Episode 145: Muse and Dots: Putting Personal AI Assistants to Work
Co-Host
Aytekin Tank
Founder & CEO, Jotform
Co-Host
Demetri Panici
Founder, Rise Productive
About the Episode
The next app you build could have just one user: you. Aytekin shares how he’s using AI to build software for himself, simply by describing what he needs. And that same idea goes beyond software. Want to learn something? Ask for a training video made around that exact topic. Personal tools, personal learning, built around you.
Like I was already a heavy user of ChatGPT, and I didn't think I needed another agent because I have so many agents and projects on ChatGPT. But it turned out I really needed one.
Hi, my name is Dmitri Bonichi, and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast, brought to you by Jotform and featuring our very own CEO and founder, Idan Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show.
Hello, and welcome back to another episode of the AI Agents podcast. In this episode, we're going to be diving into personal AI assistance. Idan, I don't know about you, but it feels pretty nice to have a little henchman, just me telling it what to do. It's enjoyable. I enjoy it.
Yeah. Hello, everyone. We've been using our personal agents last week to prepare for this, and I loved it. I'm so in. I was already a heavy user of ChatGPT, and I didn't think I needed another agent because I have so many agents and projects on ChatGPT. But it turned out I really needed one. I called it Miso, my Dots agent. I gave it a name and a nice avatar, and we'll both do demos to show it to you. Basically, Miso is kind of like my chief of staff or assistant. It has access to everything: my projects, email, messaging like Slack, WhatsApp, and text messages, and the memory in my ChatGPT threads and projects. It's kind of like a chief of staff, and I'm constantly asking it to find things out. It's also really useful for my personal life. I can ask whether my kids have basketball games or about their schedules because it has access to my calendar. It's become a really helpful assistant that's helping me figure things out, and I'm loving it. I know you're also spending a lot of time with Muse. I'm more of a Dots user, and you're more of a Muse user, so we'll compare them.
Yeah, absolutely. It's really cool. I think a personal agent is a new step forward. Neither of us has access to Duet, so we'll let you know what it's like when we do. It's an example of a proactive agent that has access to different plugins and can take action from somewhere other than your computer. It will proactively help you, learn from your communication, and essentially become more and more your assistant. This is something we've wanted for a while. The products available right now are pretty much Muse, Dots, Grok bots, and Duet, if I'm not mistaken. These companies are trying to figure out how to get you into their ecosystem, and I think this is a great way. I got interested in Muse because my girlfriend was nerding out about it on our trip to Italy. She was saying how interesting it was, and I told her I'd explained MCP a year and a half ago. But it is a useful tool for a lot of people. I was bashing it when maybe I shouldn't have, so I'm excited to see the pros and cons with you and showcase some workflows we've built.
I think one difference between these agents and regular ChatGPT threads or other LLMs is that when you're using ChatGPT, you ask a question, it works, and you wait for it to complete. It feels like a single thread. But when I'm talking to my Dots agent, I can ask it to find something out, and it tells me it's going to check. Then I'm not waiting for anything. In the background, it has a computer and browser, and it's working in the cloud. I can continue talking to it, and after three or five minutes it comes back with its research. It's also proactive. It might tell me the kids have a show outside, but it's going to rain Thursday. It feels like a real person because people can tell you things when you're not waiting for them. It doesn't feel like asking a machine questions and waiting for answers. It feels like talking to a person with a personality.
Yeah. I think the personality side of it is what will get a lot of people in the door, more than the free-form aspect. Tell us a little more about Dots and its suggestions. Muse is big on making suggestions and bringing you along. What does Dots do to bring you along with suggestions?
You mean, does it make suggestions?
Yeah. What's cool about Muse is that it makes the world a little more accessible because it sees what's going on in all your connectors and recommends things.
Yeah. It proactively comes back to me and makes suggestions. It knows my calendar, what I'm doing, and what I'm talking about. Sometimes I'll ask a question, get an answer, and then a day later it'll come back with something else, like another basketball camp it found. It thinks in the background, so it feels different from giving orders and getting answers. You have to try it. It's free, and it's incredible. I feel like it's also really good when you're outside, driving or walking. You have a computer and browser available while you're out. I can ask it to do something, check my emails, or find a Chinese restaurant. It's great when you're outside because with ChatGPT you have to give all the context, but Muse or Dots already knows your context and remembers your history, like what food you like and what you do. It can make suggestions for you. That's another pro.
Yeah.
Yeah, I agree. What connectors do you have set up with Dots that were a big game changer for you? Muse seems to have more personal connectors. For example, you can connect iMessage, which isn't available with other agents, and even take action with it.
Email and calendar are big for me because my ChatGPT is constantly editing my calendar and looking at my text messages. iPhone mirroring lets my computer look at my messages, and computer use can access them. Muse and Apple seem to be allies now, so Apple allows access to iMessages, but there's not much of a direct way to do it. My computer has phone mirroring, WhatsApp Desktop, and Slack installed, so ChatGPT can use the computer to access text messages, WhatsApp, or whatever I want it to look at. I have continuous loops that monitor these things. For example, if there's a new schedule for one of my kids' basketball games, ChatGPT puts it on my calendar. What's great is that Miso, my Dots agent, is like a chief of staff with access to my whole ChatGPT setup: memory, skills, plugins, projects, and code projects. It can look at my projects and tasks, discover new things from my messages, and update me. When there's a new basketball game, it goes onto my calendar, and Miso finds it and informs me. It's a continuous loop that helps prevent me from forgetting things.
Oh, yeah.
Yeah.
Yeah, I love that. Could you open up your screen real quick? I want to see what the visuals look like for a couple of things. There's a lot of computer use stuff that you're using, which is really cool. So, this is its own computer, right?
Yes. This is Miso. You can customize the icon and give it a name. It lives here in ChatGPT. I want to show you its computer. It has a Linux computer running in the cloud. To show what it can do, let's say I'm driving and ask it to go to our homepage, jotform.com, and give me feedback. It opened the homepage. I asked it to show me the first piece of feedback: the AI prompt on the left and 'Start for free' on the right. Those are two competing headlines, which I think is great feedback. It's an A/B test anyway, and it's doing pretty well. I can ask for the next one, and I instructed it to show me on the computer too. I can also take over the computer and do things myself. As you can see, Miso is thinking. I'll move my browser to the next section so you can see it.
Okay, let's move the browser to the next section so you can see it.
Yeah, I can already see it.
Mhm.
It's been waiting for a while, so maybe it's stuck. Let's see.
Second video. Okay. 'Powerful forms get it done. Repeat the signup pitch instead of showing proof.' So, 'Powerful forms get it done.' Actually, I don't think it's looking at this page. Can you show me on the browser?
I'm looking at your computer. At your computer.
It probably scrolled down and is talking about something at the bottom. It's been a while. Maybe it opened a different browser or computer. I think it's just showing screenshots here, although I can see the cursor blinking. Let's see what it's going to do.
Okay.
Go to jotform.com and show me the feedback.
Maybe I need to close the computer and open it back.
Okay.
Okay.
Yeah.
As we can see here, it has different tools like Blender, Linux tools, and an image editor. It can use those too. This is a computer living in the cloud.
Okay, now it's opening ChatGPT.com. Let's see what kind of feedback it gives.
Okay.
It's scrolled down and is giving feedback about 'Powerful forms get it done.' It's saying, 'How come it's powerful? You didn't explain why it's powerful.' Okay, next feedback.
I could also speak to it, right? I don't know if I shared it with voice. Should I do that?
Sure.
Yeah, let's try that. I'll reshare with sound. Actually, I had sound.
Okay.
Okay. When you say 'start call,' it looks like...
I've been using it on my phone a lot. It's like a phone. That's so funny, actually.
Hey there. How's it going?
Hey, Miso. How are you doing?
I'm all right. What's up?
I'm getting feedback from you while looking at the computer. On the top part of the Jotform homepage, what other kinds of feedback do you have? How can we improve it?
Okay, let me take a look. Pulling up the homepage now. I'll focus on the top section, the main promise, and what's pulling your attention. In the current hero area, 'Powerful forms get it done' feels a little vague. I'd test a tighter outcome, something like, 'Create a form, collect payments, automate what happens next.' Or maybe let people try a quick example before they sign up, so they experience the value first. Then include one strong proof point right there, like a customer quote about saving time. But wait, I should slow down. We're looking at a different hero with...
He's really going fast, huh? Want one piece of feedback at a time?
So, feedback three: only support 'Create forms with AI' with the real outcomes, like payments, approvals, and integrations. Then we'll hold there until you say next. Want to dig into that one?
Sure. Go to the next one.
Okay. One at a time.
Next, I bring one strong piece of customer evidence right up to that promise, like a short quote with a concrete result.
Sounds good. Miso, thanks.
This is an approach I use a lot, especially while walking or driving. I'll say, 'One thing at a time, and wait for me,' because otherwise it will talk a lot. I prefer that way. It looks good because I gave it an avatar, and I'm getting used to it, becoming friends with my personal agent.
That's the thing about this: these tools are so helpful that you build an anthropomorphic relationship with them. I'm curious, are you making artifacts with this stuff? I do that in Muse. It's similar to artifacts in the Claude environment. Have you been doing anything like that?
I've been building a lot of websites with ChatGPT, mostly internal websites. When I'm planning something, I'll build a website. For example, I'm picking summer camps for my kids for next summer, so I'll ask it to create a website. Last week we were talking about software or websites for one person, and that's what I'm doing. I've built over 10 websites for my own use cases. I'm constantly giving them feedback and asking it to research different parts, add summer camps, and compare prices. I'm continually improving these websites to make decisions easier and avoid forgetting things. When I look at the website again, I remember where I was and can keep going. It's a great way to manage projects.
In terms of artifacts, I open different projects for different websites or things I'm building. Miso is kind of like my chief of staff, and I ask it to go to a thread and do something for me. It helps me with different things and lets me access different projects. I have so many projects that I'm starting to lose track of things. One downside of Miso and these agents is that they're single-threaded. As the thread grows, it's harder to find things. Usually I don't even have to scroll; I can ask what the answer was, and it remembers. It's not a big problem, but the single thread keeps getting longer.
Absolutely. I'm going to show you some of my examples. They're interesting on the personal side because this stuff saves you time. Muse has a bunch of connectors, and unlike other tools, you can connect it to a lot of your Apple ecosystem. Apple Reminders is the first time I've seen something capable of handling Apple Reminders tasks, which is nice. I might use it for day-to-day task management. I have a list of options here: Notion, Outlook, Granola, Todoist, Evernote, 1Password, Tailscale, Google Drive, Google Sheets, and Google Slides. 1Password is surprising because most tools don't have a password app connector. There are a lot of good ones, though it doesn't have as many as other tools. Google Tasks is cool for personal task management too. Muse has some connectors that aren't available elsewhere, but it also lacks some of the more productivity-focused connectors. It does have OpenTable, Spotify, Instagram messages, health apps, and some business tools. Stripe, Slack, Vercel, and Canvas are great, but it's at a different point. I've put together a couple of personal use cases. This is the main chat; I haven't changed the name because I like Muse. I remember making fun of my girlfriend for being interested in it because it was cute and plushy. I said they got all the girls by making it cute and plushy. But I have some scheduled tasks going on. I'm a big triage guy and don't like having anything that isn't working toward inbox zero. I have a couple of side chats that get referenced in the main chat, which I love. I have one that's...
Actually, what I'd love to do is make an artifact for this. I have a couple of things to help me organize my stuff. One is an email clear-out. I haven't been checking my business email as much because I just don't want to be in there. I want to be in front of the terminal. So, I've proposed this thing called an hourly email triage batch. It goes through my Gmail and clears out as much as doesn't matter. It has an artifact next to it, the email triage dashboard. It's done about 27 emails in the last few minutes.
With email, I do the same thing. I'll ask it to tell me the most important emails once a day.
Then I don't have to check all my emails unless there's something important.
Yeah. As you can see, there are some pitches and other things that might matter. There might be a newsletter I care about, but a lot of this I don't need to read. I keep marking them as done. I like to have these things running. Right now, I'm going to have it consistently mark my inbox as read. That's a great starting point for me. It will have to go through all 200 at once first, then do it hourly. That's fine; I'd rather do that. I have a couple of other fun ones. With iMessage, I leave things unread to remind myself to check them. Every once in a while, there's a muted group chat I need to realize I don't need to be in anymore. I opened it and saw 84 unread messages. That doesn't make sense because there are no notification symbols; I only have one or two texts from my girlfriend or my mom. If my mom reaches out and I want to remember to call her later, I don't respond until later that day. How do I have 84? Muse was able to track this over time and show the backlog by age. I have no unread threads from the last seven days, so the rest are old muted stuff taking up space on my phone. Some friends have a bunch of unread texts, so I can ask it to clear old groups I don't need. I want it to be proactive and send reminders when it thinks they matter, and ask me before archiving stale conversations in the future.
With Muse, I feel like I can be very conversational, more so than with other tools. That's what I like about these tools: they let you be more personal. The proactiveness is really cool. As you can see in the sidebar, there's a scheduled message triage daily digest. I have one scheduled daily at 9:23 and another later in the evening. It's helpful to have it consistently clear my texts and emails. I also have Google Calendar tasks. I asked it to schedule a chanting lesson for Wednesday, and it did. Another great thing is that I can use voice to ask it to schedule a nightly 'What did you do today?' check-in to clean up my Google Calendar. I'm a perfectionist about making it look nice, so I'd like it to ask what I did and during which hours, in an easy conversation, so it can fix my calendar if it's off. What time did I wake up? Go to bed? When did I work? Did I go out to eat or work out? It can give me an end-of-day digest so I can tell it what I did and it can fix my calendar. That's huge for me because I'm so particular about it, but when I get busy, I don't want to drag things around. It has access to all my calendars, so it gets me. That's something I really like about Muse.
Oh, look at Muse. It's actually working on the...
Yeah, I like that it gives me a thumbs-up. I'm not going to lie; it makes me smile.
It has headphones while it's working.
Yeah, that's what Isabella, my girlfriend, really liked. She was like, 'Look how cute he is. He's working.' I don't know how he's typing with those hands.
Little hands. Yeah.
Okay.
I really like this next one. It's going to be fun with my girlfriend too. OpenTable is a really good platform for getting ideas for places to eat.
We'll talk to her, and then you'll...
Oh my God, they can go on dates.
Yeah, they can go on dates.
No. Tell your Muse to talk to my Muse.
Yeah, tell your Muse to reserve my girlfriend's Muse. As you can see, there are some great spots in Chicago I hadn't considered, and there are options to reserve through OpenTable. I could tell it to make reservations too. It's funny because my girlfriend is here. I told her about it, and she said, 'Oh my gosh, it's so cool.' Could you recommend the middle-of-the-pack Asian place, Milipac Laauo Sewan, to Isabella in my iMessage and say I can book an OpenTable through Muse? Would you like that?
Nice.
For 7 p.m. tomorrow? You might be saying this is excessive, but if you're like us and love living in the terminal, this is great.
This is great because I like talking to my girlfriend in person. That's what I prefer. This makes her life easier. What's another fun one I'm doing? I'm trying to cut unused subscriptions.
Yeah, I'd rather have it make an artifact, not a PDF. I have a lot of things where I'm not sure I need some of these domains. I'm fine with ChatGPT and a basic Gemini; I don't need Gemini Ultra right now. I feel like I could save a couple hundred bucks a month on things I don't need. I want to cut those things. I can connect it to Plaid. I'm not going to get too deep into finances, but it's a really cool tool if you're interested. You can add more than one account and connect multiple accounts. For my business and personal finances, I can have it analyze the data. It's view-only access, but it's a secure platform from what I understand. It can help build these things. As you can see, I have all my artifacts in a library, and the three most recent are here. It's making the cut list.
Let's get back to this chat with the dinner spots.
Oh my God.
How about something middle of the pack?
Oh my gosh. Let's see what she says. Did she respond?
Oh my gosh.
Rather than opening up my text on stream, live on YouTube.
Live on YouTube. YouTube.jpeg.
Let's see. I'm booking a date right now. If you don't think I have...
She did. Did you draft this message with Muse? I'm demoing for the podcast.
Yeah.
It understood it, I guess, because of the dashes. I see the dashes.
Yeah.
Do you see those dashes? I don't think you write those dashes yourself, right?
What are the M-dashes?
No, of course not.
M-dashes, yeah. That's so funny, though. It's weird that I can text through AI because that's never been a thing for me. That's really nice while driving, right? You can text while driving.
Yeah, because the iMessage situation is awful. CarPlay takes so long; it's very bad. But let's take a look at the updated email triage list. You're supposed to keep the artifact updated the whole time. Please keep it updated with any progress, in or outside the loop, because I want to see what it is. It just said it cleared 174 messages, which is crazy.
Let's see if she responded. This is going to be hilarious.
Um...
Her Muse will get back to your Muse's date.
Let's do it. It's going to update the artifact, I'm sure. As you can see, when it's spinning like that, it's updating. It's got a good feel to it.
It's got a good feel to it.
I'm sure she's working. Anyways, I'm going to recommend these to her every week and ask which date spots she thinks would be fun. Then it's like, 'I did the fun thing with a person I care about.' The task is still going, so it'll probably take a second to update this. That's probably my last example. What do you think about Muse?
I think it looks great. The only thing for me is that I'm so used to ChatGPT now. It's something separate, right? All my plugins, skills, and everything live there, so I think I'm going to stick with Dots. But I think it looks great. Did you find any cons?
Yeah. The cons are that it's not as business-focused, and there aren't that many use cases yet. Computer use is still just okay because it's new. ChatGPT's computer use is exceptional compared to other tools, but I don't feel like Muse is on par with ChatGPT yet. Claude's computer use is obviously nonexistent and bad. Comparing these two, I'm excited to see how they compete over time. When I say three, I mean Claude and Duet too. I'm fascinated by something Zuckerberg said: 'If we don't place in the top two in the AI race, our company is finished.' He thinks if they don't get top two, they'll lose and won't be a company anymore. Have you heard him talking like this?
I haven't heard that, but I think it makes sense because usually the top two have most of the market in everything. It's like Coke and Pepsi, Visa and Mastercard. The two are competing and constantly improving, and people who don't like one go to the other. Usually the top two get the majority, like 90% of the share, and the others get small pieces. But I wouldn't count Grok out. Grokbot was earlier, I think. Many people say Muse and Dots are kind of clones of Grokbot. It came before, and it all started with OpenClaw. It wasn't easy to use, though. They took about six months to create great OpenClaw clones, and they're much better than OpenClaw. It was hard to use and didn't have all these connections. My guess is people aren't going to pay twice for AI. I'm paying for multiple AI tools because I want to try them all for my business, but if I were using them for myself, I'd stick to one. I'd pick ChatGPT, Claude, or Grok. Muse is also a contender, but I doubt it's going to beat OpenAI or Anthropic. They're focused on this, while Muse is a secondary product for Zuckerberg. If they lose, it doesn't matter. They had an open-source LLM, Llama, which was pretty good, and we used it for a while, but they abandoned it. My guess is that it's a secondary product that isn't that important. For OpenAI or Anthropic, this is life or death. Zuckerberg already has a huge, successful company, so if Muse isn't successful, it doesn't matter. That's why I wouldn't bet on Muse. I think it's good, but I'll stick with Miso.
Yep. Well, guys, with that being said, let us know your thoughts and what you want us to cover next. I really appreciate you making the time. Make sure to try out these cool AI assistants and let us know which one you like best. Thanks for watching, and we'll see you in the next one. Peace. Bye, guys.
Stay Ahead with the AI Agents Podcast
Get the latest insights on AI agents, their future, and developments in the AI form industry.

