If you don't have an audience for your SaaS, I would not spend a bunch of time building an audience. By the time you build the audience, I don't think SaaS as an industry will really make much sense anymore, just given that AI can virtually rebuild anything and everything today for tokens on the dollar, so to speak. What I would do instead is build a two-step process. Rather than building your SaaS, building an audience, and then selling it, I would just build a SaaS and then sell it. You do this by going outbound, not inbound. You don't need to build an audience in order to go outbound — you literally just find people that would get a lot of value out of the product you're selling and go talk to them. There are a variety of different ways to do that; obviously Maker School is a big chunk of the outbound approach. You just reach out to people that have some sort of buying signal and then put your product in front of them in a highly curated way.
Let it ride. Do not publish. I think he's talking about a video I published on my main YouTube channel a while ago that went super viral — we're at like 1.3 million views now — and I did indeed give it some space to grow. That's generally what you should do: give it a little bit of space to grow, not too much space. I think a lot of people take this too seriously, but on the main channel where you have a lot of growth occurring on a specific YouTube video, just give it 24 hours. Don't publish anything in the next 24 hours after that video and you'll have significantly higher odds of success, because your channel can only compete in browse to a certain extent — you can't fill every video slot with your video. Logically speaking, if you have multiple additional new videos, your subscribers will be forced to pick between one or the other, because some people will only ever watch one. So you just kind of have to give it a little bit of room to breathe.
I would not identify with a particular political party, and I also wouldn't necessarily identify with YouTubers that just so happen to be liberal or progressive. If you really wanted people that are liberal or progressive, you should just go straight to these big CEOs that run these companies. You'll probably look at me and be like, 'What the fuck are you talking about, Nick? Why would Sam Altman be a liberal or progressive?' Have you read Sam Altman's blog? Is there literally anything more liberal or progressive than the claims he makes in his blog? This guy was talking about AI 5 or 10 years ago — AI is going to be used to read all legal documents, generate all scientific advancements, rectify disparities between all races, people's creeds, colors, genders, whatever the hell. This guy is so far ahead of the curve. It's so interesting to me that he's now considered this big tech oligarchy guy despite actually espousing probably the most liberal, progressive, and some might say extreme views of all. So if somebody says something that you like, obviously I wouldn't retroactively be like, 'Ah, but are you liberal or progressive?' and if the answer is no then you're not the sort of YouTuber I want to follow. But yeah, I don't know any, because I don't really listen to YouTubers on a political basis.
There isn't going to be a lot of AI video production demand on Upwork, but there is going to be a lot of video production demand on Upwork. What you need to understand is that AI is not necessarily something you can only sell if somebody says AI. You could fulfill the needs of people that just need video in general, and then show them how you could use AI to do it. You could also sell the outputs of AI, not necessarily AI. When I started out in lead gen, for instance, I didn't say 'I'll build you an AI-powered lead gen system,' I said 'Hey, I'll book you 20 meetings guaranteed in the next 60 days, or you don't pay.' I was selling the output, which is quite valuable in business and has always been quite valuable — not the means to get there. So that's personally what I would do. Do not look for people asking for AI agent creative pipelines. Literally just build a creative pipeline that uses AI. It's just normal — it's part of the day-to-day of working in business nowadays. Nobody will really care.
It really just depends on what your hardware is. If you have the hardware capacity to run a Gemma 31-billion-parameter model, you should do that, but it's always a trade-off between tokens per second. You can figure a lot of this out by heading over to Llama or Hugging Face and clicking get started — I have a free account over there. Scroll down to models and you can click on virtually whatever model you want. This one here is 700 billion parameters, so you're not going to be able to run that. First, filter by local, because some of these have cloud offerings too. Then you'll see something like Bonsai has 2B, 4B, 9B, 27B. What you could try doing is setting up the 27B; if that doesn't work, move back to the 9B; if that doesn't work, move back to the 4B. Obviously that's a vision model, so it's probably not exactly what you want. A bunch of these Neotron ones are pretty good. Qwen 3.8 I think is basically state-of-the-art right now, so you could go 27B and see if that runs. A lot of people host quantized models on Hugging Face, so you could get a quantized Qwen. You can also get like Qwen 3.5 4B quantized — an older model, but because it's quantized it runs on significantly smaller hardware. The optimization reduces the model weights from 8 gigs to 5.2 GB on disk, meaning you can literally run this thing on an 8 GB computer. So figure out your computer and adjust accordingly.
Fairly straightforward. You just plug into an image model — assuming these are static images, an image model like GPT Image. I think you could probably generate images via Sunburst for like 10 or 20 cents depending on the aspect ratio and size. I would just do square because that'll be more performant for you and it'll still be a high-quality image. I would get a bunch of images that they're currently running, maybe from the ad library, because presumably they have a particular look. If they don't and they're just ripping whatever works, then obviously you can just rip whatever works. Then I would just generate like five of them and build an app that is essentially a recreation of the Meta feed of Instagram, Facebook, or whatever. Then I'd have that app as a link you could deliver to them, and maybe use URL params to host one app but host all of these carousels and have them dynamically pulled into the app. That's probably what I would do. You could just copy the transcript part of this video, feed that into a frontier agent, and it could do a pretty good job.
Define what you mean by scratch. If you mean scratch in terms of my ability to use AI tools, but not scratch in terms of my ability to sell, I think I could go to $10,000 a month this month — probably in the next 15 days or so — with zero knowledge of any of the stuff. Because you'll find that the bottleneck is not really the tech skills. You could learn everything you need to know in presumably two or three days of full-time, 10 hours a day blasted against the keyboard. At the end of that three-day pipe, you'll be a very different person — you'll be able to use these tools far better than most other people can. From there, I would spend the other 13 or 14 days just pounding marketing and then eventually pounding sales. I would probably sell a fixed-price offer, keeping in mind that when you sell, you're selling in the domain of ideas. You're not necessarily selling a physical thing — you don't need to create a product in order to sell it. You just need to talk to a bunch of people about something that could be a product. Then I'd blast it. Assuming my outreach machine is more or less what I talk about in Maker School, I could probably book a meeting a day, two meetings a day, so I'd probably have like 20 meetings in the next 1.5 to 2 weeks, and then I'd probably close like a quarter of them — that's a pretty low conversion rate, but I'd probably close a quarter. So I'd just have to sell like $2,500 fixed price. If I absolutely crush it on one deal that I start today, then by the time 12 days comes around I might have even done a delivery, and that delivery might have allowed me to do an additional upsell. So maybe the LTV of a client would have been like 4K and I really only would have needed 2.5. That's probably what I would do if my goal was to make $10,000 a month.
YouTube native dubbing is unfortunately not very good. You can't really do it for long-form courses automatically, so you have to do it yourself, and that's what I've been doing. I've been using ElevenLabs. It's been okay. The issue is obviously it's all flagged as AI because I'm using an AI voice dubbing tool, but it's not the end of the world. We've also experimented with specific dubbing channels like Mix Ride Bengali, for instance, where I dub and do some lip sync stuff. These are all being flagged as AI and there's like zero performance on them, which fucking blows. So I'm probably going to redo that. I'm not 100% sure.
I think we'll just see a general uplift in human learning potential. I think it'll be quite similar to the invention of the internet more broadly. The internet is a sharing resource. It wasn't the internet that led to all of these breakthroughs quantitatively, but qualitatively it was a massive lift to an average researcher's ability to share notes and see what other people were doing. Just imagine the transition from back when you needed to go to dusty old encyclopedias to reference works done hundreds of years ago, to now when you can just type into a keyboard and get virtually whatever answer you want to any question. Then think about the next stage, which is when you can have agents actually act on this stuff to do that for you. That said, I think we're going to go crazy AGI well before that.
Yes, automating LinkedIn DMs would lead to a profile ban when you inevitably get caught, so I would not automate LinkedIn DMs. You really don't have to, because as you mentioned you can only do about 20 per day depending on your LinkedIn premium plan — you can do a little bit more, probably 25. So you send 25 a day, awesome. How many of those actually accept your request if you're good? It'll be about a quarter, maybe a third. So that's going to be like six or seven a day. How many DMs are you really going to send a day, dude? Six or seven? You really need to automate that? I don't think so, not for one account anyway. If you wanted to scale past that, your only real option would be to purchase or rent accounts in some way, shape, or form. If you start a business and have a bunch of people working under you, you can also just use their accounts. But basically you're sending six or seven DMs a day. In Maker School right now, the Loom-to-booked-call rate is something like 5 to 15% — it varies because some people suck at it and some are good. Think about it that way: if it's 15% and you send six or seven DMs a day and you get one booking a day, you're now getting one booking a day per LinkedIn account. If you have 10 LinkedIn accounts, you're getting 10 bookings a day. Math is pretty good, right? And it doesn't really take too much time or energy to do these DMs. I would just recommend recording a video and making them really high quality.
I keep seeing these videos where people send thousands of emails per day. They do so using a simple tool called Instantly or Smartlead. Essentially these are sending tools that allow you to very quickly procure a bunch of mailboxes and then distribute like a thousand emails across 20 mailboxes — so 50 a day. Super valuable. That's what they do. And then they'll usually send them to leads that they scrape off LinkedIn Sales Navigator. That is the number one source right now. You can scrape LinkedIn through like 20 different ways, which is why it's such a clusterfuck — there are so many platforms.
I can't offer too much information about that. Unfortunately, operational security dictates that I should not be talking about where I'm storing my money. If you did want to know how I'm storing my money, specifically my actual day-to-day investments, I actually have that as a module inside of Maker School. Both for security's sake and the fact that I think agents are going to be scraping the hell out of all this information pretty quickly, I'm not going to give you guys where I'm banking and what my investments are. But anyway, a big app that I love to use for anything and everything task-related is Linear, and I've piped that into Raycast, which is my little launcher here and allows me to very quickly do whatever the hell I want. You can see my number one most used app is Beeper Desktop — that's because it's a unified messaging app that allows me to get WhatsApp, iMessage, and like 20 other tools all in one. Then it's really just Beeper for all comms, OBS for all my video, and Spotify, which is kind of quickly being phased out for Suno. And I take notes just using Linear. If I have a great idea for something, I'll just pump it in there. Linear is like the backend.
productivitylinearraycastbeeper
#349should ai write your cold email icebreakers?watch source ↗
Great question. I would probably flip things on Marketplace, Kijiji, Craigslist, and I would use AI in order to auto-negotiate across the board. So set up a bunch of AI agents to auto-negotiate on various purchases for you and then just flip those. Basically do the exact same thing on the other side. You could probably manage like 50 to 100 of these. The only issue is going to be coordinating pickups and drop-offs and stuff — that's going to be a little bit trickier. But you'll probably be able to make up the negotiation to purchase or request drop-offs for smaller fees as well. So if you want to have somebody drop something off and then come and pick up that same product later on for more money, it's very straightforward arbitrage. Obviously, use an address you're comfortable with.
Icebreakers are like anything else. If your icebreakers are high quality and you have a process that consistently delivers results, and then you use AI to help reproduce that at scale, then you can make a lot of money with that. But if you're AI generating really shittily written icebreakers, the probability of this working is very, very low. So you just have to generate high-quality icebreakers. If they're high-quality, you're good.
How do you handle cold email in the EU? It's really not all that difficult. One, and I'm not a lawyer and none of this is legal advice, but when's the last time you jaywalked? Congratulations, you have essentially contravened a law or a rule or a regulation. Cold email is a lot like jaywalking. If you're good at what you do, if you look left and look right and there's no cars oncoming, nobody's really going to care. These laws are also written in such a vague way that if you take their definitions at their vaguest, you're basically always going to be at the behest or at the whims of the legislators. And so if you send really high-quality customized cold emails that provide value to the people reading them, and they don't come across as spam and don't seem like a big issue, you're almost certainly going to be fine. And what's interesting is Germany, Spain, a lot of these other EU countries are starved for cold email right now. If you take the same campaign you're running in the United States, then localize it a little bit — the language is quite similar, but still respect cultural norms of deference and so on — and then run it in Germany, you'll get like a four to five x reply rate. Same thing in Spain, same thing in the Netherlands. These are all places where people are starved for this sort of outreach, and they want to connect and grow their businesses and network. They just don't currently have a way to, because regulation has a stranglehold on a lot of what previously would have been serendipitous. It's unfortunate that the EU has strangleholded people to that degree, but they definitely have. People now are so afraid to even send an email to somebody that they don't know, because they think SWAT teams are going to break in through their windows and get on the floor — 'this is the guy that sent the cold email, loser, I hate people like you.' Obviously that's ridiculous.
What would I do differently if I were starting over with my agency in 2026 or 2027? Great news for you — I just recorded a video that talked all about it. It's literally right over here. The actual title that I started with is 'How I'd Learn AI From Scratch If I Had to Start Over,' and this is actually what I would watch. I also touch on agency stuff as well. So honestly just check that video out — it's on my main channel, came out earlier today.
Yeah, it almost certainly did get soft banned. It blows because this AOT video did really well initially. There's such weird algorithmic stuff going on here. If you zoom way in, you'll see there's a huge spike up and then it just plateaus. This sort of plateau is pretty indicative of just not showing up in anybody's recommended — only showing up on your own channel. So obviously they probably didn't like the fact that it was Attack on Titan, and I completely understand. But I've since changed it so it doesn't actually read AOT.
All forms of personalization and all forms of scraping are going to be against the terms of service of the websites that you're scraping from. That's just how it is. AI agents are really changing the scraping dynamic such that virtually every website is going to be scraped hundreds of thousands of times a day. I don't really think that sort of regulation, if you want to call it that, will ever be able to keep up. So how do you scrape stuff outside of LinkedIn? You can scrape things outside of LinkedIn, but you have to understand that LinkedIn is the whole juice — that's where all of their information is, and they put it on that platform specifically for the purposes of connecting and networking with people. That's what people don't understand: 'oh, you need to opt in.' Well, dude, you're on LinkedIn, which is literally a platform for business communication. It's like, 'add this information so people can help connect with you more,' and they do. It's user-generated — it's all UGC in a way. So that information is up for grabs, and if you can get it literally by just going to somebody's LinkedIn homepage, was it really ever that private to begin with? I don't think so. So that sounds totally fine to me.
I find it funny that typically the people that were the most progressive-minded, who always default to science with a capital S, are now disproportionately the people yelling at technology for replacing the various creative aspects of their being. Not understanding that all of these industries have already been captured big time by things like recording studios, record labels, musicians. This is a business, right? Now small one-shop creators actually have the ability to completely disrupt that economic model and get back to doing cool shit again for the purposes of doing cool shit, and actually really innovate. It sucks, but really what people are mourning is more so just the loss of a way of life that they spent a big chunk of time preparing for, and I can empathize with that and I don't mean to be a dick about it. When OpenAI came out with GPT-3 in 2020, I went for a really long walk and had tears in my eyes because I was like, 'wow, everything that I've spent all this time preparing for — at the time was writing — is soon going to be automated.' If AI can already write at this level in just a year's worth of advancements from GPT-2 to GPT-3, what's GPT-4 and GPT-5 and GPT-6 going to look like? Well, we have our answers now. What happened to the writing industry, which I spent so long getting into? My whole business model was destroyed, to be clear, and everybody's business model is going to get destroyed. But that's not a bad thing, it's exciting. Because you could be the person that is actually going about creating new business models on the bones of your fallen comrades.
