One of the AI boom's most celebrated traders has just been forced to hand over the keys.

Situational Awareness, the hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, built its name on an eye-popping bet: that the pace of AI progress was underappreciated by markets. That thesis powered triple-digit returns in early 2026 and pushed the fund's Assets under Manager (AUM) toward $45 billion at its peak.

That run has now unwound. A sharp July sell-off in AI infrastructure names, including sharp declines in chip suppliers such as Samsung and SK Hynix, combined with a bad short position in software stocks, left the fund overleveraged and short on cash. Prime brokers issued margin calls the fund couldn't meet, and Ken Griffin's Citadel stepped in to absorb its entire public equities book. Assets have reportedly been roughly halved.

Looking ahead, Situational Awareness isn't disappearing. It retains a prized private stake in Anthropic, and will now operate as a leaner, private investment vehicle built around that holding.

The episode lands amid a broader AI-stock retreat, raising questions about how much of the sector's gains were leverage-fueled rather than fundamental. For Citadel, scooping up distressed AI positions at a discount looks like classic opportunistic buying. For Aschenbrenner, the reset tests whether his AI thesis can survive without the leverage that made it spectacular.

- Emmanuel Chukwuani

South Korea released new economic data yesterday and it shows that the country is moving at two very different speeds. Industrial output, spending and investment all rose in June, the first time, together, in three months.

On the surface, it looks like good news, but as we look deeper, we realise it is really just two industries doing all the work. Industrial output rose 6.4%, mostly from manufacturing, while auto production jumped 15.4% on strong hybrid and RV demand and chip output rebounded 4.5% after falling 10% in May. The swing from a sharp drop to a quick bounce is not real stability, but it is one sector dragging the whole number up. Meanwhile, services, where most South Koreans actually work, grew just 0.7%. The money that big exporters earn abroad is not reaching regular local businesses.

This gap is clearly seen in the housing sector as well. Samsung and SK Hynix have paid huge bonuses this year, thanks to the AI chip boom, and that money is flowing straight into Seoul property, pushing prices up. When a country's income rises mainly because export prices go up, not because it is producing more, that extra money tends to stay with a small group. Right now, that group is chip workers and shareholders and they are the ones buying homes.

Regulators are trying to control the situation in two ways. The government changed home loan rules, requiring banks to consider a person's bonus income over three years instead of two. This stops homebuyers from taking out huge loans just because they received a large bonus cheque. Second, the Bank of Korea increased its interest rate a day later, marking its first rate hike since 2023, specifically to cool property prices and reduce personal debt.

In short, the government is not really fixing a housing bubble, but it is cleaning up after a boom that never spread past one industry. Higher rates can slow the loans, but they cannot change who earned the money first. As these rates hit everyone, the service sector, which got none of the boom, still pays for it.

- Armaan Kapadia

Diagnostic backlogs plague healthcare systems worldwide as radiologist shortages collide with rising imaging volumes. Artificial intelligence (AI) has already shown itself to be part of the solution, helping healthcare providers scale clinical throughput while minimising errors. But three steps are needed to unlock the power of AI: administrators need to upgrade IT systems, pilot proven tools, and train their clinical teams. 

Evidence supporting AI efficacy is already emerging from clinical workflows. For instance, researchers at Northwestern Medicine found generative AI helped clinicians complete radiology reports 40% faster without impacting accuracy. Studies like these are increasingly common. AI also reduces error rates by 11% in radiology diagnoses and improves cancer identification by 50% in pathology. As a result, the global AI medical diagnostics market size is expected to grow from $4.03 billion in 2025 to $108.6 billion in 2035. 

Perhaps more importantly, AI is not a substitute for clinical decision-making. AI acts as a force multiplier that automates workflow triage so clinicians can focus on the higher-level work only, they can do. Easing diagnostic bottlenecks with AI is only part of the battle. Clinical and operational leaders must work to convert efficiency gains into long-term value. 

Here are three steps they can take today. Administrators should start by modernising their health system’s IT infrastructure to enable seamless integration with Food and Drug Administration (FDA)-approved clinical decision support tools. From there, leaders can launch pilot projects to study impact in high-volume centres like radiology departments. But before either of these steps can be successful, healthcare leaders must commit to up-skilling clinical teams through comprehensive training initiatives. 

- Shrish Yalamarti

Until next time,

The Long and Short

Markets, Made Sense