I’ll admit it. When I saw the words “OpenAI lost” at the top of a LinkedIn post, I figured it was clickbait. Then I read the whole thing and realized the data behind it is pretty interesting.
The author was looking at Feedback Bench 1.1, a new ranking from a company called Enterpret. It compares 9 personal AI agents based on what real people say about them in public posts on Reddit, X, YouTube, the App Store and the Play Store. The winner surprised a lot of people: Meta’s Muse. The original poster even says upfront that Meta isn’t sponsoring them. They just think Muse is the best agent right now.
Here’s what the expert found, one point at a time.
1. The ranking measures what users say, not lab scores
Most AI leaderboards test models in a lab. Feedback Bench takes a different route and listens to what users are saying online. According to the creator’s breakdown, it weighs three main variables:
- Popularity: how many people talk about the agent on social media. The author’s view is that “hype is a good proxy for usefulness.”
- Customer love: how users rate the agent compared with the others. A score of 0.5 is the peer average, so anything above that means people like it more than usual.
- Getting tasks done: the heaviest-weighted area, at about 25%. Users care most about whether the agent can actually do things, and the original poster agrees.
I like that last one a lot. Clever answers are nice. An agent that actually finishes the job is what people pay for.
2. Meta takes gold and OpenAI lands near the bottom
Here’s the full ranking the LinkedIn creator shared:
- 🥇 Muse (Meta)
- 🥈 Hermes Agent (Nous Research)
- 🥉 Grok Bot (xAI)
- Hark (Hark Labs)
- OpenClaw (OpenClaw Foundation)
- Instinct (Spear Street)
- Gemini Spark (Google)
- ChatGPT Dots (OpenAI)
- Poke (The Interaction Co.)
What jumped out at me: an open-source project from Nous Research beat both Google and OpenAI. Gemini Spark came in 7th, and ChatGPT Dots finished 8th out of 9. That’s a big upset for the biggest name in consumer AI.
3. “Agent” means more than a chatbot
This part cleared up something that confuses a lot of people. Technically, ChatGPT and Claude are “agents” too. The author thinks people are getting ahead of themselves with that word. Tools like Muse, Hermes and Grok Bot go further.
The post’s author highlights three differences:
- They use their own computer: these agents run on their own machine, not yours. This is called “computer use,” and it lets them get things done on their own while you do something else.
- No model picking: you don’t choose an “intelligence” level. No menus of models or effort settings to sort through.
- Built for personal tasks: think doing your taxes, booking flights, shopping online or filling out forms.
That last point matters most to me. A chatbot tells you how to book a flight. A personal agent books it for you.
Why it matters: the AI race is moving from “who has the smartest model” to “whose agent actually gets my errands done.” Feedback Bench shows that brand size doesn’t win that race. Real-world usefulness does.
4. Anthropic just joined the race
The original poster also noted that Claude just announced “Duet,” Anthropic’s own personal agent. It isn’t in this ranking yet, so the next version of Feedback Bench could look very different.
💡 How to use this
If you’re thinking about trying a personal agent, here are a few practical tips based on the findings:
- Start with one boring task: pick something repetitive, like filling out a form or comparing prices, and see if the agent finishes it without help.
- Judge by results, not demos: the ranking puts “getting tasks done” first for a reason. Test whether it completes the job.
- Look past the big brands: Hermes Agent came in 2nd. Smaller and open-source options are worth a look.
- Watch the newcomers: with Duet on the way, the top spots could shift fast.
I think this ranking is a nice reminder that public opinion can tell you more than a polished benchmark chart. Check out the full LinkedIn post to see the author’s take and join the conversation about which agent deserves the crown.