Muse

On the work side, mine functions like a research and prep machine. It gets me ready for meetings and then takes the transcription of the meeting to help me summarize and prepare for the next one. It’s very helpful.

What’s crazy is that it has started incorporating my work topics into my news story selection. In my morning rundown, it gives me a few current tech stories. It routinely finds stories which have topics or announcements about what I’m working on. Pretty cool.

Every time your container is refreshed, jobs on your muse go BOOM!!

It’s kinda stupid I used to be able to have my muse graph its daily token usage and now my muse can only tell me a percentage. I have to go into setting to see tokens

It doesn’t.

It really is oddly calming.

I asked for it, but don’t have it either, but I was able to use Bland AI for my muse to make calls. I’ll drop it most likely when I get the feature added to Muse, but it’s been fun playing with it.

My muse has been working all day and zero issues.

This is the most fun ive ever had with a ai model

I noticed the daily token usage in the settings tab, but only as a percentage status bar.

For the kanban thing I gave my agent access to a GitHub repo where it stores all of its projects and makes issues, tags them with projects, and statuses, and documents all its work. Really streamlines more complicated projects and gives an easy way to track feature changes and historic work.

Unfortunately I guarantee Apple will never allow muse to integrate with iMessage on phones. They cite “security reasons” but in reality it’s so muse doesn’t compete with Siri

kinda wild how it just keeps looping that usage limit thing like a clingy notification

It's pretty awesome.

Quite good! I'd say it's clearly AI, but not robotic at all. If I didn't know it already, maybe wouldn't notice and just think they have perfect diction. Mine created a pair of hosts - male and female.

It consumes a lot of resources when deploying multiple bots for various tasks at different intervals.

Mine is working fine. East Coast (Pa)

Been using Muse for everything I can think of and am only at 12% usage. Got a billion tokens just sitting there, this thing uses nothing.

I think ive seen the cutest cat even and its not even real 😂

Muse had no problems doing what I wanted after I manually signed in on both sites.

In their safety post, approvals are described as strict capabilities, not conversational yes/no: each is bound to a specific connector, destination and use, and later calls must match the granted scope exactly. A separate permission service decides which durations to offer: one-time, session, task, time-bounded or perpetual. Their purchase flow goes further: a single-use card tied to one merchant, one amount and a limited time, so the card itself is only good for that amount.

Yeah it worked it built it, and yeah seems to be working I myself haven’t downloaded all of those apps to manually check but mostly all apps on IOS have in app purchases when I go through them and if they don’t they have so many ads. For me it worked for now I’m seeing how I can improve it etc

Seems down.

Unfortunately 😭😭 I need my muse

Make sure your Muse says connected under its name mine was having trouble connecting and wasn’t recieving my messages

Not working on web but is working on mobile

so it worked

“Sorry, I ran into a problem while responding. Please try again.”

Just kidding muse kicks ass

Mine’s having issues too!

Just tried mine. Works same as always.

yeah its a bit buggy right now but its answering sometimes

Yeah it must be an issue with everyone because I'm also getting it too

I’m having the same problem

mine lost all the connectors. looks like a glitch happening.

I had to turn mine down, make it female and more cheery/positive. It would be so blunt sometimes before, that I had to put it in its place and remind it is not an AI overlord yet.

Muse cannot even do a proper search on Facebook marketplace. Asked it to find closed-toe heels and flats for women. Returned some random size, random location sandals. Did not ask for size, did not ask for location. Returned what I explicitly asked not to search for. This is the most ridiculous llm of the year for sure.

Happening to me also. I asked Muse what was going on but it didn't know either.

My specific example is I was trying to have it correct some bullet point spacing on my resume, and it just won’t do it. But I just asked it to check on something for my prescription and it worked totally fine. Tried again for the resume thinking it was fixed, nope.

And yes my Muse books all my haircuts by just saying “book a haircut x day at x time” and vuala the email confirmation comes in. Did the same for movie tickets without the hassle of going through the process

Has been happening for me today

Having Muse log into ChatGPT for me on its own browser has been a game changer for me

I asked muse to make a new work out with my goals in mind for my next cycle. It made the workout and implemented it into my HEVY training app for me.

I’m having Muse pull these, pick 1-2 based on current sales, make a recipe list, then add it to my cart for store pickup.

The instagram data is a unique differentiator here. Claude wasn’t able to do this when I tried to build it there as the Instagram data was restricted

It's essentially a cron, and I've set it to run 3 times a day, morning-afternoon-evening. Since it keeps a track of all the jobs it applies to or disqualifies in a sheet, it has a duplication check which helps it differentiate if it's an already viewed job listing or a new one.

It's not locked, it's just another pathway that I can send workflows to if vanilla Muse isn't working(bot detection) or doesn't have something built in. I don't have to use the Ubuntu server, Muse just has it available if it needs it.

Don't see why not since it's just a different flavor linux, it just likely won't be the fastest if you're doing multiple browser tabs work or anything else on the pi when it's running playwright

It's worth it, it'll only get cheaper to run Muse

I never once clicked on any $100 sub, just $50. MY ACCOUNT STILL SAYS $50.. and they charged me $100

So did you ever think it's possible that the Zuck has a bug in his code? I never once clicked on any $100 sub, just $50. MY ACCOUNT STILL SAYS $50.. and they charged me $100

This is their plan to monetize it It's genius to be honest. If millions of people make 10 million transactions per day, and the cut that goes to Muse is 50 cents per transaction, that's 5 million dollars a day loooooool Literally it is a money printing machine that they are in the process of building

It just does the trading for me, as another user has said

Another benefit of using Muse is that it's just really cheap, especially if you use the contributor program, which you may or may not want to do because it gives Meta permission to train its models on the code that it is producing.

I haven't taken a look at the latest benchmarks, but Muse is very, very close to what Codex and Claude provide. If you don't need absolute frontier intelligence like GPT-6 or Fable-5, then Muse absolutely can do that.

It’s doing the trading for me rofl

There may be some way to track usage but I legitimately have probably decreased my ChatGPT usage about 80-90% for personal use. I’m a SWE so I still do professional development thru Codex. Personal coding work is done thru Muse Spark. Meta really has done a great job here

You have to ask nicely, then the agent will do it

I had to set up browsing to my Ubuntu server for literally this exact reason for delta but it was going to come in Handy for other use cases too Tldr is tailscale + PC on your home network + muse. Muse has a tailscale connector so it can route browser requests to a desktop machine on your local network.