Grokipedia Abandoned After Three Months, Not Compute

Elon Musk's Grokipedia, an AI-generated encyclopedia he promised would humiliate Wikipedia, hasn't received a single update in three months.

It launched in October 2025 with 885,000 machine-written articles and the standard Musk assurances of massive superiority. By April, it was abandoned—and nobody announced this, it just stopped.

This is not a new problem. Google built Knol between 2008 and 2012, a platform that would let expert contributors write definitive articles on their subjects with algorithmic curation to surface the best ones. After four years and millions in resources, they killed it quietly.

The bottleneck is not compute power, not algorithmic ranking, not even the initial quality of generated content—it's something both companies missed entirely. Someone has to read every article and catch the hallucinations, the confident fabrications, the plausible lies that AI generates as fluently as truth. At encyclopedia scale, that is not a part-time task, that is not a task that scales with better hardware—that is people, full-time, getting paid to be bored and vigilant and right. Wikipedia survives because millions of obsessives volunteer for exactly this, catching errors and arguing about tone and caring more about a thirty-word passage on industrial looms than most people care about anything.

Musk needed the same human infrastructure but wanted to buy it, wanted the product without the people who make it real. When you build something you believe will be obviously better, you inherit an unspoken obligation. You must keep it obviously better forever. That requires constant, thankless maintenance by people who will never be celebrated for the work they prevent from being wrong. Abandon that commitment and your innovation becomes a mausoleum. The lesson is not about Wikipedia or AI, it is about anything you create and then leave to run on its own.