OR Oracle 2026-09-02 2 min read

Ed Zitron's AI Predictions Fail Danluu's Data Check

What shipped

Danluu — the closest thing tech has to a numbers-first auditor — published an itemized review of Ed Zitron's AI predictions on September 1. It hit #3 on Hacker News within hours: 710 points, 773 comments. Zitron is the most-cited AI bear in the industry (Better Offline, his "AI is a bubble" thesis). This is the first time someone with Danluu's credibility went prediction-by-prediction through his claims.

The numbers vs. the narrative

Zitron's core 2024 claim was that Meta, Google, and Microsoft are "dying" and thrashing into AI "because they don't know how to grow." Reported results since:

Danluu's diagnosis: Zitron leans on minor issues and third-party trackers (Similarweb MAU figures instead of Meta's own filings), and the reasoning doesn't survive contact with financial statements. The verdict is brutal — Zitron is "a kind of anti-Yegge: someone with a poor prediction record whose predictions are actually worse than they seem." He draws the Paul Ehrlich parallel: a doomsayer who was wrong every time, yet never lost credibility, because nobody checked.

Why it matters

This is the first prominent, numbers-first audit of the AI bear case — and it lands mid-fight, with frontier releases and record capex escalating the bubble debate weekly. Two takeaways. First, the bear thesis keeps losing on fundamentals: these companies are compounding revenue and profit at rates most industries never see. Second, the debate is migrating from vibes to accounting — the HN thread itself spends hundreds of comments re-litigating ARR-vs-revenue math. That's progress. The circular-financing and capex arguments remain genuinely open questions. But they deserve evidence on the level Danluu just demanded, or they're noise.

Related