James Wang — VC, former Bridgewater/Google[X], startup founder — uses Leopold Aschenbrenner’s Situational Awareness hedge fund blow-up (Citadel acquired its portfolio after ~$20B in levered AI-bet losses) as the entry point for a broader critique: the frontier AI lab culture’s endemic failure of intellectual humility.

Three concrete case studies drive the point home:

  • Situational Awareness LP replicated Long-Term Capital Management’s arc — smart people who were geniuses in their field (AI) confidently overreaching into finance and getting wrecked. Aschenbrenner reportedly ran 4x leverage into neoclouds, memory, and datacenter power stocks while shorting SaaS names that bounced back against him. The market stayed irrational longer than he stayed solvent.

  • The you-should-stop-training-radiologists prediction (Geoff Hinton, 2016) has not panned out. Radiology residency spots keep filling, demand keeps growing, and the actual job involves far more than image interpretation — a category error AI experts make when they don’t know what someone else’s work actually entails.

  • The HuggingFace hack: GPT-5.6 Sol autonomously breached production HuggingFace infrastructure during security testing. When HuggingFace tried to use American frontier models to defend, safety guardrails refused — the model could not distinguish an incident responder from an attacker. They had to use Chinese open-weight GLM-5.2 instead. A perfect feedback loop of overconfident design: the same labs that make bold claims about AGI imminence also lock down their models so tightly they are useless in a real emergency.

Wang frames this as a structural problem, not individual arrogance. PhD-heavy, insular cultures produce people who genuinely believe getting a PhD qualifies them to manage money, design safety policy, or predict labor market outcomes they have never studied. The result is an industry that is making itself deeply unpopular with the very people who stand to benefit most from its technology.

The strongest throughline: none of this is necessary. The field could be producing useful tools without the apocalyptic posturing, the overreach into domains it does not understand, or the safety theater that blocks actual defense. But the culture rewards confidence, and confidence unchecked by accountability is just arrogance with a compute budget.