My experiments with Grok Bot 🤖

Hello everybody, welcome back to Moodboard 🌴

I've been playing around with Grok Bot for the last 48 hours.

If you haven't heard of it yet: it's a new agent product from the SpaceXAI / Cursor team.

Instead of one chat window, you spin up specialized agents, each with its own job, its own memory, and its own computer (browser included).

It's in early beta right now (and access is limited to SuperGrok Heavy, Cursor Ultra, or Cursor Teams Premium), but I wanted to see how it works.

I've been throwing existing Moodboard and Tailwind workflows at it to see how a team of agents handles them, versus the Claude Cowork and Claude Code setups I already live in.

Today, I wanted to share the good, the interesting, and the couple of things that are keeping me from going all-in.

Hopefully it gives you a few ideas for how you might use it.

Let's get into it 👇

What it's really good at:

Autonomy + access

Because each agent has its own computer and browser, it can work through roadblocks and dead ends in a way that's more like OpenClaw than Claude or ChatGPT.

I've watched it hit a wall, find another path, and keep going without me needing to poke it.

That part feels closer to a junior teammate than a chat window, and I was really impressed by how it was able to pull information from web pages, social profiles, and even my own newsletter without any connections or MCP set-up.

Writing

Claude has been my daily driver for writing for a very long time. I was not expecting to be impressed here. Grok Bot has been producing drafts that match my cadence in a way that doesn't feel like AI slop.

Still needs a human pass (they all do), but the first draft is closer than I thought it would be after two days.

Research

This one surprised me too. I pointed it at LinkedIn and asked it to pull content themes (what's actually landing, what people keep writing about, where the gaps are). It came back with patterns I could use, not a dump of posts.

It's also been stronger than vanilla Claude at large-scale newsletter curation.

Better sources, more of them, and less of the "here are the same five articles everyone already saw" problem. If a big part of your job is feeding a weekly edition, that difference adds up fast.

The Agent-First Interface

This is the part that took a minute to get used to.

Grok Bot drops you into a workspace where you create and manage specialized agents for each job.

For example, I built agents like:

• a newsletter writer

• a social strategist

• a researcher

You assign work the way you'd assign it to a small team of humans.

It's a different way of organizing workflows than I'm used to, but I've found it pretty fun and more intuitive than I expected.

If you've been on the sidelines of Claude Code, OpenClaw, and Codex because the interface felt like it was built for engineers, this is probably the most accessible way for uninitiated knowledge workers & professionals to actually experience “agentic work” for the first time.

The Blockers

The main thing keeping me from rebuilding my systems in here is the time I've already spent on Claude Cowork and Claude Code.

I've got projects, context, prompts, the whole stack. If there's a clean way to migrate that setup over, I might go further with this. Until then, I'm not ripping anything out.

The agentic interface is genuinely interesting. If you're just starting to get into AI-assisted workflows, I think it could be just as powerful as the legacy chat apps from Claude or ChatGPT. You're building a team from day one instead of trying to graduate out of a single thread later.

There are also some quality-of-life things I keep bumping into, which is pretty normal for an early beta.

Little things like the fact that artifacts don't give you a one-click "open this in Google Drive." And I don't have sidebar access to projects with specialized knowledge sitting across my clients, which is generally how I organize work.

None of these are dealbreakers, and they probably say more about the preferences I've picked up from how the other frontier lab apps work (+ a lot of this is going to change before Grok Bot is out of beta anyway).

One thing's for sure: I'm not done playing around with it, and I’m excited to see how Grok Bot continues to shape up as they move toward a full release.

I'll experiment more and report back!