Grok Build is here. Plan mode isn't security. | Promise & Risk of AI #shorts #ai #xai #grok

Añadido: 2026-05-18

[00:00:00]XA, I just shipped Grok build yesterday, their first coding agent. The internet has questions. Here's the 60-second honest read. It runs in two phases. Plan mode drafts a step-by-step plan you approve before any code touches your repo. Diff review surfaces every change.

[00:00:17]$300 a month on Supergrow heavy. What it gets right. Real competition for Claude code. A plan-first architecture that addresses the agent trust problem. A different model family for ensemble use.

[00:00:30]What it does not fix. Plan mode is a process safeguard, not a security control. Shell access plus file write equals a contractor with sudo on your workstation. Same risks, same governance gaps as every agentic CLI. The full 12-minute breakdown is live on the channel. Should you switch?

[00:00:51]Link in the first comment.