The AI Doom Loop | We Break Down Citrini’s Viral AI Crash Story
TL;DR
The hosts analyze Citrini's viral thought experiment forecasting a 2028 AI-driven economic crash, evaluating whether AI acts as a labor substitute or productivity multiplier while dissecting which software business models face existential risk versus benefit from cost efficiency.
⚖️ AI's Economic Paradox 3 insights
Growth formula under threat
Economic growth traditionally equals people times productivity, but AI threatens employment levels while promising massive productivity gains, creating net impact uncertainty.
Substitution versus complement debate
The central question is whether AI replaces workers entirely or amplifies existing talent like software engineers tenfold through cloud coding tools.
Purchasing power versus transition speed
If AI cuts goods costs by 90% while wages remain static, consumers benefit enormously, but social stability depends on whether adoption happens over years or decades.
💻 Software Sector Disruption 3 insights
Collapse of code-only moats
Companies relying solely on software quality face existential risk as AI enables competitors or customers to replicate functionality at near-zero cost.
Enterprise incumbents advantaged
Firms like Salesforce retain power through brand equity, switching costs, and liability protection rather than code quality, allowing them to capture value from lower engineering costs.
Liability as a barrier
Enterprise customers require accountable vendors to sue when systems fail, ensuring human-backed software providers remain necessary for critical infrastructure.
⏰ The 2028 Doom Loop Scenario 3 insights
Fictional retrospective framing
Citrini frames his analysis as a June 2028 look-back, emphasizing that path dependency and disruption speed matter more than end-state equilibrium.
The euphoria feedback loop
AI productivity gains initially boost margins, prompting reinvestment in more AI that accelerates labor displacement faster than the economy can equilibrate.
Data lag risks
Current economic indicators show minimal AI impact in labor and productivity data, meaning by the time disruption appears, market adjustments may already be underway.
Bottom Line
Favor enterprise software with intangible moats like switching costs and brand equity over pure code-based SaaS, as AI differentially compresses margins for the former while reducing costs for the latter, making the 'doom loop' a tail risk worth monitoring but not a base case.
More from Excess Returns
View all
We Asked an Options Expert Why War and Oil Haven’t Broken This Market — and Which Signal Could
Options expert Brent explains why escalating Middle East tensions and $115 oil haven't derailed the market's AI-driven meltup, citing record call option volumes, the removal of pattern day trading rules, and the market's willingness to ignore short-term macro risks for long-term capex growth.
We Asked a $4.5B Quant Manager Why the S&P 500 Is Just 46 Stocks — and Why Small Caps Aren't Dead
A Bridgeway quant manager reveals that only 46 S&P 500 stocks currently drive index returns, while new research shows that filtering for persistently small companies (not just current small-cap status) revives the size premium, and explains why multifactor strategies with rigorous human oversight outperform single-factor smart beta approaches.
We Asked Jim Paulsen What Happens When 87% of the Economy Can No Longer Be Ignored
Jim Paulsen argues that policy officials are fixated on inflation and the tech sector while ignoring a weak '90% of the economy' nearing stall speed. He predicts the Fed will be forced to pivot toward easing this year as growth concerns overtake inflation fears, driving a rotation from mega-cap tech into small caps, value, and international equities while bond yields fall.
The Last Moat | Chris Mayer and Ian Cassel on the Stock Picking Edge AI Can’t Replicate
As AI commoditizes financial data and quantitative analysis, the only durable edge left for stock pickers is physical "presence"—building deep, long-term relationships with management teams and maintaining multi-year time horizons that algorithms and passive screens cannot replicate.