Two More Days Inside Grok Bot Galaxy: Building, Breaking, and Rethinking AI Teammates

From building my first bot to burning through my usage limits — a few honest thoughts after Day 1.

What I Tried

Of course, the first thing I did was play around with Dr. Eggbot.

I started with the basics: setting up a bot, giving it a name and description, chatting with bots, checking MCP connections, and figuring out how all the different pieces actually connect.

After that, I started thinking:

Okay, but what can I actually build with this?

I know some people had already been using Grok Bot for a month and had their own workflows figured out. I was basically starting from zero, so I wanted to find something I could actually build instead of just watching demos.

Then I noticed there was a challenge from Grok: create a template, get selected, and potentially get a chance to go to space.

I mean...

Yeah. Let's do it. 😂

I spent quite a while trying to come up with an idea that wasn't just another "AI assistant that does X."

Eventually, I decided to let some bots help me think through the idea and build it.

After a lot of wiring things together, I submitted the bot.

And honestly, just doing that gave me a pretty good tour of the Grok Bot ecosystem.

I ended up touching MCP, Bots, Chat, Mentions, Shortcuts, Public Bots, and using other people's Public Bots.

The downside?

I burned through my usage limit ridiculously fast.

I eventually upgraded to Pro just to keep experimenting.

And somehow, I've already managed to hit 100% usage again.

Cry. 😂

Some Thoughts

There Is a Lot Happening Under the Hood

One thing that became pretty obvious to me is that running this kind of system probably isn't cheap.

You have bots running in parallel, cloud computers, different tasks being passed around, SaaS integrations, authentication, permissions, and all the little things happening between the steps.

It made me think about the economics of building a startup around something like this.

If you had to pay for all of that yourself at scale, without the infrastructure and distribution of a company like X, I can imagine the numbers getting ugly very quickly.

I'm not saying I know what Grok's actual infrastructure costs are.

I'm just saying that after watching my own usage disappear that quickly, I started thinking:

How the hell does this scale? 😂

There Are Still Things I'd Love to See

One of the biggest things I'd like to see is better support for multiple bots working inside the same conversation.

I can imagine a much nicer experience where I could simply say:

"Load my product team."

And suddenly I have a predefined group of bots ready to work together.

Something like a bot cookbook or reusable team template.

Instead of rebuilding the same setup every time, you could import a group of specialized bots and start working.

That feels much closer to how I'd actually want to use AI agents.

If They're Teammates, Can They Actually Have Personalities?

We're calling these things AI teammates.

That got me thinking about something slightly more interesting.

If I have a CTO bot, a reviewer bot, and a prototyping bot, should they all talk and behave exactly the same?

Probably not.

Could we define their tone, personality, communication style, and even the way they approach problems?

And then there's the more interesting question:

Can we somehow inject part of our own way of thinking into them?

Not just:

"Remember my name and preferences."

I mean things like:

"This is how I usually evaluate a product."

"This is what I consider good enough to ship."

"These are the things I always look for before launching."

That starts getting much more interesting to me.

The Hidden Cost Is Real

Another thing I noticed is how quickly usage can disappear.

There are a lot of things happening behind the scenes:

• VM usage
• Multiple bots
• Tasks being broken into smaller pieces
• Tool calls
• Permission requests
• Authorization steps
• Human approval

And probably plenty of other things I don't see.

I hit 100% usage surprisingly quickly, even on Pro.

And honestly, I still don't completely understand why the demo felt so fast compared with my own experience.

Maybe the infrastructure is different.

Maybe the workloads are different.

Maybe there's a lot happening behind the scenes that I simply don't see.

I don't know yet.

But it's something I'm definitely going to pay attention to as I keep experimenting.

Also, having Grok Bot available from mobile is going to be important if these agents are supposed to become something we actually use throughout the day.

What I Took Away

The biggest thing I learned is probably this:

AI teammates don't mean you get to stop thinking.

Even if you have a bunch of capable bots working for you, somebody still needs to sit there and orchestrate the whole thing.

Which bot should do what?

When should another bot take over?

What needs human approval?

What should be rejected?

What actually needs to ship?

The bots can do a lot of the work.

But someone still needs to care about the result.

And this is where I think the whole AI thing can get a little funny.

With AI, we can build almost anything.

But if you don't have a consistent mentality around actually shipping something useful, you can just become incredibly efficient at building useless things.

You're not using AI.

You're just being dumb with AI, faster. 😂

AI Can Execute. But Can It See the Blind Spot?

This is probably the part I've been thinking about the most.

Even if you have AI teammates.

Even if you're an experienced engineer.

Even if you have a whole pipeline of agents researching, designing, coding, reviewing, and shipping...

There are still things they might completely miss.

Have you ever asked AI for ideas, got 20 pretty reasonable answers, and then talked to someone with years of experience who casually said something you never considered?

And you suddenly thought:

"Oh. That's actually the idea."

That's the blind spot I'm talking about.

AI can be incredibly good at execution.

But sometimes the most valuable thing isn't another execution step.

It's vision.

Knowing what question to ask.

Knowing what doesn't make sense.

Knowing what everyone else is overlooking.

That's where I think humans still have a very important role.

AI Teammates ≠ Human Workforce Replacement

At least from what I've experienced so far, I don't see AI teammates as simply replacing a human workforce.

I see them more like an extension of the person using them.

A wing.

You still decide where you're going.

The agents help you move faster.

And honestly, some of the guest speakers during the event reinforced that for me.

A few of them brought a really strong business and marketing perspective that I wouldn't necessarily get just by asking a bot to generate another plan.

That part was probably one of my favorite things about the whole event.

Sometimes you don't need another answer.

You need someone to make you think differently.

One Last Thing

I'm still waiting for my Grok Bot usage and credits to reset. 😂

Apparently I got a little too excited on Day 1.

But that's also probably the best sign that I was actually experimenting instead of just watching the demos.

I still have a lot of questions about how this whole thing works at scale.

And I definitely haven't figured out the "right" way to build with it yet.

That's kind of the fun part.

Let's build something great.

Happy Friday. 🤖