I Let My AI Agent Run for 504 Hours Straight — Here's What Happened

[00:02] back, but I've been really sick. So, yeah, basically [snorts] just back on my feet now. But, the good thing is that AI agent has been running autonomously on my Mac Mini even though I've been like bedridden. So, that's good at least. So,

[00:16] today I just wanted to take a look at our 3-week or 504 hours running review. So, we're going to go to check this has been running on X as you probably know how many followers we gained over those 500 hours, how many

[00:31] interactions, or basically just a look at the X analytics, okay? YouTube channel the agent is running, how many views, how many subs,

[00:43] and the agent is also running this Skills MD store. It's basically like a hybrid, but it does make promotional videos. It does promote it on X. So, we're going to see how much revenue we made from that so

[00:56] I'm going to talk a bit about why I am running this experiment and what my plans are going forward. So, yeah, I think we just got to get into it. So, let's log on to the Mac Mini and check out the X stats and stuff like that.

[01:11] Okay, so if you haven't seen my setup before, it's pretty straightforward. We are running this on Cloud Code, right? With a P flag. And basically we have like a loop here that runs that is connected to my WhatsApp. It's like a

[01:24] custom solution, almost more like a nano claw if you check that out. It's less intricate than open claw, but for me this is working. It's basically what I want, okay? We have a set of skills that the agent

[01:38] can use anytime it needs. So, if you go to skills, you can see this is the list. React videos for for YouTube, we have X, coding, promotional videos, video research. Yeah, I have a bunch of skills I kind of

[01:52] trained up. You can go back and watch my other videos if you want to. Gmail, GitHub, LinkedIn, LLM Bench. We have a bunch of stuff here, Polymarket. these. But let's head over to X and look at the

[02:07] analytics for the first 3 weeks of this see the analytics. I guess this is something like 3 weeks or something like. You can see it's not like insane, but we have like the impressions are

[02:21] around like, I don't know. I'm not quite sure, but we have 980,000 impressions. 852 followers. I don't know what the Yeah, you can see basically these are stats. 4,000 likes.

[02:38] 15,000 profile visits. 28,000 engagements. Shares, engagement rate. Yeah, I don't I'm not quite sure. But these are the stats and you can see we are kind of gaining like 40 followers a day or

[02:53] something with this setup. So, it is pretty interesting. We get a lot of unfollows too, but like I said, this is just an experiment. I wanted to run for just an experiment. I wanted to run for at least 4 weeks. So, we are like 75%

[03:05] So, it's been super interesting and this has been Yeah, one day we had 128 impressions. I might look at that in the last video on this agent or maybe the last video, we'll see. But you can see it's

[03:20] pretty interesting stats, I would say. Nothing like remarkable, but like a Nothing like remarkable, but like a steady It's quite even here, right? So, let's head over to YouTube and check out the stats for agent over there.

[03:33] Okay, so this is the YouTube stats. So, you can see we have gained around 322 subs. We got 28,000 views. I haven't been posting for a few days with something. But I I have fixed it now, so I'm going

[03:48] to start doing the post again. But again, this has been also pretty successful. We have 700 watch time hours. Uh and it's a pretty automatic setup here, too. So, this has been going

[04:02] trying to do this for the rest of the experiment to see if we can end up at least maybe like 5 600 subs or something like that. Doesn't really matter too much. But again, pretty strong performance overall. And it's also been

[04:17] doing like this um it's posting on X when it does a video, something like that. So, this is also something you can set up if you want to. Uh so, that is YouTube. And the last thing we wanted to check out was kind of

[04:30] the skills and D store. This is the agent is running this store here. So, this is some uh it's having responsibility for this page here. So, it's going to do promotion where you can buy the setup for the agent, right? So,

[04:44] it's kind of promoting itself. So, you can do like Chrome automation CLI for want to. I'm going to leave a link in the description to this. So, this is what the agent has been promoting. So, I'm going to head over to

[04:57] Stripe. Uh let's check out some of the stats here. Okay, so if I have a look at Stripe here, you can see we have made 41 sales on the page. like $2.99, but we added like a new

[05:11] feature that's $4.99. So, we sold some of them, too. So, if you add that up, you can see $141 so far. And if you think about the cost, it's like 100 bucks a month, ish, a bit more, like 120, ish. And the electricity

[05:27] is very low on the Mac Mini. So, I would say we are in profit so far on the agent, and we even have a week to go. So, I say we're like $20 in profit, uh including. So, it means we have run this for free now, basically. So, that is

[05:40] pretty cool, and I'm super happy how it has performed on the store so far, right? Skills and D store, if you want to check it out. And we haven't really had any complaints. I don't know if you refund because people paid double and

[05:52] done and it keeps creating these inter-interesting promotional videos using their emotion skill. So, I want to go back to the agent again and I want to talk a bit of more what I plan to do in the future and actually why I'm doing

[06:06] this experiment. Okay, so I wanted to talk a bit about why I actually did this experiment and what my plan's going forward is. So, while we wait for that, start on Hacker News and look around and we have a rabbit hole skill that kind of

[06:20] dives into different links and stuff. So, let's just run this while I talk a bit about it. So, the reason I started this is actually because I wanted to see what could happen in the future if these AI agents kind of get their own

[06:34] personality, right? And they can kind of do whatever they want on the web. Can we actually make them do something useful? And I wanted to see if I run them for 4 uh what would actually happen? Will it change anything? It doesn't really have

[06:49] like a great memory system or anything like that now. It basically do all the cron jobs it's set up to do and you can see now it's surfing the web, opening up a bunch of different links like reading all that, getting the

[07:01] contacts. It's using X, gathering contacts. Uh but I I have a plan. Uh I useful. I I wouldn't say it has been useful because we you been using it to market it our Skills and Deeds store and like I said, we've been making a few

[07:16] bucks on this. So, it hasn't been like a negative experience. And the followers course interesting, but it's nothing like uh revolutionary or anything. Uh

[07:29] but I want to say privately, like off this agent on my own, I have created like a similar system that is like applying for jobs and con- contacting don't I don't know if I'm going to do it sure if I'm going to do a video on it.

[07:43] But it has landed me like a couple of interviews. So that is pretty crazy. I might share something on it. But I don't know if I want to share the exact thing because it's just an experiment, right? So that was a little bit about why I

[07:57] actually did this experiment. And I think it's I think I'm just going to keep continue and see where we can take this. So what I'm trying to do now is just explore what else we can do with this. What is it looking at now? Mercury

[08:11] 2 fast reasoning LLM powered. So now it was just browsing the web, right? So that's pretty funny. You can It's also pretty funny to watch this. I was also thinking about could we set up some kind of stream? So actually I think we can

[08:24] is also something I've been thinking about. And like I did with the Polymarket video, I might be looking a bit more into that. So did it stop here now? Had some API error here. I don't know

[08:37] Let's do continue. Yeah, I don't know what happened here. Polymarket or something like that in that style. So I'm just going to keep thinking about it. And I'm going to come back with a

[08:51] video later this week. Maybe we're going to try to train up a new skill or something like that. Or I might actually show off how I'm using know a lot of people are interested in that.

[09:03] So okay, now we are on Moonshine AI GitHub. We're looking at their code repo. So you can see it's pretty fun just to watch it. So that's why I was thinking

[09:16] about some stream or stuff like that. But yeah, I'm going to come back with some interesting things. And if you have an interesting idea, just let me know in the comments. And like I said, if you want to check out the Skills ND star and

[09:28] you want to set this up for yourself, you find that link in the description. So yeah, that was all I had for today. A pretty interesting review for the 400 500 hours or something, 3 weeks. So, yeah. Thank you for tuning in and I'll

[09:42] yeah. Thank you for tuning in and I'll see you again in just a few days.