MediumI handed Muse my Inbox. And I’m Amused! | by Tabinda Khan | Oct, 2026 | Medium
Sitemap
Tabinda Khan

Tabinda Khan

Product Marketing Leader. Part strategist, part student. I use AI for efficiency and curiosity. I write about what I experiment and learn along the way.

Press enter or click to view image in full size

I handed Muse my Inbox. And I’m Amused!

3 min readOct 8, 2026

--

20,865 emails, one reckless experiment, and what Meta’s new AI agent did.

20,865 Reasons to look away

One of my Gmail accounts was sitting at 20,865 messages. I’m not proud of that number. It was the kind of backlog you stop seeing after a while, when it just becomes wallpaper.

This week I decided to deal with it. Not by reading them. By handing the whole mess to this new agent that kept popping up while I was on Facebook. It looked cute and started telling me what it could do for me, so I gave it a try.

Pick a Number, Any Number

Here’s what it did. Once connected to Gmail, it scanned everything in the inbox and grouped it by sender: 2,700 LinkedIn notifications, store promos, Uber receipts, years of it. Then it gave me numbered groups, and I just said “trash all 24” or “24, 52, 7, 43.” By the end of it, the inbox was at zero unread. Trash in Trash!

The Great 371 Resurrection

It wasn’t all perfect. At one point it trashed 371 emails I hadn’t meant to lose. I told it to get them back, and it restored every single one.

Get Tabinda Khan’s stories in your inbox

Join Medium for free to get updates from this writer.

Ctrl+Z is a Love Language

I delegated aggressively because every action was reversible. Gmail’s trash holds everything for 30 days. That’s what let me move fast and say “trash it anyway” without agonizing over each group. Trust in an AI agent isn’t about it never making mistakes. It’s about mistakes costing nothing.

That’s the part the think pieces miss. Everyone argues about whether agents are smart enough. Wrong question. The real question is whether the environment lets you recover when they’re wrong. Give me reversibility and I’ll hand over the keys. But Muse can also send emails and make purchases, and those can’t be undone with a click.

From Party Trick to Morning Routine

For now Muse is a part of my routine. It prepped me for a consulting call with fresh industry data seven minutes before I dialed in. Yes, it has access to my Google Calendar.

A small one from tonight: I asked it to draft a content piece while using it on my phone, then told it to save the draft to my Google Docs. It created the doc and filled it in.

Errands, Not Fireworks

None of this is glamorous. It’s errands. But that’s the point. The most useful technology disappears into the boring parts of your life.

I also wondered whether Muse cared more about collecting data on me than helping me. So, I did opt out of having my Muse interactions to be used to train Meta’s models.

Amused, Not Dazzled

Meta lets you give the agent a custom name and avatar. I’m still forming my opinions on all of this. But I know one thing: Muse didn’t amuse me by being brilliant. It won me over by being fast, proactive, and intuitive. The future isn’t an Agent that’s never wrong. It’s an AI you can afford to let go wrong with options that you feel you are in control of.

--

--

Tabinda Khan
Tabinda Khan

Written by Tabinda Khan

Product Marketing Leader. Part strategist, part student. I use AI for efficiency and curiosity. I write about what I experiment and learn along the way.