Meta's Muse tops the App Store on UX, not new agent capabilities
- Meta's Muse sits at the top of the US App Store free-app chart above ChatGPT and Momo, even though its core capabilities have existed for over a year in Codex, Claude Code, Hermes, and OpenClaw.
- Muse collapses the interface into one chat: no model selector, no chat vs work toggle, no slash commands, and no mention of MCPs or cron jobs, which the author calls the fix for ChatGPT's growing maze of choices.
- The post credits design, product marketing, and Meta's ad-driven model for funding the tokens, and argues Muse may be the first time agentic AI clicks for an average consumer.
- OpenAI shipped dots, its always-on agents inside ChatGPT, while the piece was being written.
Hacker News opinions
You can get root on the VM your instance runs in, hook it to your tailnet, and run a torrent seedbox on Meta's own infra.
Persistence is the real problem. bashrc, bash profile, cron, systemd all get wiped the second the VM resets. Has anyone found a launch script that survives that?
I don't get Muse. ChatGPT already books tickets for me and spins up a cloud browser when there's no API, and now OpenAI has Dots as the always-on version of the same thing.
It's basically a Meta-hosted OpenClaw VM for every Facebook user. It cuts the friction for your parents to run an agent with zero help.
UI/UX is the whole story. Same capabilities underneath, but a lay person can actually understand and use it.
Facebook's install base is 3x ChatGPT's. That alone is enough to get this to the masses.
The article says it plain: good product design, personality mechanics, and ad-funded tokens. Technically there's nothing new.
This is the Dropbox comment of the AI assistant era. At least it's not quite as out of touch.
Claiming Google has no product chops because they haven't won consumer AI is wild. YouTube, Maps, Chrome serve billions. I think Muse gets crushed.
Those are all 20-year-old products they maintained. Keeping an entrenched product running doesn't mean you can launch a new one. Their consumer AI strategy is incoherent, just NotebookLM and nano banana hits.
I asked Gemini to play jazz on YouTube Music with my premium account on the same device, and it started a Spotify playlist. It couldn't even open Maps in navigation mode.
Privacy is what blocks me. Do I hand Muse my identity, mail, calendar, and phone so it can book appointments and pay bills?
Most proactive agent use cases need data I won't give up. I want a self-hostable version I can deploy to my own AWS or Fly and point at a ZDR provider like Fireworks.
Local models on an iPhone will get good enough eventually and the privacy problem just goes away.
None of this is valuable unless you're fine with your data being used against you. Meta ends up knowing your family's health, your bills, your doctors, who services your loans.
The ChatGPT clunkiness rant is right times 100. The sidebar is a jungle, and I never remember if it's Codex or chat, local or remote.
Tibo's on X asking what should go in one more sidebar, which tells you everything.
They just copied Claude's Cowork split because it sounded enterprise-friendly, then pushed it onto regular users.