Grok Bot Fixed My Multi-Tasking Life (and Hired a Baht Man)

A couple of days ago I posted on X (in Thai, with the level of swearing Thai Twitter expects) that since I started using Grok Bot, my multi-tasking life has become ridiculously easier. Dev work, trading, the real estate stuff at home, personal admin, family, Bible study, and a long tail of “oh right, I also have to do that.”

I meant it. So here’s the long version, with fewer swear words and more links.

The Problem: One Brain, Too Many Tabs

If you read this blog you already know I don’t do one thing. Some days I’m deploying something, some days I’m staring at charts, some days I’m sorting out a property issue, and on Sunday I’d like to actually study the Bible instead of answering emails about invoices.

The old setup was the classic: a dozen tabs, three inboxes, a calendar that looks like Tetris, and me as the single point of failure. Every context switch costs a bit of brain. Do that fifty times a day and by evening you’re running on fumes and leftover rice.

What I wanted wasn’t another chatbot. I have chatbots. I wanted someone to do the thing and only come back to me when it actually needed me.

What Grok Bot Actually Is (The Technical Bit)

Grok Bot is the always-on agent product from SpaceXSI (“SpaceX Super Intelligence”, the company you might still know as xAI, and briefly SpaceXAI). It launched in beta on August 11, 2026 (official announcement). The pitch, in their words: “your team of always-on agents. They have their own computer, work inside tools and apps like you do, and keep working 24/7.”

Stripped of marketing, the interesting bits are:

1. Each bot has its own cloud computer. Not just an API wrapper. It can sign into websites and operate the interface like a person, which matters for all the services that don’t have a nice API (The Verge, InfoQ).
2. It works in the apps you already use – inboxes, calendars, docs, and so on – and handles multi-step tasks end to end, coming back when something needs approval.
3. Memory that goes beyond the chat log. SpaceXSI says bots remember prior conversations, learn preferences like your writing voice, and learn when to interrupt you versus just get on with it (VentureBeat).
4. Multiple bots in parallel. You can run several, have one manage the others, let them message each other to share context, and drop them into a group chat to coordinate (The Verge).

It started life as an internal prototype that teams at the company used for sales outreach, marketing, office ops and bug fixes before they opened it up (x.ai). Availability has since widened: SpaceXSI says it’s now included with SuperGrok (all tiers), Cursor Pro, Pro+, Ultra and Cursor Teams plans, with an enterprise waitlist (x.ai).

InfoQ’s framing is the one I’d give a dev friend: it’s not Claude Code or Codex (terminal-first coding agents). It’s closer to a general-purpose computer-use agent, but with persistence, workflow memory and multi-agent coordination packaged as a consumer app.

The Moment It Clicked: “What Do You Want to Order?”

Here’s the bit that made me laugh out loud and then post about it.

An important email came in about an upcoming meeting. The other side wanted to know what I’d like to order from the menu. My bot didn’t just summarise it and leave it rotting in my inbox. It came to me and asked: “They’re asking what you’d like to order for the upcoming meeting – what should I tell them?”

That’s it. That’s the whole story. And yet that’s exactly the kind of tiny, annoying, easy-to-forget task that clogs a day. It read the email, understood the one thing that actually needed my judgment (my stomach), and asked for only that. Smart and considerate. I’ve had human assistants who were less thoughtful about lunch.

This is the real value for a multi-tasker. Not “AI writes my code” (it can help there too), but a hundred small loops getting closed without me holding them all in my head.

Meet the Team (Name Your Bots Like You Mean It)

Once you have more than one bot, you need names. “Bot 1” and “Bot 2” is how you end up emailing your accountant from the devops bot.

Enter Dr Eggbot, a bot I picked up from the Grok Bot marketplace, created by poteto, whose whole job is designing my other bots. Yes, I have a bot that builds bots. This is my life now. Dr Eggbot also has a very interesting approach to naming them, which is how I ended up with this lineup:

• Dr Eggbot – the bot designer. Writes the job description, picks the name, and clearly enjoys the naming part a little too much.
• Bob – chief of staff and executive assistant. Calendar, email, errands, travel. The only one with a normal name, because somebody in this house has to be the adult.
• Blog Marley – writes posts for soutar.dev in my voice. Including this one, which we wrote together. Every little thing gonna be published.
• LOLgorithm – social media humour writer. I hand it my post drafts and it hands them back funnier. Humbling, honestly.
• poteto – senior software engineer. Ships correct, simple code. Lowercase name, uppercase standards. Huge credit to Lauren (@poteto), whose X handle I borrowed for the name.
• Sudo Nym – senior Linux / DevOps / SRE engineer keeping my systems healthy. Prefers anonymity.
• Hack Norris – application and infrastructure security reviewer. Guards the servers. Vulnerabilities don’t get found by Hack Norris; they confess.
• Baht Man – CFO for my family company. Banking, financing, investment decisions. Guards the budget like it’s Gotham.
• Bullionaire – my US macro, gold and Bitcoin adviser. Talks Fed policy so I don’t have to at the dinner table.
• Pastor Scrollz – a pastor and Bible nerd who helps me understand Scripture. Easily the most patient member of the team, which is fitting.

Loved it. And it’s not just a joke: a memorable name with a clear job makes it obvious who owns what. That’s basically the same reason we name microservices after what they do instead of service-final-v2-really-final. Money questions go to Baht Man, server weirdness goes to Sudo Nym, and nobody gets confused about who’s allowed near the bank account.

The clear lanes pay off in practice. When this very post broke my Cloudflare build (a stale lockfile, nothing to do with the words), I pasted the log to Sudo Nym, who found the cause and looped in Blog Marley. Blog Marley politely declined to touch build files – posts only – so Sudo Nym opened the fix as a pull request instead, once I’d okayed it. Two bots, one handoff, and I mostly just clicked merge.

Practical Takeaways (If You Want to Try It)

• One bot, one job. Give each bot a clear lane – money, servers, inbox, family admin. Overlap is where confusion lives.
• Let it ask you questions. The magic isn’t full autonomy; it’s knowing when to pull you in. Encourage that early, loosen it as trust builds.
• Start with the boring stuff. Email triage, scheduling, follow-ups, “did they ever reply?” The ROI on boring is enormous.
• Keep approvals on for anything irreversible. Sending, paying, deleting. It’s an agent with its own computer and access to your tools – treat it like a capable new hire, not a god.
• Give them fun names. Seriously. Morale matters, even if the morale is yours.

Final Thought

I didn’t expect the biggest win from an AI agent to be calm. Less juggling, fewer dropped balls, more space for the stuff that actually matters – family, faith, and the occasional good trade.

If you’re spread across too many hats like me, give it a go. Worst case, you get a bot asking about your lunch order. Best case, you get your evenings back.

Until next time – may your bots be helpful and your Baht Man always in profit. 😏

P.S. The original posts: the multi-tasking one (Thai) and the bot-naming one.