I tried Meta Muse for a week, and it didn't go how I expected

I’ve been slow to warm to the benefits of AI, but that transition is definitely happening. I’m more comfortable using it daily, and I enjoy finding ways to put these new tools to work. If you had told me a year ago that Meta released a new agentic agent, I wouldn’t have cared, but as AI gets more accurate and easier to use, I’m more open-minded.

That’s what makes Muse genuinely exciting, and whether you love it or hate it, Meta’s done something with Muse that other AI companies haven’t yet meaningfully recreated. If you’re dipping your toe into AI, Muse should be your first stop for several reasons.

First impressions matter

Muse does a great job of putting the user at ease right away. There’s no intimidating blank page to fill, and I don’t feel the need to watch a YouTube video to optimize my prompts. On first boot, you’re greeted by Muse’s mascot, a creature that is a cross between a Teletubby and a Labubu. Thankfully, you can change it later, but initially, it guides you through Muse’s various functions.

I love the level of customization Muse provides. It’s presented as a text conversation, with your assistant sending little emojis in response or to acknowledge messages. I quickly changed my avatar to a friendly-looking orange cat named Lotus, and I do get a kick out of him typing away at his laptop while completing tasks.

Don’t underestimate the importance of aesthetics. AI remains daunting for many users, and Meta’s packaging will prove more successful than you might think. Of course, Muse has to do more than just look friendly to keep users’ attention for more than a few minutes.

There’s plenty of substance under the hood

Your little Muse avatar likes to explain that it has a computer at its disposal, which is a user-friendly way of telling you that it’s an agentic AI agent. Instead of just asking Muse questions, you can set specific tasks and even have it complete a purchase for you — though it stresses that a human gets the final say before any money goes out the door.

I still don’t know how comfortable I am with AI making purchases for me, but I’ve gotten solid use out of other Muse functions. I have mine set up as a personal RSS feed. Every couple of hours, it scours the internet for top stories across a wide range of topics. It’s fantastic that I can really hone in on what Muse sends me, and after a couple of back-and-forth messages, I get a pretty diverse list of articles across many of my interests.

Everyone is busy, and it’s easy to let some of your hobbies slip. I love collecting vintage electronics from Palm PDAs to Apple PowerBooks. I gave Muse instructions to scan eBay twice a day for the best deals it can find. The initial list it brought back wasn’t impressive; items were either in poor condition or didn’t quite match what I wanted. Thankfully, I could explain to Muse that I only want items that are new in box or in excellent condition, and I told it to sort by newly listed and to show only auctions with a Buy It Now option. It worked perfectly. I get a wonderfully curated list of new electronics twice a day. What normally would’ve taken a precious 30 minutes at night now waits for me when I’m done with work. More time to peruse and select, and less time searching.

I also connected my Gmail inbox and sent Muse on a deep dive. I get a ton of PR emails per day, and I wanted to filter and scan for the ones with product offers for me to review. Muse actually found an important one I missed, and it tracks whether I responded and what the PR agencies sent back. It’s a small task, but important nonetheless. Stuff like that adds up, and it’s why Muse feels like it’s turned a corner for average users in a way other AI agents haven’t, even if they have similar capabilities.

I’ve gotten better at splitting up duties

The more comfortable I’ve gotten with AI, the more I realize different models have their own strengths and weaknesses. It’s not simply a matter of picking one over the other. You may have a personal preference, but I’ve tried to create a symphony of different models working together. I like Claude for coding, and I have a running ChatGPT conversation for general reasoning and information. Muse absolutely adds to that mix. It lives on my home screen in a way no other model or agent has on my smartphone so far. It’s earned a spot, and while I still maintain a healthy skepticism of AI and its dangers, it’s important to keep an open mind.

Meta is offering pretty generous token packages on Muse’s free tier to start. With everything I have running, I only use about 6% of my weekly allotment per day. If you find yourself consistently maxing out, 500 million tokens per week will run you $20 per month, with a $100 tier for those who need 3 billion tokens.

I honestly don’t see running Muse for anything intense enough to justify $100 a month. However, for $20, it’s one of the first agents that produces value I can see weekly, in a package that is easy to use and understand. Meta is on to something with Muse, and don’t be surprised if it’s the AI agent that catches on the best.