AI Too Dangerous to Release, Perfect Investment Thesis

The fastest-growing product category in world history is apparently so lethal that its creators won't let humans touch it. OpenAI and Anthropic have achieved the holy grail of venture capitalism: infinite upside, zero accountability, zero product demos. Investors are now betting billions on black boxes with warning labels, which is either genius or a masterclass in marketing through fear.

Within 60 days, we've learned that AI models are so powerful their makers fear releasing them, yet so advanced they're now building themselves. This is not a paradox—it's a business model. When your competitive advantage is that you refuse to show anyone your product, you've transcended traditional notions of evidence-based investing. The fastest-growing category exists in a state of perpetual beta, forever about to change everything while remaining safely immune to actual scrutiny.

The press releases write themselves: "Responsible AI development requires extreme caution," reads as "Please continue valuing us at $100 billion while we withhold the thing you're paying for." Self-improving models are particularly elegant—investors can't verify they work because they're updating faster than due diligence. It's the ultimate bull case masquerading as a risk warning.

In venture capital, the best products are the ones no one can actually evaluate. This is simply taking that principle to its logical conclusion.