The Exponential Horizon

Andrew Kang, writing on X:

Being overly concerned about bubbles is foolish. Trying to time the market is foolish. Short term fluctuations and corrections will always happen but are entirely noise since we're in such close reach to the singularity.

We will compress more technological progress and economic growth in the next 2 decades than in the entire history of civilization combined.

There's something structurally interesting about an argument that begins "for those who have experienced at least one market cycle" and then proceeds to explain why everything those cycles taught you is wrong.

The tell is in the unfalsifiability. If prices rise, the thesis is confirmed. If they fall, that's "noise." If ASI arrives in 2027 or 2029 or 2035, "it doesn't really matter." This isn't analysis. It's eschatology with a brokerage account.

Here's what's frustrating: there is a real thing happening. AI capabilities are improving fast. Development velocity at companies using AI tooling well has genuinely increased — not by the "triple digit percentages" asserted here without evidence, but meaningfully. These are observable, measurable phenomena worth understanding.

But Kang does what the AI hype cycle always does: takes a real trend and extrapolates it through six layers of speculation until you arrive at "data centres in space" and "multiplanetary colonisation" within a decade, presented with the same confidence as next quarter's earnings.

The dot-com era is instructive here, and not in the way Kang thinks. The internet was transformative. It did change everything. People who said "this time is different" were right about the technology and catastrophically wrong about asset prices. Pets.com didn't fail because the internet was fake. It failed because "the technology is real" and "this price is rational" are completely different claims.

The most revealing line: "Traditional valuation models are unequipped to price these changes." Every bubble produces someone explaining why the old tools for measuring value no longer apply. Sometimes they're right. But when someone tells you to stop measuring, it's worth asking what they're selling.

Kang's piece will resonate with people who are already invested and want permission to hold. It will not help a single professional make a better decision about how AI actually affects their work. That's the gap — and it's wide enough to drive a business through.