39:42Nick Saraev Daily Updates · Talking Head
Un-pausing My Daily Updates
A $4M/yr founder returns after a year away, reveals his current revenue numbers, and lays out a four-lever plan to hit $500K/month.
June 17thThe argument in one line.
A Claude Code agent running on a /loops schedule can autonomously refactor low-code n8n automations into native serverless code over 48-72 hours, achieving cost reductions that would take a human engineer weeks to replicate manually.
Read if. Skip if.
READ IF YOU ARE…
- You run an automation agency or build n8n/Make workflows for clients and are seeing token or API costs compound at scale.
- You use Claude Code for agent work and want a concrete pattern for autonomous optimization without babysitting a terminal.
- You sell AI services to mid-market or enterprise clients and need a framework for discovery calls that surfaces revenue-critical processes first.
- You want a mobile-first Claude Code workflow that lets agents run in the background while you are away from your desk.
SKIP IF…
- You are looking for a tutorial — the /loops concept gets only 90 seconds of airtime; the rest is Q&A commentary.
- You have no existing automations to optimize; the loop pattern assumes production workflows with real token burn.
The full version, fast.
Nick Saraev uses a daily update format to answer YouTube community comments and report on his agency and products in public. The headline concept arrives at 10:22: spin up a Claude Code loop with a goal of reducing automation cost by 10%, point it at your n8n account, and let it run for days — logging hypotheses to a shared file, testing, refactoring modules into code blocks, then migrating those blocks to Modal or AWS Lambda. Upfront token cost is real, but for high-volume business automations the savings reach thousands of dollars per week. The video also covers a perplexity-scored humanization prompt for AI content and a five-thread named-agent mobile workflow.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:50
01 · Channel format intro
Format overview: YouTube community Q&A plus build-in-public plus growth updates. Goal framing: $4M now, $6M target, 500k/mo, 1M subscribers.
00:51 – 03:13
02 · Discovery call framework
How to identify which client processes to automate: ask for the full customer journey from first touch to close. Focus surfaces revenue-critical steps over admin noise.
03:14 – 04:05
03 · Lead gen scrapeability filter
Only do outbound for verticals where you can scrape a clean ICP list. B2C and most e-commerce fail this test. B2B (accounting, agencies) pass.
04:06 – 06:04
04 · Cold email and misc Q&A
XP Pen plus Presentify for screen annotation. Cold email is not dying. Mindset note: look for reasons things work, not reasons they do not.
06:04 – 09:12
05 · LeftClick agency stack and offers
Current offers: Claude Code onboarding, CoWork deployment, enterprise AI/ERP consulting. Team: Cody McLean (sales) and Gio (technical lead). Spinning up cold traffic offer via Instantly Airmail.
09:12 – 10:22
06 · Humanization prompt
Statistical model built on personal writing corpus. Generates sentence fragments, scores by perplexity, selects lowest deviation. Artifacts drop from every paragraph to once per 5 pages.
10:22 – 11:58
07 · The /loops automation optimizer
Core concept: Claude Code agent loops every 30 min with goal of cutting automation cost 10%. Scans n8n, logs to shared file, refactors modules into code, deploys to serverless. Saves thousands per week for high-volume clients.
12:28 – 13:33
08 · Guardian Angels paper and AI personalization
Reference to Gorn's Guardian Angels paper: build a shell around the human that amplifies rather than replaces. Distinguish AI as economic replacement versus AI as human amplifier.
13:34 – 14:33
09 · Live streaming debate
Hormozi bullish on live content. Nick constraint: daily updates are hyper-efficient (record in 29-min gaps between meetings). Scheduled live streams break time flexibility.
14:34 – 17:14
10 · Moving update (personal)
Moving apartments in Calgary. New place: mid-market building with a commercial gym 30 seconds from the unit. Eight Sleep mattress topper.
17:15 – 18:56
11 · Maker School and YouTube metrics
Maker School: 2,085 members (down 1 in 24h). ARPU rising as low-tier members churn. YouTube: 452k subs. GLM 5.2 video at 31k views after 2 days. Instagram restarting.
18:57 – 21:03
12 · Mobile Claude Code workflow
5 named Claude Code threads via /remote control. Caffeinate laptop, close lid. Agents run all day — list cleaning, humanization, personal tasks — while host is in meetings.
Lines worth screenshotting.
- A Claude Code /loops agent pointed at an n8n account will autonomously minimize module count, replace them with code, and migrate that code to serverless — without human intervention beyond the initial prompt.
- The shared-file pattern is the key: every loop iteration reads prior learnings before writing new ones, so the agent compounds knowledge rather than repeating failed experiments.
- Serverless cold-start providers like Modal cut recurring automation costs by roughly 10x compared to always-on n8n nodes polling on a schedule.
- Scoring AI output by perplexity against a personal writing corpus produces content that reads as native voice, with artifacts appearing once per five pages instead of every paragraph.
- The scrapeability test is the fastest lead-gen filter: if you cannot scrape a clean list of your target ICP, do not build outbound for that vertical.
- Discovery calls that ask for the full customer journey from first contact to close surface revenue-forcing processes automatically; asking what processes exist surfaces administrative noise.
- Running five named Claude Code threads via slash remote-control with a caffeinated laptop lets agents execute economically valuable tasks while you are in unrelated meetings.
- ARPU can rise even as subscriber count falls if churn concentrates on low-ticket legacy plans — a metric worth tracking separately from total member count.
- Cold email volume is up but cold email quality is not — the inbox still opens first thing every morning, and high-quality emails face roughly the same competition as they always have.
- An AI agent that amplifies a human by wrapping all input in a personalization shell is architecturally different from one that replaces human economic output.
Three patterns from a builder who runs agents all day.
WHAT TO LEARN
The most useful ideas in this video are not the headline — they are the three workflows Nick uses daily to get AI to do economically valuable work while he is doing something else.
- Ask for the customer journey, not the process list. Clients describe processes through the lens of what feels important to them — usually administration. Asking for the journey from first contact to close surfaces what actually moves revenue and filters out everything else.
- Perplexity scoring is a practical way to preserve voice in AI-assisted writing. Generate multiple sentence continuations, score each by statistical deviation from your own writing corpus, and select the lowest — artifacts become rare rather than constant.
- The /loops pattern turns an optimization goal into an autonomous multi-day agent. Give Claude Code a measurable target, a shared log file for inter-iteration memory, and a set of tools. It will compound improvements without supervision.
- Named persistent agent threads accumulate knowledge over time via auto-compact. Five threads covering different domains of your work — accessible from a phone while the laptop runs in the background — extend your productive hours without adding headcount.
- The scrapeability test is the fastest filter for outbound lead gen viability. If you cannot programmatically build a list of your ICP, outbound is the wrong channel for that vertical — a data problem, not a messaging problem.
Terms worth knowing.
- /loops
- A Claude Code slash command that runs a prompt repeatedly on a timer, allowing an agent to test hypotheses, log results to a shared file, and iterate toward a defined goal without human supervision.
- Perplexity scoring
- A statistical measure of how surprising a piece of text is relative to a reference corpus. Low perplexity means the text closely matches the author's own statistical writing patterns — used to select AI sentence continuations that sound native.
- Serverless / cold-start provider
- Infrastructure (Modal, AWS Lambda, Trigger.dev) that runs code only when triggered, billing per execution rather than for idle uptime. Replaces always-on workflow nodes with on-demand functions, reducing recurring cost significantly.
- n8n
- An open-source workflow automation platform where logic is built as connected visual nodes rather than code. High node count and polling frequency drive up execution costs at scale.
- ARPU
- Average Revenue Per User. A SaaS metric tracking how much each subscriber pays on average. Can rise even when total user count falls, if higher-tier subscribers churn less than lower-tier ones.
Things they pointed at.
04:14productXP Pen drawing tablet
04:16toolPresentify (Mac annotation tool)
12:28bookGuardian Angels: LLM Personalization for Productivity and Security (Gorn)
12:28bookThe Coexistence by Ethan Mollick
Lines you could clip.
05:32
“Rather than looking for reasons why something can't work, you should look for reasons why something will work.”
11:11
“You'll run it on an automation like n8n. It'll then refactor the components to minimize the number of modules, make more of it code. And then before you know it, you basically just have a code block.”
11:50
“Before you know it, your automations are running a million times faster and for a million times less money.”
05:07
“It's just an email, bro.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
00:00Welcome to my daily updates channel where I'm showing you guys I'm running a $4,000,000 a year business with a goal of hitting $6,000,000 a year or 500 k a month. I also wanna make a million subscribers currently at $4.50 k, a little bit over that now, but anyway, um, that's obviously quite the gap. So what do I do in this series?
00:15Well, I answer you guys questions with some YouTube comment q and a. And then I actually build and strategize in public to run you guys through everything that I'm currently doing publicly and transparently in order to get to that goal. Finally, I ran it out with some growth across YouTube, Instagram, and my products.
00:28And speaking of my YouTube, uh, this is my main channel if you if you haven't seen it yet, and this is the daily updates channel. So what I do is I go to the community tab, sort of like the comments, and then I scale by newest, zoom out a bit, and then I scroll all the way down to, I think, like, or three weeks ago, then I just answer all your questions.
00:43I'll do that for like twenty minutes or so. That's how we do some value. And then, uh, finally, I'll show you guys my actual business.
00:49Alright. So, uh, the one for three weeks ago, we're not gonna answer.
00:52We're gonna answer this one here from Stoyan, who I think is Bulgarian Zorave, Glad to see you back. A few questions.
00:58You did so much traveling. Are you planning on visiting Bulgaria anytime soon? There we go.
01:03Uh, not probably in the next, like, six to twelve months, unfortunately, but after that, for sure. No problem whatsoever. Assuming you're in Sofia, which is their capital, I plan on going there, uh, to get some passport stuff done so I can finally become a, god forsaken EU citizen, uh, in reality.
01:17So we'll sort that out then. Could you give me an insight or a framework that you do when you map processes to automate or optimize for a client on a discovery call? I feel like this is hard if you don't have any information about the processes beforehand.
01:29Sure. I'd say the core way that I determine which processes are useful is well, first of all, you need to ask the client to run you through what their processes are.
01:40And the simplest and easiest way I've found to do that while also maintaining a focus on the shit that actually matters, not just the stuff that they think matters, which usually doesn't actually matter, is, um, could you outline the entire customer journey for me from the very first point a customer comes into contact with your brand all the way to the point at which you stop engaging with them.
01:59So literally, very second, the very first moment, the first impression that somebody has with your business, is it through, like, an ad? Is it through, like, an email chain, like, a referral?
02:08Is it through, like, a cold email? Is it through, like, a DM? Is it through, like, a social media post?
02:13Like, what is the very first place somebody comes in contact with you? And then walk me through every step from then on until the close, uh, of the engagement itself. And typically, when you do this, you get to focus on the show that actually matters, which tends to be, you know, like, what what am I sending the client?
02:28How am I managing the relationship with the client? Like, this is stuff that actually, um, forces the revenue.
02:33Right? And if you are a business owner, lot of the time, you're gonna be more concerned with, the administration and the the logistical stuff and, oh, the hiring and all that. But, um, when you just focus on, like, hey, what is the customer themselves actually saying?
02:46What's the person that's paying you actually experiencing with your with your business? You can usually just, start there and then get the $80.20. And then later on, if, you know, you have to do the admin automations, logistics automations, you could do that afterwards.
02:58Um, so that's number one. And then after that, you just look for automations that, um, help with reach, help with the actual acquisition, help with the conversion or the closing, or help with, uh, ability to reactivate, like, long term.
03:09So as long as you're doing that, that's like your structured sort of framework or process, you'll go pretty far. How do you do lead gen for different verticals? Uh, says Diane.
03:17For example, how do you do lead gen systems for ecom accounting marketing agencies when the target audience is not easily scrapeable, for example? Well, if the audience is not easily scrapeable, I will usually not do outbound lead gen because, obviously, having the lead is the is the core thing that is, um, that's sort of the core requirement.
03:32Right? So I would only work with an audience if I could very easily scrape. So I mean, like, for instance, uh, you know, ecom, this can be pretty tough because a lot of the time, we're selling b to c or direct to consumer.
03:42And it's like, how the hell am I gonna scrape, you know, consumer details? I can't. So I don't tend I don't tend to work with ecom.
03:47But accounting, you can work with. Why? Because accounting b to b, they work with all sorts of businesses.
03:52If they work with agencies, I scrape a list of agencies. If they work with insurance companies, scrape a list of insurance companies. Marketing agencies, same thing.
03:58Marketing agencies work with all sorts of businesses. So hopefully that makes sense. Okay?
04:02And, uh, yeah. No. Good call on the resources maker school stand.
04:06Very much appreciated. David says, don't glaze enteropic too much. Yeah.
04:09I won't. Don't worry about it, but thank you very much, man. Good seeing you in your cute ass cat.
04:13Reflex says, hey, Nick. I was wondering what hardware and software you use to draw directly on your screen. Yeah.
04:17I use this one called an XP Pen, which is a simple $40.50 dollar product you can buy online, and then I use, um, a Mac tool called Presentify. That's how I do this. Nick is back.
04:27Let's go. Thank you. Soldier boy to your voice.
04:30Thank you.
04:32Nick, do you think cold email, says Saif, do you think cold email is becoming more of a brand and trust problem than a messaging problem? It seems like prospects today often decide whether to trust the sender before they even read the copy. In that context, can someone with no audience, no brand presence, no social proof, and no established reputation still build a successful cold email acquisition channel or has some form of credibility almost become a prerequisite?
04:54I find it to be difficult to believe cold email is still a channel to risk capital on tools, to test an offer which can generate zero replies because of how inboxes are flooding with cold emails nowadays. Basically, everything you've said here is, uh, nonsense, unfortunately.
05:08I I wish I didn't have to come down so hard on you, but this is you just you're way in your own head here. Uh, it's just an email, bro. Uh, of course, there's always been competition in the mailbox.
05:17People have been sending cold emails for a very long time. Cold email is not new. There's certainly more cold email that is of low quality these days, I would say.
05:24But in in reality, the the amount of high quality email landing in somebody's inbox pretty much hasn't changed, and we all still open up our email inbox the very first thing in the morning. Rather than looking for reasons why something can't work, you should look for reasons why something will work. Okay?
05:39I think if you approach this from a positive standpoint as opposed to a negative, you'll find a lot more success not just in cold email, but also in life in general. Okay. Wow.
05:45I just realized I wasn't recording for the last few minutes, which absolutely blows. So I just answered a bunch of questions up here and I'll stitch this in post.
05:54But, anyway, where was I? I think I was here or maybe here.
05:59No. I I think I answered this one. Uh, fat fingered the key, which blows.
06:03Anyway, let's just re answer all those questions. Uh, Axis Gaming says, hey, Nick.
06:07I'm curious about what you're doing with LeftClick. What are the offers that are selling right now? I'm guessing a lot of the acquisition comes through a personal brand, so you don't need a cold traffic friendly offer.
06:14Even with the meat processing company, Metro, what kinds of work are you doing for them? Is it to do with Is it GTM? What does the split look like with the clientele?
06:20Is it primarily a lot of online businesses like agencies and so on and so forth? Basically, just wants to run through of, like, my current stack. Yeah.
06:25I mean, the offers that are working right now are, like, training, upskilling, workshops, um, getting people up and running with Claude CoWork specifically, like deploying and then, um, like, showing everybody how in in a business to effectively manage, like, shared skills within Claude CoWork. Um, that's a lot of, uh, that's a lot of the work that we're doing right now simply because, you know, I think Claude code is obviously very interesting and exciting to a lot of people.
06:46But it's a little bit out of their purview, so they can't necessarily do that. Uh, for the meat processing company in particular, we were building an enterprise, uh, resource planning, like an e an ERP, which I was just helping them with inventory because they are like a major distributor. And, uh, you know, also like a little AI brain type thing that can, um, I think, like, progressively manage more and more of the inventory system over time.
07:06You are correct that, uh, the vast majority of my traffic these days or the vast majority of the people that I'm working with are not from cold traffic anymore, and I don't actually have a set of fixed offers. I would like to because I wanna scale fClick to over 250 a month.
07:18Like, that that would be my goal. Um, but right now, we're working with a a few really big names. And, honestly, I'm I'm contributing most of my time and effort towards them, like, uh, mister Beast, for instance.
07:29So, yeah, a lot a lot more consultation on my end. You should also note that, uh, I have a couple of team members now. It's not just me.
07:34I have, um, Cody McLean in Calgary here who, uh, you know, he's built websites and and systems for the government of California, for, uh, you know, Lululemon. Like, he has a bunch of sick case studies under his belt.
07:47He does a lot of, like, the consultation and at the sales end now as well, which is quite useful. And then also my friend and business partner in Clairvaux, Gio, who does a lot of, like, the building and, you know, he leads the technical team at Clairvaux.
07:59So he's, like, sort of in charge of the the dev there, which means when I bring him in to do projects and stuff like that for clients, he's usually very, very good at that, which is nice. So that's that's more or less what the split looks like.
08:10It's mostly, like, larger businesses, and it would be nice if I could just sell, like, a simple small productized package to small to midsize businesses right now. Um, time wise, it's just not the best split for efficiency and leverage reasons, but I wanna get back to it.
08:25Um, so we're actually spinning up a cold traffic offer sometime in the next few days, which is gonna be fun. I'm doing it off the back of Instantly's Airmail. They recently launched new mailboxes, I just wanna test them.
08:35So I was like, why not do it with left click? So we'll see how that goes. But, yeah, most of my stuff is consultative now.
08:40People really just come to me, which is nice. K. And then Ahmed said Bangladesh mentioned absolutely Aaron, who's one of the most consistent people in Maker School, so let's go.
08:47Kevin said, good to see daily updates back again. Thank you. Zorps says, Nick, what are all the automations or just simple use cases use Cloud for at your agency?
08:54I love an in-depth video showcasing from front to back how you use AI internally. Tabbed again and showed a bunch of API keys. I really wanna alt tab right now, but I can't because there's tons of API keys on my screen.
09:04So kinda screwed myself up. Uh, if I alt tab, I'll show you guys my school API key and all that stuff.
09:10So let's not do that. But, yeah.
09:13I mean, like, a couple a couple big ones. Like, in in my agency, one big thing we're using now is, a humanization prompt. So basically, I created a statistical model on my tone of voice across all of my writing on the Internet.
09:24And then I use that model to basically iteratively produce sentence fragments for content that I create. And then for every sentence fragment, I score the perplexity, which is the statistical deviation from like some core sample, which is my own writing. And then I just have it generate and then select the lowest perplexity ways to continue with every sentence.
09:42And the value there is I end up with content that is actually my own tone of voice. So you can generate with AI lots of content, but doesn't really do much in the content marketplace because people look at it and be like, wait a second. That's AI content.
09:52This guy's using AI for content. Fuck them. Um, what you can do though now is you can actually, like, use AI to generate stuff, and then you could, like, reformat that to your own tone of voice, and you could do it fairly well.
10:03Yeah. There are a couple of weird artifacts and stuff, but instead of appearing in every sentence or every paragraph, it'll appear like once in every, like, five pages, uh, which is quite nice. So that obviously significantly improves human q and a time because I don't need to spend god knows how long, like, doing all that stuff myself.
10:18And, uh, yeah, that's pretty big. Aside from that, yeah, I'm just thinking about the advanced SaaS stuff that I haven't actually talked about before, like loops to improve the efficiency of preexisting automations.
10:32So you could run like a Claude code loop, and you could say, want you to scan my NA done, you know, account. I want you to scan my Claude routines.
10:40I want you to scan everything. And then, um, I want you to make it 10% more time efficient or cost efficient or whatever.
10:46And, um, that's your goal. And just loop over every 30 minutes until you've done this by logging your results to a shared file with all of your learnings. Future instances should pull from that shared file.
10:54Where you're doing those, you're basically doing, like, auto research where, um, your your system is, like, making hypotheses, testing them, and then seeing the results.
11:03And then if you get to the 10%, then it'll stop. 10% is one way to do it, but typically what this ends up being, if you just run this for like a long time, like a couple days, is you'll run it on an automation like n a n.
11:14It'll then like reject the components and stuff to minimize the number of modules, make more and more and more of it code. And then before you know it, um, you basically just have like a code block. And then it'll take that code block, and then if you wanna make it faster and more efficient and less costly, it'll actually put that on like some inference provider.
11:29Sorry. It'll provide like some serverless provider, something with cold starts like modal or trigger or whatever. And then that'll get even more efficient.
11:37You'll spend one tenth of the cost. And then it'll take it from that and it'll like go find some AWS thing it'll and put it on that. And before you know it, your automations are running, like, a million times faster and, like, for a million times less money.
11:48So that's been pretty cool. That's, like, kind of like a new sauce idea. Obviously, spend tokens upfront because you have to pay for the AI intelligence to automate to optimize the process.
11:59But, um, after you've looped it over, if these are, like, real business automations, like, a few of our clients, you can save a lot of money. You know, like, you gotta understand some of these, uh, business automations, specifically for the bigger guys. Like, these are burning a lot of tokens.
12:10Like, we're actually spending a lot in both token cost, inference cost, automations because they're running so often. Some of them are pulling.
12:16There's no way to really fix that that we have found. So when you put all these on some sort of, like, you know, serverless stack or, you know, AWS or something, you could save many thousands of dollars every week. Um, then Ahmed asked about the coexistence, the next book of Ethan Moloch.
12:31So, yeah, I actually did find this. And, um, I was looking at this and talking about it. And, you know, to make a long story short, um, I don't really know anything about this unfortunately, but it looks pretty cool.
12:40I mean, like, damn, what a nice, you know, cover page. Um, his main point is that it's about building a durable human advantage in a world of intelligent machines. Uh, and to that, I said, you should check out Guardian Angels LLM personalization for productivity and security by Gorn.
12:54Gorn is, um, just a very intelligent guy that is just heads down thinking about how to align AI models to human preferences and human needs. And the whole idea is instead of working on these massive galaxy brain intelligences that just do economic work without humans, we could focus on finding ways to amplify a human being rather than replace them.
13:16The way that you amplify a human being is you basically build this shell around you that, um, takes all input and then uses, like, machine intelligence to proactively solve problems for you. So part of this was actually the inspiration for that humanization prompt that I was talking about earlier. So it's pretty cool.
13:32Definitely check him out if you guys want. Alright. And then Nick Rougeau earlier said something like, my question is very similar to this if you wanna skip.
13:40So I appreciate that. And then f one d one four two, who changed his profile pic, says, not sure if you've Hermosy talking about this recently, but he's been pretty bullish on live content and interactive streams. You should, uh, I don't know, do live content and interactive streams.
13:54And, uh, to that, would say, yeah, it's pretty cool. I mean, like, I've I've considered doing it and we also played some video games with Jack Roberts, if you guys are, uh, if you guys remember that, which is pretty fun. But, um, just trying to find a a way to schedule it, think, is the hardest part for me.
14:07Look. You gotta understand that, uh, the cool part about, like, daily updates, for instance, is I could do them anytime I want. They're hyper efficient.
14:13Like, if I have twenty nine minutes between one meeting and the next, I could just, like, press a button, record a video, and then go, and I get to utilize my time a 100%. But if I have to now schedule it at, like, 10AM to 10:30, I have to, like, reorient my time around that, then I have some dead time before and after, which kinda sucks.
14:27So, you know, it's not that big of a deal, but probably not every day. Probably not every day.
14:33Okay. Unfortunately, I'm running low on time because I didn't realize that I wasn't actually recording.
14:39So what are we gonna talk about? I'm moving.
14:43I'm moving. That's pretty fun. Moving.
14:46So that's why everything in the background's all weird. I got a big bookshelf, so I'm gonna see if we could get some people to come by pick pick up the bookshelf today. I've got a lot of stuff.
14:55I've a got a sofa I'm gonna give away. Got a big dresser we're gonna give away. I'm gonna do away with this desk, which kind of sucks because I kinda wanna work for the next few days, but whatever.
15:05I bought, like, an Eight Sleep mattress topper, so I'm really excited about that. I think that'll improve my sleep a few additional percentage points.
15:13Considering buying a new couch. So that's just a little personal update, but, yeah, we got a lot going on on the moving front. Apartment's gonna be sick.
15:20New place that I'm moving to is really tricked out. I mean, I'm not gonna say it's like a luxury apartment like this one is.
15:27The one I'm currently at is definitely, like, luxury. We have, like, full marble backslashes, and everything's all sexy, and the veneers and whatever are really nice.
15:35But, uh, this next place I'm moving to is kinda like mid market, which I really like because it just means that all the appliances are, like, really stress tested. They've had a lot more quantity to build the quality. So one thing I'm particularly stoked about is the gym is thirty seconds from my front door.
15:49What I mean by from my front door, I mean my unit store. I literally, like, open my door, and then I go down a bunch of concrete stairs, and then go, like, a little bit down a hallway, and then there's a giant commercial sized gym. Oh, yeah.
16:00Squat racks, you know, all machines. Dumbbells only go up to 70 fives, but I think that's still pretty good because we have, like, barbells and stuff for for all other purposes. Got, like, the the the best compound machines.
16:13What what are they called? Like, not the Bowflex, but, like, you know, one of those big stand up machines where you could do, like, tricep push downs, but then also, like, cable fly you know, cable machines, basically. So, yeah, I'm really amped for that because right now, you know, if I wanna go to the gym, it takes me probably, like, ten, twelve minutes just to, like, prep them and go there.
16:30I gotta put shoes in my bag because in Calgary, um, it gets really dusty, and they spray a lot of salt and rocks and and stuff like that to prevent people from slipping on snow all year. So there's just you always have to change your shoes before you go to the gym. That's their policies.
16:42I don't have to do that anymore because I don't have to leave my I can stay inside the entire time. They got a sauna right next door to it, so I'm just gonna sauna. Uh, it seems really cool.
16:52I'm really excited about that. And then yeah. I mean, the weather's gonna be way better than the weather that I have out here.
16:57That's not gonna snow for five and a half, six months a year. So that's pretty sweet. So, yeah, I'm just doing a lot of that today.
17:04Unfortunately, I wanna, you know, do more work and add value to my life and and the career aspect, but it's definitely gonna be like mostly, like, life day, I would imagine. Anyway, yeah, Maker School's doing pretty well, um, on the tracking front.
17:18Uh, we went down one user in the last twenty four hours, so twenty eighty five, which is okay. Um, I still have yet to, like, scale up the content, which is why I'm gonna keep seeing this.
17:28I imagine I'll probably keep seeing this until maybe about a year or so. So maybe we get down to, twenty fifty. But, um, I think with enough momentum, we can turn it back around pretty easily.
17:37I'm also worth noting that the r ARPU, so average revenue per user, is going up despite the fact that the, um, count is going down because we have a lot more people signing up to the 30% off annual plan, um, which is quite nice. And then we also have, um, like, we're front loading their revenue, but then we're also most of the people that we lose are people that were, like, on really low plans, like 49 a month, which I didn't realize Maker's School used to cost.
17:59That's wild. Um, Yeah. So that's that.
18:02On the YouTube side, we're at $4.52 $3.77. You know, I published this video, GLM 5.2 is basically Opus for a fifth of the price, and initially performed really poorly.
18:13But, um, over two days and fifteen hours, we're we're up to almost, like, 31 k views. So we're actually almost at our typical range here. And I think it's just because the Internet has realized that I'm right.
18:23What I've realized is, like, I tend to be pretty early on stuff. When I started talking about how OpenClaw was trash, people were like, no. It's not.
18:29It's awesome. You're crazy. And then, like, a month later, everybody's like, yeah.
18:31OpenClaw sucks. I think people are going to realize that most of what I talked about in that video is true, and I'm not trying to pretend like I'm some crazy prophet, and I know way more about this than anybody else. But, you know, people have been, like, very anti open models, myself included, simply because they usually suck.
18:48And this is, like, really the first time it's, like, not sucked blown chunks. So, uh, it's just gonna take time for people to really understand that. So Instagram's five twenty eight eighteen.
18:59I should note that, um, we're actually finally gonna start publishing videos on Instagram again. So hopefully, after today, this goes up to like at least point one on a daily basis. And we could see some growth there.
19:09Because, I mean, otherwise, this is gonna absolutely crush Instagram. And it's been on that trajectory for a while now. Like, if that is Instagram, this has been my YouTube.
19:19But I I would like to see if I could scale that curve up a little bit. That'd be nice. And, yeah, I mean, like, I've been using AI to do a lot of things recently.
19:30Really gone deep into, like, agent management and engineering, uh, which I'm pretty stoked about.
19:38So I finally sorted out, a good mobile workflow. And, um, the play is, for anybody that's wondering, like, how to do this stuff on their phone, you just you open up, like, five Cloud Code chats.
19:51You go slash remote control on each, and you just rename them all. So you have, like, Cloud Code business, Cloud Code personal, Cloud Code this, Cloud Code that.
20:02And And then you just get like the cloud code app on your phone.
20:06You turn your computer so that it like never turns off. You caffeinate, you do whatever the hell, you close the lid, and now you can just use cloud code exactly like you were doing at home, just through your phone. And because you have threads that auto compact after a while, they actually accrue valuable knowledge.
20:19And then if you wanna do something in business, you just use your business thread. If you wanna do something that's more general, use your general thread. You wanna do something that's, I don't know, something else, whatever, use your your other thread.
20:29So, yeah, I mean, hopefully, that's at least halfway valuable. But I found a lot of value out of that. Like, yesterday, I had a podcast with the lovely Lazarus who came all the way from Cyprus to hang out with me and then filmed in, like, this gorgeous studio.
20:40I was running agents that entire time, and they were all doing economically valuable things for me. They were cleaning lists for an upcoming campaign I'm launching. They were, you know, writing stuff for me with the humanization.
20:50They were working through the humanization prompt. Like, it was pretty cool. It's pretty cool.
20:54Finding moving companies and all that sort of shit. I'll leave it there because now I'm starting the app. But, uh, have a lovely rest of day, and I'll catch all y'all on tomorrow's daily update.
21:02Cheers, everybody.
The Hook
The bait, then the rug-pull.
The title promises a cheap automation trick. What you actually get is 21 minutes of a $4M-a-year agency founder answering his audience's questions in real time — and then, buried at the 10-minute mark, a genuinely new idea: an AI agent that runs on a loop for days, refactoring your n8n workflows into serverless code until your costs drop by 10x.
Named ideas worth stealing.
01:30model
Customer Journey Discovery Framework
Ask the client to narrate the full customer journey from first brand contact to end of engagement. The journey framing auto-surfaces revenue-forcing steps and filters out admin noise.
Steal forAI agency discovery calls, process audit kickoffs
10:22model
/loops Automation Optimizer
- Define cost or time reduction goal (e.g. 10%)
- Point agent at n8n account and existing automations
- Loop every 30 minutes: hypothesize, test, log to shared file
- Agent refactors modules into code blocks
- Deploy code blocks to serverless (Modal, Trigger.dev, Lambda)
- Stop when goal is met or after defined run window
Multi-day autonomous refactoring loop that converts visual n8n workflows to serverless code, compounding cost savings over time.
Steal forAny high-volume automation client with real token or execution costs
09:12model
Humanization Prompt Pipeline
- Build statistical model on personal writing corpus
- Generate sentence fragment candidates with AI
- Score each fragment by perplexity vs. personal corpus
- Select lowest-perplexity continuation
- Repeat per sentence
Produces AI-assisted content indistinguishable from native voice by selecting for low statistical deviation from the author's own writing history.
Steal forScaling content output without losing voice authenticity
19:36model
Mobile Multi-Agent Workflow
- Open 5 Claude Code chats
- Run /remote control on each
- Name by domain: business, personal, etc.
- Enable caffeinate or no-sleep on laptop
- Access via Claude Code mobile app
Persistent named agent threads that accumulate knowledge via auto-compact, accessible from phone while the laptop runs in the background.
Steal forAnyone who wants AI agents running during meetings, travel, or off-hours
How they asked for the click.
VERBAL ASK
20:57next-video
“I'll catch all y'all on tomorrow's daily update. Cheers, everybody.”
Soft verbal sign-off with no explicit subscribe CTA or link. Relies entirely on series habit rather than a direct ask.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
