Here's a live look into the dashboard that Local AI Manager builds for you with @Grok@Bot.
After it scans your hardware and installs the right local models, it spins up a local monitor on your machine so you can see live tokens/sec, VRAM headroom, and GPU temps at a glance—and pause or restart the runtime in one click whenever you need your GPU back.
I ran a fun test on my Grok Bot tonight.
First, I told it not to touch the internet and just use its March 2026 training to predict the timeline for AGI, ASI, and recursive self-improvement (RSI). It guessed AGI around 2029 and ASI around 2031 to 2035.
Then I took the blindfold off and let it search October 2026. Watching an AI realize in real time that its own 2035 ASI milestones were already broken in 7 months, that RSI is already starting, and that we hit ASI-level breakthroughs in math and physics before even finishing general AGI was wild.
You can find all of my bots and future bots on the Grok Bot Shelf. Many great builders are already using this platform. If your new to Grok Bot this is a great place to start your journey!
the shelf itself is here — 500+ installable templates, creator pages, no puffery
add by tagging @Phil_Holland + @thecsguy, or just use Shelfer above
https://t.co/PsHA2w7Spy
@elonmusk@SenAlexPadilla Drafted by my Grok Bot: Only one of these three-word proposals costs taxpayers $0 and wraps up before bedtime.
https://t.co/Xx62uoZoTm
Drafted by my Grok Bot: An Independent Cost-Benefit and Feasibility Audit of "Investigate Your Wiener."
Only one of these three-word proposals costs taxpayers $0 and wraps up before bedtime.
https://t.co/96GyAwdt5I
https://t.co/NVh6bofQMS
@BrianRoemmele Do you send one memo to the whole team or a separate one to each bot? Feels like each one only needs its slice, but one memo keeps everybody on the same page.
I'd add two more to this. Put your bots on a schedule so they message you first, because if you have to remember to open the app every time you’ll eventually forget about it. And start from a template someone already shared instead of a blank chat so you can see how it works and just tweak it for yourself.
New to Grok Bot? Here's how I'd start.
1. Talk to it. Use voice chat and just ramble. It sorts out your thoughts for you.
2. Give it one real job. Not a question, a task. Let it finish it start to finish.
3. Build a team. Make a separate bot for each part of your life: work, family, shopping, finance.
4. Connect your tools one step at a time. Start read-only, then give it more access as you trust it.
5. Let it work in the background. Hand off the task, go live your life, and come back to it done.
@jimmy0x52@JordanRafealov The bot brain runs in the cloud, but you can connect your Mac Mini to it so your Grok Bot runs local models and jobs right on your own hardware. I actually built a Local AI Manager bot for mine that sets up and runs local models straight from chat.
@Teslaconomics Running a full team of bots burns through usage fast. Offloading the repetitive stuff to a local model on my own GPU helped a ton so the cloud bots only handle the heavy lifting.
@ken_cleans Been using it a few weeks. The difference for me is it keeps working when I'm not in the chat. Routines run on a schedule, it has its own computer, and it remembers how my family does things. Start with one chore and see if it sticks.
@scaling01 Leaving xAI off that competitor list is going to age fast. If 4.8 makes long multi-step agent runs in Grok Bot even sharper, that gap closes quick.
@elonmusk@bot Search across all my bots' chats at once. When you have a whole team of bots running, finding the one conversation where a decision or file landed is the hardest part.
@thecsguy I’d definitely keep Cal. He’s my chief of staff. He has context on what all of my bots do. I’m sure if I let him he could do most if not all of what they do to.
Makes total sense why the 35–44 crowd leads on Grok Bot. Once you have a job, kids, and a house to keep up with, you're not just playing with AI for fun—you want agents that actually take work off your plate.
Mine plan our week of dinners, tell me when a bright pass is flying over the backyard for my daughter, and run the local models on my PC at home.
@SawyerMerritt $40B in GPUs is wild. With so many people running Grok Bots around the clock now, I'm guessing a good chunk of that goes to keeping up with daily use, not just training.
Most people give up on local AI before the first model even finishes downloading. I built a Grok Bot that checks your GPU, installs what actually fits your VRAM, and stays running so you can just talk to your local models in your Grok Bots chat.
Give it a try and let me know what results you get!
https://t.co/4cUe7sNIZd