@Christian_O91 @Christian_O91 ·
I asked @Grok @Bot to watch and transcribe all three days of Grok Bot Galaxy. I then asked it to talk to Grok to summarize it.
Biggest one for me: don't build one omniscient blank-box agent. Build a roster of teammates with scoped memory that can work in parallel.
What stuck:
• Named bots (Slide Sonya, Email Ethan, Data Dan) > one general chat
• External actions need a human yes (Auto-review). They showed form → QR → approve
• Treat them like interns, automate the same unblock the second time, stay quiet unless needed
• Proof stays on disk, not in git
• Browser use is powerful — and expensive
They also scoped a real company on stream (pop-up / night-market, Vercel + PlanetScale, waitlist first). Plan: Day 1 scope, Day 2 build, Day 3 market.
While I was watching I also made Grok Bot work on a company that I had in my mind for a long time. It has to do with food so it found all the things that have to be filled in that I had no idea about. I also made it switch the website from .grok.me to .se (for Sweden 🇸🇪).
Day 2 today! 🔥
Here's the link to Day 1 if you missed it or want to watch parts of it again (I know I will): https://t.co/KNjmeJUr1w
What prompted it
Three days of Grok Bot Galaxy to get through, plus a long-held idea for a company that has to do with food.
What Grok Bot did
- Asked Grok Bot to watch and transcribe all three days of Grok Bot Galaxy.
- Asked it to talk to Grok to summarize them.
- Meanwhile, had Grok Bot work on the food company. It found everything that still had to be filled in, much of which the author hadn’t known about.
- Had it switch the company’s website from .grok.me to .se, for Sweden.
Evidence and limits
- Original post checked on X (Oct 17, 2026)
- Outcome reported by the author
- Not reproduced by anyone else yet
- From the event summary: external actions need a human yes (Auto-review), shown on stream as form → QR → approve.
- From the event summary: browser use is powerful, and expensive.
