SpaceXAI engineer Lauren Tan on using Grok Bot for self-verification
She says 99% of people treat Grok Bot as a chat box while the 1% let it run the app, read heap snapshots and verify its own work, producing 2,000 PRs last month with almost no human review.
SpaceXAI engineer Lauren Tan: "99% of you still use Grok Bot like a chat box. The 1% hand it a feature map and let it verify its own work." That one skill - verification - is the $300K+ line item in every AI salary band right now. Not prompting. Not vibes. Verification. Her agents run the app, take the traces, read the heap snapshots, hill-climb the fixes - before she wakes up. 2,000 PRs last month. Almost none reviewed by a human. The company she calls "a carnival" is hiring on exactly this. And the free talk where she builds it from scratch is still up. Bookmark it. The prompt era is ending; the verification era is billing.
This post quotes another post: x.com/0xfuckpoverty/status/2092708922788303067