is no point in nostalgia for the prior international framework and that sovereignty and power politics have returned as the dominant logic. The weight of the statement derives from its source: Carney is among the most institutionally credentialed internationalists in the world, making his willingness to mark that era as closed a meaningful shift in elite consensus.

The second notable Davos moment came from China's vice premier, who argued China does not want to remain the world's factory and wants to absorb more imports. The claim is not new. Beijing has made similar commitments for years without delivering, and the structural reason is well understood. China's fiscal system is highly decentralized, with provincial leaders racing to maximize export-oriented industrial output to hit targets set by Beijing. The result is a prisoner's dilemma at scale: every province wants its own BYD or battery champion, making a system-wide pivot toward domestic consumption politically difficult to engineer regardless of what central leadership signals publicly.

The question of what China could realistically import at scale has no clean answer. Chips are the obvious candidate, but Beijing's stated goal is import substitution, not permanent dependency. Even in sectors like luxury goods, caviar, and sparkling wine, a December Financial Times report documented China rapidly developing domestic production capacity. The pattern repeats across categories, leaving trading partners with shrinking lists of goods China cannot or will not produce domestically.

Ken Griffin's skepticism about AI-generated analysis, expressed at Davos, reflects a view that current models still produce low-quality substantive output. His example: a commodities trader submitted an AI-generated report that impressed on format but failed on content. The counterpoint is that the pace of model improvement is compressing fast. Andrej Karpathy, who was unimpressed with AI-generated code a year ago, has since said he is struggling to keep up with model capabilities. The more actionable data point is that corporate executives are now citing specific workflow improvements from AI rather than pilot programs, a shift in tenor that appears consistent across financial services. A major bank's CEO, in a conversation recorded for the Odd Lots podcast, offered concrete examples of cost and time reductions that were absent from equivalent conversations twelve months earlier.

A Wall Street Journal report noted a divergence: c-suite executives report significantly greater time savings from AI tools than non-executive workers. The most plausible explanation is that executive workflows, particularly synthesis of competitive intelligence and preparation for meetings, map well onto current model strengths in research aggregation and summarization.

On Japan, the CEO of PIMCO, the world's largest bond fund manager, pushes back on the hyperinflation narrative surrounding surging Japanese government bond yields. His argument is that thirty years of stagnant growth are giving way to a genuine reinvestment cycle, which is generating an inflationary impulse similar to the post-ZIRP rate normalization in the US. The key evidence: the Nikkei is hitting all-time highs in US dollar terms, not just in yen. That distinction matters because the Venezuelan stock market rose in local currency terms while collapsing in dollar terms during its debasement period. Dollar-denominated strength in Japanese equities makes a currency collapse thesis harder to sustain for now.

Joe Weisenthal of Bloomberg's Odd Lots used two weeks of early-morning vibe coding sessions, roughly one to two hours per day at 5 AM, to build havelock.ai, a tool that classifies text sentences as resembling either written or spoken language using lexical markers drawn from linguistics research. He trained a machine learning model without a formal coding background entirely through AI-assisted development. The tool is functional, if imperfect, and Weisenthal tested it on the Carney and Trump Davos speeches, finding measurable differences in orality scores that tracked the perceptible register gap between the two speakers. The project illustrates a broader point: AI tooling is now lowering the cost of building useful niche software below the threshold that previously required commercial justification.