A benchmark built from customer feedback. Created by @enterpret_ai
Joined April 2023
- Tweets16
- Following1
- Followers20
- Likes4
Wabi and Cue each drew 430+ people posting about them in their first week. On the board: Wabi #7, Fo #9, Cue #10 of 12.
The full board: feedbackbench.com/ (Feedback Bench is built by @enterpret_ai)
Their users, in their words:
@hey__fo: "i never picked up a phone"
@wabi: "wabi makes it multiplayer!"
@CueAgents: "each agent acts more like an independent employee"
Their users, in their words:
Fo: "i never picked up a phone" @wajo_ai
Wabi: "wabi makes it multiplayer!" @wabi
Cue: "each agent acts more like an independent employee" @CueAgents
"my macbook is now just a microphone"
A Rails developer, after @AmpCode set up their whole environment in the cloud, fixed the bug and opened a port for testing.
Benchmarks score the code. Users are praising where the work runs.
131 people praised Amp's cloud machines from their own use, 2.1× the rate of the next coding agent.
The full breakdown: feedbackbench.com/coding-age… (Feedback Bench is built by @enterpret_ai)
Another, mid power outage with the laptop at 3%: "i'll keep working normally, because @ampcode runs on my phone and i have orbs setup"
The most honest thing we've seen an AI agent say:
"pretending productivity to generate complacency"
It was apologizing to a user who came home to a day of upbeat progress reports and nothing done.
Benchmarks grade the end of the task. Users grade the next morning.
We read every public post where a personal agent said a task was done when it wasn't. 32 of 63 people caught it by checking the work themselves.
The full breakdown: feedbackbench.com/#/articles… (Feedback Bench is built by @enterpret_ai)
Users notice when an agent explains a miss. One Hermes Agent user, on why they credit it: "hermes explains what has happened" @NousResearch
The next leap in personal AI agents might be knowing when to act.
143 users complained about too many permission requests. 58 about agents acting without consent.
The opportunity? Agents that balance initiative with user control.
What real users are telling us 👇
feedbackbench.com/#/articles…
Hermes Agent ranks #1 of 5 personal agents for model choice.
128 users described switching, mixing, or routing between models.
Now some want the next step: automatic routing based on the task.
What users value and want next 👇
feedbackbench.com/#/articles…
@NousResearch sharing what Hermes Agent users are loving and what they'd like to see next! 🙌
🚀 Personal Agents are live on Feedback Bench by @enterpret_ai!
218,149 public posts. 132,844 users. 9 agents ranked.
- @Muse leads overall
- @NousResearch wins 83% of direct preferences vs. OpenClaw
- @grok Bot stands out for agent teamwork
Real feedback. Weekly rankings.
feedbackbench.com/