You might've noticed there wasn't a newsletter last week. It's the first one I've missed in a while, because I was on Seabrook Island with my son's fifth-grade class for a field trip about ecology (and some much-needed time with him). I could've had AI write one and put it on autopilot without a second thought. I still want to put thought and care into these, so I focused on what was most important that week instead of phoning it in.
Last Thursday night Jordyn and I went to the Westin for the Charlotte Business Journal's Power 100 reception. This year's theme was visionaries, and it's still sooo surreal to be on that list (I shared more about the night on LinkedIn).
When CBJ asked for my bold bet back in July, I wrote about the CLT Startup House and the irony underneath it.. "I run an AI company, and I think the AI age makes in-person community more valuable, not less." Greg Walter, president of Charlotte Motor Speedway, closed out the speeches by saying a list like this "isn't simply recognition of what's already been done. It's an opportunity, perhaps even a responsibility, to think about what we can do next."

It's been an assistant-heavy couple of weeks. The Thursday before, a room full of folks at House Jam (our monthly build night at the CLT Startup House) each built a personal agent. Then the day after the CBJ event I sat down with my co-hosts for Builders of Charlotte (our podcast, 4 of us building AI companies here in Charlotte) and we spent an hour on personal AI assistants. Most of what came out of both nights points at the same thing.. getting AI to carry more of the load so I can spend more of my time with people.
The episode isn't out yet, but you can subscribe to Builders of Charlotte on YouTube so you catch it when it drops.
1. Three assistants showed up in 7 weeks
If you haven't been tracking it, xAI's Grok Bot went into beta on August 11, Meta launched Muse on September 8, and OpenAI announced dots at their DevDay on September 29. All 3 run on their own computer in the cloud (i.e., they've got their own browser and can keep working in the background while you're doing something else), and all 3 can sign into your tools and get things done on your behalf.
The way I see it, they'll split by audience. Muse gets the folks who live on Instagram and WhatsApp, because it's easy and friendly and Meta is pushing it on literally every channel right now. dots gets the folks who already default to ChatGPT, because it's already sitting there. Grok Bot is the one power users will pick, because spinning up another bot for another job is as easy as asking for one. Charles (BotSource) thinks nobody wins and we all end up with a bunch of bots negotiating with each other on our behalf, which honestly might be the more likely answer.
My hot take is that Apple and Amazon are the dark horses, because they own the hardware. Amazon basically gave away Echo devices a decade ago to get Alexa into every house in America, and we've got like 8 of them in ours. My kids are trained on Alexa, and so is my mother-in-law, who doesn't use any other tech. Siri's an idiot right now BUT she won't be forever, and Apple has never needed to be first.. they weren't first to the tablet or the smartphone either.
2. The best use so far is the boring stuff
Charles was sitting in his 5-year-old's first Cub Scout pack meeting while they walked through all the forms parents have to fill out, so he told his assistant which troop it was and asked it to find the forms and fill them out. It did it right there in the meeting. Another time, while he was driving, it asked him whether he'd figured out childcare for the 2 teacher workdays coming up.. he had no idea there was no school, so he had it email the after-school program on the spot (it misspelled the principal's name, so there's still some human cleanup involved 😅).
Side note.. the amount of information schools send parents is absurd. Our kids' charter school runs most of its communication through GroupMe, and we get like 50 messages a day, so I built Jordyn a connection that reads GroupMe and pulls out only the stuff that actually matters.
I'm in Claude Code most of the day for real work (I've shared that setup in past newsletters), but I still use the personal assistants for life admin, where I just want to say "handle this" and not open a terminal. The memory systems these companies built are really good, and the forms/reminders/school emails are exactly where most of us fall behind.
3. Most people can build one in a night
At House Jam Nº 5 the question for the night was "What's one job you'd hand off tonight?" Everybody built their own agent, split into groups by experience (builders, vibe coders, explorers, and observers who just wanted pizza and to watch), and most of my group built theirs on their phones.
We used a simple definition.. an agent is a goal (one job and what finished looks like), tools (what it can reach, like your inbox, a sheet or a browser), a loop (how often it checks in and tries again, what folks in AI call a heartbeat), and a done rule (when it stops and hands the work back to you). Before anybody opened a laptop, they had to say 6 things out loud to their group: the job, what triggers it, what it reads, what it hands back, what done looks like, and what tool they'd build it in.
At show and tell, one person had never written code and used their phone to build an agent that compares health insurance options because their COBRA is running out.. it's pulling the marketplace plans, modeling what changes when 2027 rates come out, and working through the HSA rules, which is math they told the room they never would've done on their own. Another person built an agent that hooks into Claude Code and, after a set amount of time, opens the camera and makes them do jumping jacks before they can keep going (it won the $100 I'd offered for connecting an agent to the physical world, and they donated it to the next round of pizza). A third built a prototype for field service techs that records the repair work and blurs the homeowner's face so it's not recording the person, just the work.

4. The hard part is still hitting record
I've gotten in the habit of delegating out loud in meetings. I'll say "James is going to handle this part" while we're still talking, and afterward James (my AI chief of staff) reads the transcript and picks up what I assigned him. Friday morning, before we recorded, Juan from Founder Jam called me about the Founder Jam web page. I wasn't recording that call, so I told James the basic idea and he went and pulled the new site and built a landing page, first try. Normally I'd have to go back and forth with him on changes.
That still required me to stop and cue him up, though. You still have to hit the record button, and you still have to come back afterward and give direction. The next step I keep thinking about is an agent that's just on.. a mic pinned to my collar, listening all day, so I can stop doing the turn-based thing entirely.
Colin (Torta Studios) pushed back on this, and it's a fair push. When people know they're being recorded they get more performative and less honest, so an always-on mic could make everybody a little more curated. Where I've landed is that your phone is already listening for its wake word, Alexa's already listening, and privacy is mostly an illusion in 2026. I've kind of embraced that everything I'm doing is already recorded, so I might as well get some benefit from it.
That doesn't settle the consent question, and I think it'll be the hottest topic of the next couple years. Some folks will say "turn it on, what do I have to hide," and others will put on the tinfoil hat and head for the woods.
5. Give it context before you give it access
Whichever one you pick, context is what makes it useful. When I set up Muse and Grok Bot, I didn't connect all my stuff. I dropped in a couple of my context files (a primary file about me and StealthX, my writing guidelines, my email voice) and each one basically became a lighter version of James with way less access. I don't trust Zuck with my inbox, so Muse got the files and that's it. Within like 20 minutes it was rolling.. it read through my newsletters, turned a few of them into Instagram carousels, and (after I approved them) posted them to the StealthX Instagram, which we honestly barely post on.
If you haven't started writing down how you work, now's the time to start stacking those bricks. I walked through what goes in a context file back in January, and the same files have moved with me across every tool since. Colin's version for work is simple.. find 5-10 docs about you, your role and your company, upload them, and then tell it what you're trying to get done.

Wrapping up
I wrote on LinkedIn last week that I barely touch my computer anymore. Most of my day is meeting with folks, and the work moves in parallel because I talk to AI for a few minutes, it drafts something, and I review it before I walk into the next conversation. What I want from all of this is more time jamming with people at the house and more time with my kids, while the work keeps happening.
It's the same bet I made with the house. Start by picking one assistant and giving it one boring job this week.
Putting this into action
If you haven't tried a personal assistant yet, pick one and connect 3 things: your calendar, your email and your notes/reminders. Muse has a free tier, dots needs ChatGPT Pro or Business Premium, and Grok Bot needs a paid SuperGrok or Cursor plan. Be careful what you hand over, especially on a free tier.
Take a 10-15 minute walk and ramble to it in voice mode about everything you're behind on. Most folks don't know what to ask for at first, and the models are really good at turning a ramble into a plan.
What I'm reading
Introducing dots. OpenAI's announcement. Pay attention to the controls, where you decide what it can do on its own, what needs your approval and what's off limits.
Meta launches personal AI agent Muse. The plain-English version of what Muse does and where your data lives.
Grok Bot will tap Claude, Midjourney and Suno. Musk says Grok Bot will route each task to whichever model does it best, including competitors' models.
Onward & upward 🤘
Drew
P.s. A couple things coming up at the CLT Startup House. Community Conversation Part 2 is this Wednesday, October 14 at noon, picking up where "Is AI gambling with our lives?" left off (RSVP here). The Builder Block Party is Saturday, October 24 starting at 5, with food trucks, live music and demo tables, and we're working on a table where an agent listens to the room and builds things from the conversation in real time.. come see it. The next House Jam is Thursday, November 5, if you want to build one with us.





