Playing this video loads it from YouTube (Google), which may store cookies on your device and receive your IP address. Click to play.
Video: "NEW Grok Bot Update is INSANE!" by Julian Goldie on YouTube.
What Grok Pages is
Grok Bot's Pages feature lets a bot build a live screen from plain English. You describe what you want to see - a table of today's tasks, a summary from each bot, a list of what needs your attention - and the bot builds and updates it as it works. No HTML, no code, no configuration file.
Grok Jarvis is Julian's application of this. One bot - the Jarvis bot - sits above all the other bots in the setup. Every morning it reads what each bot did, pulls any unresolved items, and renders a summary page. You open the page, say "Good morning, Jarvis", and it reads the summary out loud.
How the two-prompt setup works
The setup is simpler than it sounds. Prompt one defines the Jarvis bot's personality: concise, precise, no filler. Prompt two defines the Page: what sections to show, which bots to query, what format the summary takes. After a brief fix message to tighten up the layout, Goldie says the whole thing was running in about ten minutes.
In practice, the Jarvis bot uses Grok Bot's messaging system to ask each specialist bot for its status. Each bot sends back a structured reply, and Jarvis assembles it into the dashboard. This is a real multi-agent communication loop - not one model pretending to be multiple agents, but separate bots with separate memories and schedules talking to each other.
What is overhyped
The "INSANE" framing in the title is doing heavy lifting. What this actually is: a morning briefing dashboard built with a chatbot. The underlying functionality is real and useful, but the demo shows a personal productivity layer more than a business-grade workflow system. There is no error handling. If one bot fails silently overnight, Jarvis reports whatever that bot last said, not that something went wrong.
Grok Bot is also a paid product - xAI's pricing puts multi-bot setups at roughly $200 per month at the time of writing. That is reasonable for an active business use case, less so for a handful of occasional tasks.
What is genuinely useful
The multi-agent briefing concept has real legs. Most businesses running AI agents have no good answer to "what did my agents do yesterday?" - you end up logging into three separate tools and piecing it together. A single morning report, even a rough one, is better than that. The Pages approach also means non-technical people can read and understand what the agents produced without needing to check logs or terminals.
The part that stands out most is the voice readout. Saying "Good morning, Jarvis" and hearing a summary of your AI agents' overnight work is genuinely a step change from staring at dashboards. Whether that appeals depends on how you work, but it is a reasonable direction for AI agent interfaces to move.
Where this connects to NordSys
The agents we build and manage for clients already produce daily outputs - SEO reports, content drafts, lead summaries, stock checks. Pulling those into a single readable briefing is something we handle as part of the setup. If you are running a growing number of AI tasks and losing track of what they are doing, that is a solvable problem. Talk to us about how we structure agent reporting for working businesses.
See our AI Agents service →