I Stopped Googling. I Didn't Mean To.
When I started testing Muse, Meta's new personal agent, and Instinct, the invite-only agent everyone seemed to be swapping codes for, I had no intention of giving either one access to my real life.
No real email.
No calendar.
No bank account.
I just wanted to poke around. See what they could do. See where they broke. Figure out whether “personal agent” was actually different from another chatbot with a slick interface.
Then something unexpected happened. I started using Google less. Not because I decided to. It just… happened. And that turned out to be much more interesting than any benchmark.
🔍 The first sign: I stopped searching
It took me about a week to notice. I was barely opening Google.
When I had a question, I started asking Muse instead. Not because I had consciously chosen it as my new search engine. It had simply become the thing I reached for.
Then came a very ordinary problem:
I needed airport parking for a day.
Normally, this would have been a Google search.
“Best airport parking.”
Then I'd compare prices, look at reviews, figure out shuttle times, and decide whether saving a few dollars was worth the extra hassle.
This time, I asked Muse.
It already had my flight details. I was flying out at 7:00 AM and returning at 10:30 PM.
Muse searched the parking options and came back with a few choices. But instead of just dumping a list of prices on me, it actually made a recommendation.
It reasoned through my day.
I'd need to be at the airport around 2 hours before my departure which is 5:00 AM, which meant I'd probably leave the parking lot around 11:00 PM on the way back.
One option was closer and the another was about a five-minute walk from the terminal and was roughly $10 cheaper, there was another option where the shuttle was involved, but the price difference was less than $5.
Muse's conclusion was essentially:
Why bother with the shuttle at odd hours for less than $5 in savings when you can park five minutes away, walk to the terminal, and still save $10?
That felt less like search and more like having someone actually think through the decision with me.
I picked the option. Then Muse opened the booking site in its own browser and started filling everything out for me.It got all the way to the payment page before stopping. The only reason it didn't finish was that I hadn't saved a payment card in its vault.
That was the moment it really clicked for me.Google would have helped me find parking.Muse had helped me figure out which parking made sense for my specific trip and then tried to book it.
That's a subtle but important difference.With search, I do the work after getting the answer.With an agent, the answer is only the beginning. It already knew my schedule. It considered the trade-offs. It made a recommendation. And once I made the decision, it started doing the work.
For something as mundane as airport parking, it felt surprisingly close to having a personal assistant. And once you've experienced that, going back to a page of blue links feels surprisingly manual.
📱 Instinct felt more like a person
On the other hand Instinct takes a completely different approach. There isn't really an app to open.You just text it , through WhatsApp or iMessage, the way you'd text a friend.
I started on WhatsApp.A few days later, I messaged it from iMessage instead. It replied with something along the lines of:
“Glad you found me here.”
It knew it was me. Nice little detail, But then I started thinking about why it knew.
What exactly was it remembering?
What information followed me from one conversation to another?
And how much of that memory would I actually want an agent to have?
That's one of the strange things about personal agents.
The features are impressive. The invisible parts are what make you stop and think.
🎂 I gave them a birthday invitation. Neither noticed.
Both Muse and Instinct are positioned as proactive assistants.
The promise isn't just:
“Ask me anything.”
It's closer to:
“I'll notice things for you.”
So I created a dummy Gmail account and sent it a birthday invitation.
Then I waited. Nothing !!
Neither agent proactively surfaced it. I tried again from a different email address.Still nothing.
My guess is that they may be much more focused on structured events, like calendar entries, than information buried inside email.
But regardless of the reason, it changed how I interpreted the word “proactive.” It doesn't necessarily mean:
“I'll notice everything important.”
It may mean:
“I'll notice certain things I'm designed to notice.”
That's an important distinction.When I actually asked them to look through the email, things got more interesting. Muse found the invitation and correctly noticed that it looked like a placeholder since I sent it from my personal ID to myself .
It asked what I wanted to do, drafted an acceptance, and then stopped.It waited for me to send it.
I liked that.
Instinct didn't really treat the email as an invitation. It seemed much more interested in finding something corresponding to it on the calendar. Different approaches. Different assumptions about what “helpful” means.
🧾 Then I tested receipts and the roles reversed
Next, I put a couple of expense receipts into the inbox. I also added an email with “URGENT” in the subject line. This time, the results flipped.
Instinct found both receipts. It also noticed the urgent email and suggested that I take a look without me specifically asking about it.
Muse found one receipt. It didn't mention the urgent email. To be fair, I hadn't asked it to look for urgent messages. But then I asked Muse whether it had missed anything. And this was the part that bothered me.
It didn't say:
“Yes, I missed something.”
It simply told me it could see two receipts.
So the information was apparently there.
It just hadn't surfaced it.
That distinction matters. Because if I'm trusting an agent to help me manage my inbox, “I found it” and “I told you about it” are two very different things.
🔓 The permissions question got real very quickly
This was probably the biggest surprise of the whole experiment. When I connected the dummy Gmail account, Instinct asked for access to my entire Google Workspace.
Muse's connection was much narrower.
I was very glad I had created a dummy account. It also sent me down a privacy rabbit hole.
I went back and read Instinct's terms after the test. As I understand it, disconnecting a service doesn't necessarily mean data that has already been indexed is automatically deleted; a separate deletion request may be required. That's the kind of thing I want to know before connecting my real inbox, not after.
Muse's security architecture sounds more thought-through on paper.But the strongest privacy model, where your data is encrypted with a key that only you control is still limited to trusted testers, with broader availability planned later in 2026.
So I'm holding off on connecting my real accounts for now. And honestly, I think that's a good thing.
We're entering a world where the most useful AI isn't just going to know things.It's going to have permission to do things.The permission model may end up mattering more than the intelligence.
⚖️ So… which one am I using more?
I honestly don't know which one is “better.”
And after a couple of weeks, I'm not sure that's even the most useful question. What I do know is that I'm using Muse a lot more.
Not because I decided it was the winner. It just keeps finding its way into more parts of my day.
It gives me my morning brief.
I ask it to find things to do with my family on the weekends.
It keeps track of reminders for me.
And right now, it's even tracking my flights for a year-end trip. 🤞
None of these are particularly mind-blowing tasks. That's actually the point. I'm not using Muse for some futuristic AI demo. I'm using it for the little things I would normally have to remember, search for, or organize myself. And those little things add up.
Instinct has impressed me too, especially with how proactive it can be. But Muse has become the one I naturally reach for. I didn't choose it. My behavior did. I have heard it can identify the booking and check-in on my behalf, haven't seen that one yet, that I will be interested in.
🔮 I think the interesting part is just beginning
We're still very early. Right now, Muse and Instinct feel like experiments in what a personal agent might become.But I'm starting to think the biggest change won't be some spectacular AI capability.It will be the gradual disappearance of small tasks from our mental load.
Finding something to do this weekend.
Remembering to do something tomorrow.
Figuring out where to park.
Keeping an eye on a flight.
Summarizing what happened overnight.
Searching for something that would normally take five Google queries.
One task at a time.
Eventually, you stop thinking about the agent as an app.It just becomes part of how you get things done.That's probably the biggest thing I've learned from this experiment.
I started by testing two AI personal assistants. I ended up changing how I use the internet. And I didn't even notice it happening.