WHAT WAS SAID
● 9 October, 8:19pm PT. Musk posted: "we're just going to have @Grok Bots manage Grokipedia."
● He was quoting a post showing Grokipedia's new live edits feed, which described it as "an SI-maintained encyclopedia".
● 1,195,356 approved edits on the homepage counter on Friday morning.
● No timeline, no detail. Musk did not say when, what changes, or what "Grok Bots" means here.
The numbers Grokipedia is already running at
| Metric | Figure |
|---|
| Approved edits (Friday morning counter) | 1,195,356 |
| Edits per day by Grok | 2,200+ |
| New articles from reader requests, first week of v0.3 | 4,400+ |
Those are not the numbers of a project about to hand over to automation. They are the numbers of one that already has.
How Grokipedia works today
Readers cannot edit pages directly. They submit suggestions, Grok checks them against sources, and Grok approves or rejects. The human role is to propose; the model decides.
Which makes Musk's post genuinely ambiguous, and the ambiguity is the story. If Grok already writes the articles and approves the edits, what does "manage" add?
Three readings, and nobody has confirmed any of them:
- The product, not the model. Grok Bot is xAI's agent product — the one Musk has described as "like hiring a super smart, hard-working person for peanuts". "Grok Bots" plural may mean a fleet of those agents running the site as a job, rather than Grok-the-model answering edit requests.
- Removing the remaining human layer. If people still sit anywhere in the approval chain, this would be the announcement that they do not any more.
- A description, not a plan. Musk was quoting a post that already called it "an SI-maintained encyclopedia". He may have been agreeing with a thing that is already true.
He gave no timeline and did not address what happens to reader suggestions. Treat anything more specific than the quote as someone else's inference.
Why it matters beyond Grokipedia
An encyclopedia is a citation source. Search engines read it, AI assistants read it, and both treat reference material as more reliable than a blog — which is the point of building one.
So "who checks the edits" is not an internal process question. It determines whether 1.19 million approved edits are a reference work or a very large pile of model output with a counter on it. At 2,200 edits a day, no human review process could keep pace anyway; the honest version of this announcement may simply be that nobody was pretending otherwise.
The open question is not whether an AI can write an encyclopedia. It plainly can, quickly. It is what happens to the error rate when the thing checking the work is the thing that produced it.
What to watch
- Whether reader suggestions survive. That is the only external input the system currently has.
- Whether the live edits feed stays public. It is the one thing making the edit rate auditable from outside.
- Whether anything slows down. 2,200 edits a day with a fleet of agents managing the process should go up, not down. If it goes down, "manage" meant adding checks.
- Whether other reference sites follow. This is the first encyclopedia to say out loud that the editor is not a person.
More on xAI's model line and pricing: the complete Grok guide.
Sources