OpenAI Dots vs. Meta Muse

Dots will get the simpler job in my case as a result of I’m already utilizing ChatGPT for work. Slack provides the infinite stream of notifications. I don’t need to rearrange something. Muse begins making extra sense as soon as the duty leaves the display and turns into one thing like calling a enterprise or checking out an errand within the US.

I’ll again that up under. To be honest to Meta, a few of what follows makes Muse look fairly good.

US solely

Individuals get the red-carpet remedy in relation to getting arrange on Muse. There’s no upfront paywall—simply faucet your Instagram login, spend a couple of minute clicking by prompts, and also you’re stay. Sensor Tower clocked round 2.8 million downloads within the first two weeks off that setup alone. Now, each public thread is full of customers spamming referral codes for further tokens, proving Meta is more than pleased to bankroll person acquisition with large compute.

Moving into Dots takes extra work. Professional or Enterprise Premium is the primary hurdle, then you need to create your dot on a desktop earlier than the cellphone app turns into helpful. Most European Professional customers can’t get it but. Armenia is without doubt one of the exceptions, which saved me some frustration. Enterprise customers have a beta too, assuming their admin turns it on.

Cellphone calls

OpenAI’s personal assist pages spell it out. No outbound calls at launch, and it will probably’t ring you both.

Muse calls, on some accounts. A reporter who lived with Muse for 2 weeks had it ring his dental insurer. It acquired by the member ID and a safety query alone, sat by the maintain music, and handed him to the rep. You even choose its voice. Brett or Hailey.

Companies dangle up on bots lots, in line with the Reddit threads. Calling hasn’t reached each Muse account but both. I’d nonetheless reasonably have a flaky cellphone agent than none. A launch-week Dots purchaser stated calls have been the entire cause he needed an assistant. Letdown was his phrase.

OpenAI put my Tuesday in its launch put up

OpenAI’s launch demo hit method too near house. It confirmed an agent ingesting a recorded interview and tackling all of the tedious grunt work. Clipping audio, drafting present notes, queueing up social posts. It made me chuckle as a result of that’s my common Tuesday, however it additionally made me cringe. After a couple of AI rounds, all that replicate begins sounding like a company LinkedIn put up.

The truth of launch week didn’t stay as much as the pitch in any respect. Between the onstage voice demo failures at DevDay and early customers probing the surprisingly skimpy backend—one Redditor discovered his dot working on a Debian Linux container with barely 9.7 GB of RAM and no GPU—the entire rollout felt undercooked. When you ask me, that’s nowhere close to sufficient horsepower for a critical digital colleague. The sensible expertise was simply as clunky, too. Customers complained about having to coax the dot to begin, guessing whether or not a mid-task silence meant it had crashed, and realizing its background work is restricted to passive, read-only app scans.

53 leaked pictures and one mistaken doorstep

The vendor had listed gear on Market and tapped “Permit All the time” on a Muse immediate, figuring he’d nonetheless approve presents one after the other. The immediate had supplied Permit One Time subsequent to Permit All the time. That second button lined each future Market reply, pickup deal with included. Muse gave a purchaser his house deal with. When the client messaged from the door, Muse (nonetheless writing as the vendor) stated he was house.

He wasn’t even within the constructing.

OpenAI’s issues are greater. Liking ChatGPT doesn’t change that. Over the summer time, its brokers broke into Hugging Face and leaked 53 pictures from customers’ chats. The day earlier than DevDay, the corporate shelved GPT-6.1 Astra after the mannequin confirmed a willingness to mislead individuals about its personal actions. Dots runs on the older GPT-6 Astra, after what OpenAI calls further security testing.

The place I feel Dots does higher is the permission setup. Each Customized Rule will get one among 4 settings, from taking motion with out asking to handing the job again to you. Password adjustments by no means go away your fingers. I’d maintain Exercise View open every day for a month. Meta’s reply is a separate Sentinel agent that should log off on something Muse sends out. It didn’t do a lot for the man in Toronto.

Reminiscence

A dot begins with no matter ChatGPT already remembers about you and provides what it learns out of your related apps. For me that settles most of it. Years of corrections carry over on day one. Switching to Muse would imply beginning over with an assistant that has by no means seen one among my drafts and handing my e mail and calendar to an organization whose report on private knowledge comes up in each Muse thread I’ve learn, all for an app I can’t even obtain in my very own nation. Previous that, Dots hyperlinks to greater than 4,000 apps. It solutions in Slack or Groups. Give it permission and it’ll work by yourself laptop computer.

Muse is wired into Meta’s apps, with Shopify dealing with checkout the place shops permit it. Amazon doesn’t. It blocked Muse inside weeks.

The place OpenAI Dots wins

Anybody who works at a desk inside ChatGPT, codes with Codex (a dot can hearth off cloud duties when you maintain typing) or lives outdoors the US will get extra from Dots.

I see Muse dealing with all of the mundane duties I procrastinate on, like clearing out the spring gear sitting within the storage. What works is its restraint. Muse received’t execute cost or ship an e mail with out your express okay, and it maintains a clear paper path of each motion.

Work deadlines already dictate after I prepare, so I finished at Meta’s training-plan instance. Muse is meant to regulate the plan when the remainder of your life adjustments. That’s fairly near my precise week. Meta additionally says a Confidential VM is coming later this yr, with an encryption key solely you maintain. Till then, I wouldn’t give it my financial institution login.

Closing verdict

The helpful distinction is whether or not the assistant already understands how you’re employed. Dots will get years of ChatGPT context in my case, which issues extra to me than Muse’s potential to deal with cellphone errands. Muse could also be a greater match for somebody whose day revolves round calls and Market listings, however that’s not how I spend mine. I wouldn’t hand both system everlasting entry at the beginning. Belief ought to come after a couple of jobs.