SAN FRANCISCO — xAI has launched Grokipedia, an online encyclopedia written and continuously updated by its Grok artificial-intelligence models, pitching it as a faster and more transparent alternative to crowd-edited references.
Elon Musk unveiled the project as a "truth-seeking" knowledge base that reads primary sources, cross-checks competing claims, and attaches citations to each statement it publishes. The launch catalog spans hundreds of thousands of entries covering science, history, technology and current events, with the system rewriting articles automatically as new information appears.
An Encyclopedia That Writes Itself
Grokipedia is built on the same models behind the recent Grok 4.6 release that rolled into Tesla vehicles, tuned for long-form factual writing rather than conversation. Instead of waiting for volunteer editors, the system drafts an entry, pulls supporting documents, flags where sources disagree, and leaves a visible trail so readers can follow the evidence themselves.
That approach leans on raw computing power. xAI has said Grokipedia runs on the enormous training and inference capacity it assembled this year, the same infrastructure that recently came online when the company brought its Colossus 2 supercomputer to full power. The scale lets Grok refresh entries on fast-moving topics within minutes rather than days.
Why Musk Wants It
Musk has argued for years that widely used reference sites drift toward a narrow editorial viewpoint and resist correction. Grokipedia is his answer: a reference that, he says, favors sourced evidence over consensus gatekeeping, and that is cheap enough to maintain at a scale no volunteer community could match.
The strategic logic also fits the group's broader push into the enterprise. A rigorously sourced, continuously updated knowledge layer is exactly the kind of grounding that business customers want behind an AI assistant, and it dovetails with the recent move to sell private Grok models to large corporations. A trusted factual backbone makes every downstream product — search, chat, coding help — more reliable.
Accuracy and the Road Ahead
The obvious risk is the one facing every large language model: confident errors. xAI says it has leaned hard against that, requiring a citation for each factual claim, exposing a public edit history, and letting readers flag disputed passages for re-checking by the model. Early entries pair clean, readable summaries with footnotes dense enough to satisfy a researcher.
Whether Grokipedia can match the trust that established encyclopedias built over decades will take time to judge, and Musk has invited exactly that scrutiny by making the sourcing visible. For a company that has spent the year stacking compute and shipping models at a rapid clip, the bet is characteristically ambitious: that an encyclopedia written by a machine, with its work shown, can out-update and out-cite the human-edited web. Readers can judge the first edition for themselves at xAI's site, and the company says the catalog will keep expanding daily.
