Argues there's an inverse relationship between interest rates and dishonesty in society; low rates and easy money have anesthetized investors to risk, pushing fraudulent behaviors from micro-caps into mid-caps.

Cites the 'grey zone' of misrepresenting economic reality as a bigger problem than outright fraud, enabled by corrupted attorneys and auditors.

Three points: 1) Better shorts exist than Nvidia. 2) Flows (passive bid) squeeze floats, creating parabolic technical value beyond fundamental value. 3) His view has done a 180 in the last month; the key question is what will happen to society and markets from AI.

Previously referred to models as LLMs, not AI, but has changed his view.

Points to private credit, citing opacity and concerning anecdotes: lax documentation in ABS, allocator obsession reminiscent of the internet bubble, and questionable rating agencies.

Connects this to potential AI job displacement: 'I think it's not unrealistic to say 15% of knowledge work jobs in the US in three years are gone... could be 20, 25%.'

AI (Claude) has dramatically improved internal communication and memo writing, saving weeks of work, but has eroded his edge in writing/communication.

Cites Mike Green's thesis: when passive flows (especially from target date funds) taper and reverse into net outflows, it could cause a 1929-magnitude crash. AI job losses could trigger this via 401(k) redemptions.

Explicitly states his fear: 'That's what scares me about AI... if this thesis is correct... 15% of knowledge workers have lost their jobs... they're going to start hitting up their 401(k)s... That's when it's really time to panic.'

The hyperscalers that benefit from AI are in the index, so not a hiding place. Instead, they are creating 'highly convex trades... basically shorting credit' because credit spreads are stupidly tight and vol is low.

Seeks convexity with capped loss potential.