YT YouTube2Text

YouTube2Text → Transcripts → How I Built a $100K AI SaaS Without Knowing How to Code (5 Easy Steps)

How I Built a $100K AI SaaS Without Knowing How to Code (5 Easy Steps) — Transcript

by Liam Ottley · 6,811 words · 959 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00If you have ever wanted to build your
0:01own SAS but felt like you weren't
0:02qualified or didn't have the skills to
0:04do so, this video will completely change
0:06your mind on that because we have Ben
0:07Webb coming on here who until recently
0:09was completely non-technical and since
0:11then has been able to make over $100,000
0:13with this very, very interesting dental
0:15sass that he's built. And this is an
0:17awesome example of people coming to the
0:18AI agency route, finding a great use
0:20case, building a software, and in Ben's
0:22case, feeling like he's completely out
0:24of his depth, but pushing through anyway
0:25and building an incredible business as a
0:27[music] result. In this video, Ben
0:28breaks down his entire process of how he
0:30found his idea, how he validated it, the
0:32exact tech stack that he uses that costs
0:34less than $100 per [music] month to be
0:36able to build his own SAS and walks you
0:37through the process of building the
0:38front end, the back end, and so on
0:39[music] as a complete beginner. So, this
0:41is a really, really valuable one for
0:42anyone looking to build any kind of SAS.
0:44Hope you guys enjoy this one. All right,
0:45Ben, great to meet you, man. Thank you
0:47for coming on. Um, you've had a pretty
0:49awesome run here with your uh vibe coded
0:51SAS in the dental industry.
0:52>> Yeah, thank you for having me on, Lamb.
0:54Back on in the flesh this time. I'm
0:55really excited for this one. That's
0:56going to be a banger. So, what I've got
0:58for you guys today, as you can see on
0:59screen, is a five-step, a simple and
1:02actionable framework that you can take
1:04to go from idea to a SAS MVP, even if
1:07you're someone someone like myself who's
1:09never touched code before, doesn't have
1:11a developer or a massive budget. Moving
1:13into step one of the framework, I
1:14actually want to start with my story
1:16here to really give you guys an
1:18understanding of uh what a realistic uh
1:20experience and expectations of this was
1:23cuz mine certainly was not a linear
1:25story. So random Bulgarian dentists
1:27almost paying Microsoft $1.5 million. So
1:30basically I met this gentleman um on the
1:32other side of the world in Eastern
1:33Europe just about as far as far as you
1:35can get from Melbourne came in through
1:37my school group and we kicked it off. We
1:39had a common interest in AI and
1:40automation and after talking for a while
1:42we had this idea to train an AI X-ray
1:45diagnostics model. was something that he
1:47wanted for his own clinics actually as a
1:48diagnostic tool and I liked the idea of
1:51it for a SAS um but I at the time had no
1:54idea how I was actually going to
1:55leverage it um because the AI diagnostic
1:58side of of dentistry is dominated by
2:00some massive um corporate companies that
2:02I have no business competing with. Now,
2:04we gave this idea some serious
2:05consideration and we actually went to
2:07Microsoft as yours team and got a
2:10shockingly a $1.5 million quote to train
2:14this model, which was slightly more than
2:15I was expecting. And to be completely
2:17honest with you, I actually considered
2:20uh cutting out both of my kidneys and
2:21selling them to make this work. But uh
2:24moments before grabbing a scalpel, we
2:25actually had a bit of an epiphany that
2:27we'd already done the hardest part,
2:28which was because he had been a dentist
2:31for many years running his own clinics,
2:32we actually had all of the data we
2:34needed to train the model. He had
2:35thousands and thousands of X-rays
2:38sitting around basically gathering
2:39virtual dust. So we hired a dev team and
2:42got it done in 45 days, which was far
2:44far less than I was actually expecting.
2:46>> So that's just using the pairs of like
2:48input and output. Like this is the X-ray
2:50and this is the human like human
2:52diagnostic off the back of it.
2:54>> Yeah. Exactly. So we use a platform
2:55called I'll drop the source here called
2:56RoboFlow. So this is basically a um
2:59computer vision machine learning kind of
3:01thing
3:01>> training platform.
3:02>> Yeah. Yeah. So I just want to I just
3:05want to say quickly here this is a bit
3:07of a unrealistic situation where I've
3:09been presented a unbelievable resource
3:12basically from the sky. So most SAS will
3:15not require something like this to get
3:16started up. With that being said, if you
3:18put yourself out there, and I know you
3:19can speak to lamb, uh, you can speak to
3:21this lamb. If you put yourself out there
3:22and you network with people and you post
3:24content, things like this are just going
3:26to happen. And that's going to be the
3:27spark to starting your business. You're
3:28going to meet your business partner.
3:29You're going to get an asset like this.
3:31So, how to actually find a problem. Now,
3:34generally, the way I like to look at
3:36this is inside any industry, there's a
3:38pool of problems that exist. And then
3:40when a new model or AI software comes
3:43out, it sort of unlocks a few of these
3:45problems to actually be solved. Now, two
3:47concrete examples here to really
3:48illustrate this to you guys. Voice
3:50realism and latency technology getting
3:52finally good enough with things like 11
3:54Labs or GPT real time. And all of a
3:56sudden, you've got, you know, customer
3:58service, inbound calls, outbound calls,
4:00um, reactivation, receptionist, this
4:01massive chunk across hundreds, if not
4:04thousands of verticals basically opens
4:06up overnight. Same thing with text,
4:08image, and video generation models. Sora
4:09Nano Banana content creation which is
4:12you know this massive chunk again across
4:14hundreds of verticals opens up
4:16overnight. So I don't want to hear
4:17anyone in the comments saying you know
4:18finding a problem's too hard where do I
4:20start? Like this is the market we're in
4:22right now. This stuff is happening on a
4:23weekly basis. Okay guys very quickly. If
4:25you're an aspiring entrepreneur and want
4:26to start your own AI business and you
4:28haven't already joined my free school
4:29community it's down there in one of the
4:30links in the description below. Has my
4:32full free course on how to start your
4:33own AI agency as a complete beginner.
4:36and you're surrounded by over a quarter
4:37million people who are also striving
4:39towards the same thing. There's no
4:40better place on the planet right now to
4:42be surrounded by like-minded people and
4:43you get free weekly Q&A with me where
4:45you can ask questions directly to me
4:47about how to start and scale your
4:48business. I'll see you in there. I
4:49literally was just talking with Corbyn
4:50about this. It's I call it I call it
4:52spawn trapping like like you're
4:53basically spawn camping. The new model
4:55comes out there's like a fertile plane
4:57appears is what I call like a step
4:59change in the model capabilities. Okay,
5:01the image models are now way better and
5:03you can put text in it. I remember that
5:04was like the GPT40 image gen came out
5:07earlier this year and suddenly you could
5:08put text in it and it had like much much
5:10better image generation capabilities.
5:13That's a step change and there's all
5:14these new business opportunities that
5:16get unlocked by it. So this is a you can
5:18literally just camp the new models as
5:20they come out and wait for a wait for a
5:22really big step change to come.
5:23>> Yeah. And it's not like it's hard to see
5:24what's coming next, right? If you pay
5:26even the slightest bit of attention, you
5:27can see what the next move is. So also
5:29something to be cognizant of here.
5:31Crumbs can be worth an 8 to nine figure
5:33valuation. So what I mean by this and
5:35I'll pull from my personal experience
5:36here when I was going into dentistry the
5:38problem that I'm solving with my SAS
5:40which is um manual treatment plan
5:42creation basically is not problem 1 2 3
5:45or five in the industry but because
5:47dentistry has a large enough total
5:49addressable market that problem is still
5:51worth pursuing. It's the crumbs of the
5:53industry if you like but it still can be
5:55worth an 8 to9 figure valuation. So when
5:57you're looking at an industry, don't be
5:59surprised when the, you know, top 10
6:01biggest pain points are already solved
6:02by companies that have got there faster
6:04than you. But it's still worth, if it's
6:06the right market, pursuing the other
6:08problems in there as well. Now, also the
6:10world's completely changed because what
6:12a lot of people their attention is drawn
6:14toward when they're thinking about this
6:16is okay, the barriers to entry have been
6:18lowered from the coding side, but
6:20they've also been lowered drastically
6:22from theformational side. And again,
6:24pulling from my personal experience
6:25here, I'm not a dentist. My experience
6:27as far as dentistry goes is I've got a
6:29few teeth and I've been to the dentist a
6:31few times and had them poke around in my
6:33mouth. Yet somehow I have been able to
6:35build a SAS around dentists. Now you can
6:38leverage tools like uh Chat GBT's deep
6:40research for example to pull from
6:42thousands of data points on the web and
6:44then present all of the information to
6:46you in a concise articulate manner where
6:48you actually have all the information
6:50you need to understand the the workflows
6:52the preferences of the professional that
6:54you're building your SAS. So how to
6:56actually go about finding a problem?
6:58Well Chad GPT is a great place to start
7:00but there truly is no substitute for
7:02talking to profession like real people
7:04in the industry. So friends, families,
7:06colleagues, anyone you know who is a
7:07dentist, a chiropractor. And generally
7:10inside um the problems that you're going
7:12to be looking for will be inside two
7:14pillars. How do you get customers and
7:16how do you fulfill the service? Now of
7:17course there smaller things under these
7:19two umbrellas. You know customer
7:20service, onboarding, lead qualification,
7:22etc. But these are the two sort of
7:24pillars that you want to start with and
7:25simply ask them what makes you want to
7:27punch your computer? What makes you want
7:28to throw your computer in the trash?
7:29Like this guy here. Um there is always
7:31something. There is no industry that
7:33exists, not on earth, that does not have
7:36some things on a daily basis that just
7:38annoy the heck out of the owners. In
7:39fact, I first want to ask you, Lamb,
7:40this is obviously one of the most
7:42crucial steps to do properly, otherwise
7:44your SAS will crash and burn. So, I want
7:46to ask you, is there any pitfalls that
7:47you commonly see beginners doing here?
7:49>> It's always too big. Like, they're
7:51trying to solve way too much. And I
7:54think they're underestimating the
7:56difficulty of the dev and overestimating
7:58their ability to to vibe code. um at at
8:01the end of the day like you like start
8:04far far far far smaller than than you
8:06think and and like uh like I've said
8:09before um previously is like you want to
8:12make sure that if you're picking and a
8:13problem to solve make sure that the the
8:15gold mine is built on top of the gold,
8:18right? Like there's certain use cases,
8:20there's certain uses of these models
8:21that instantly create value in a way
8:23that's like so clear and undeniable that
8:25it's it'll be a hit right off off the
8:27rip. And those are the ones that you
8:28should of course be trying to find. And
8:29there's other software use cases where
8:31you can you know like task management or
8:35like productivity or like like maybe a
8:37little bit of scheduling here and there.
8:38In the case of image generation models,
8:40video generation models, uh computer
8:42vision in your case, uh there are some
8:44very very very immediately clear value
8:46creation moments you can use with the
8:48you can get with these models and the
8:50rookie will kind of shoot for something
8:52they think sounds interesting and build
8:53this fancy software and try to iterate
8:55until you get product market fit. um
8:57when you could probably just have like
8:58really like cracked it off the off the
9:01tea um by just building over the over
9:04the go over the gold in the first place.
9:05Now, a bit of a pro tip here. Once you
9:07have found an industry professional,
9:09this is going to be an extremely and I
9:11cannot stress this enough, an extremely
9:13high lever contact to have. So, nurture
9:15this relationship, keep in touch with
9:17them because you're going to need to use
9:18them later. So, that takes me to step
9:21number two, which is building the
9:22output. Now, I want to put an emphasis
9:25on what I said here, which is speed is
9:26key. What you're not doing here is
9:28building a dashboard, a UI, a login
9:31system. You just need to build the core
9:33output that solves the problem. Okay?
9:35Because what you need here before you go
9:36spend time and money building something
9:38else is fast feedback. So,
9:40realistically, you have two options to
9:42do this. So just just to just to clarify
9:43that you're meaning the core
9:45functionality of the app in your case
9:47the model that's able to take in an
9:49X-ray and and provide the like JSON
9:51output to uh give it the diagnostic
9:54right.
9:54>> Yeah absolutely. So for me it was more
9:57the actual report like can we is the uh
10:00the treatment report that we generate
10:01for the patients does that actually
10:02suffice. So yeah the the AI model was
10:05the hard thing to get but um yeah
10:07definitely just the core output that
10:08solves the problem and not the fluff
10:10surrounding it. So, two options to do
10:12this, make or n. Now, if you're not
10:14familiar with these platforms, YouTube,
10:16Liam's channel in general has a lot of
10:17content about how to learn these. Um,
10:19but generally, this is the approach you
10:21want to go for if your output requires
10:23heavy logic or complicated processing to
10:25be done. Now, bit of a hack to speed up
10:28this process is getting the blueprint
10:31generated from chat GPT. Now, make and
10:34ed have got I think they're starting to
10:37see some vibe automation, if that's what
10:39you want to call it, where you just
10:40prompt inside these platforms to build
10:41it. Um, I personally don't think they're
10:43all the way there yet. So, this is this
10:45is just what I use to do the heavy
10:46lifting for me. Um, but that is actually
10:48inside of chat GPT. So, I said generate
10:50an importable make.com JSON blueprint
10:53for a scenario using the following
10:54instructions. And then you're obviously
10:56going to um have a comprehensive prompt
10:58there about what you want that blueprint
10:59to to do and look like. Then it'll give
11:01you the JSON file back. So download it
11:03to your desktop and then when you're
11:04inside this platform whether it's make
11:06or or nadm import. So hit the three dots
11:08import the import the blueprint and then
11:10it will populate all of the modules for
11:12you. Now this won't be perfect out of
11:14the box. You will have to do some
11:15tweaking but it definitely will do 90%
11:17of the heavy lifting for you. Option
11:19number two is vibe coding. And now as I
11:21said I'm going to get into a lot of
11:22depth about vibe coding tech stack all
11:24of those things. But to put it simply
11:25right now, this will be your approach if
11:27you have like if you can't no code the
11:30output if it's a little more complicated
11:31or requires heavy processing. Um this
11:34will be your approach. Now the balance
11:36you want to strike here kind of goes to
11:37what I was saying before is there's sort
11:39of two ends of the spectrum when you're
11:40looking at this. This end of the
11:41spectrum is you don't spend enough time
11:44developing the output and there isn't
11:46enough substance there to get proper
11:47feedback on it. The other side, and this
11:48is what beginners tend to gravitate
11:50towards, is overdoing it, spending too
11:52much time, and you obviously want to
11:54find a nice balance there. Have just
11:55enough substance that when you go talk
11:57to a professional to get market
11:59validation, they're giving you proper
12:00and useful feedback. So, a bit of a rule
12:02of thumb, if your output is useful but
12:04ugly, perfect. If it looks good and it's
12:06pretty, but it's [ __ ] useless, you
12:08have wasted your time. Okay. Now, my
12:10first solution here, um, this is giving
12:12me stroke looking at this. uh uploads an
12:16X-ray, gets the AI detections from the
12:18models, processes the conditions,
12:20treatments, and urgency with um GPT
12:22steps, and then creates and sends the
12:24PDF treatment report to the patient.
12:27Now, funny story about this. Um I had
12:29just sold my marketing agency at the
12:30time. We just got the AI model trained,
12:32and I was super pumped and excited to
12:34start working on this shiny object
12:36syndrome in the worst way. I basically
12:38put in two backto-back 16-hour days
12:40building this, working myself to death,
12:42and on the end of the final day, I
12:44actually came down with a fever. Now,
12:45for anyone who's had a fever dream, um,
12:48you'll understand how awful this is, but
12:49I was basically in here, you know,
12:51terrified of make.com, you know, errors
12:53and broken JSON and disconnected modules
12:55and stuff. Um, it was, uh, yeah, any any
12:58founders's worst nightmare, basically.
13:00So, uh I guess the moral of the story
13:01here is take it one step at a time. Only
13:04build what you need to and um get your 9
13:07hours.
13:07>> But it is this just goes to show like
13:10when you find something that you're
13:11really interested in and you know like
13:14you can put in the you can put in the
13:15work and I think that's what a lot of
13:17people are like lowkey craving is
13:19something that they're so excited for
13:20and has so much potential that they they
13:23bang out 16our days and they and they're
13:24probably still smiling at the end of it.
13:26>> Yeah. Exactly. And for me it was like I
13:28can build a medical grade dental sass
13:30like what like and I just jumped
13:32straight into it and worked my ass off.
13:33So yeah, you're exactly right.
13:34>> Your learning of of make.com that was
13:37just YouTube tutorials and and tooling
13:39around like that and then solving some
13:40of your own problems.
13:41>> Yeah, pretty much. Um to be honest, I
13:43got most of my experience like I kind of
13:45learned in the process. I'd tinkered
13:47with voice agents, done some small
13:49things like that before, but most of my
13:50learning from make actually came from
13:52having a crack at that and and building
13:54it out. I was kind of learning as I was
13:55going along to be honest. So now is time
13:58to validate the product. So this means
14:01um get out of your cave, get out of your
14:02dungeon, go in and talk to some
14:03professionals. Okay, validation does not
14:05happen on chat GPT. Now this is where
14:08what I said before about having that um
14:10industry friend. Keeping in touch with
14:12them, nurture that relationship because
14:14here and it's going to continue to be
14:16super useful. Now most beginners think
14:19that validation looks like this. This is
14:21what I like to call the Greg approach.
14:22Now, Greg, he spends two weeks making a
14:24pitch deck. He validates his SAS with
14:26chat GPT. He's always in the mindset of
14:28adding one more feature before actually
14:30talking to anyone, and he takes ages to
14:32get any useful feedback. Okay? You don't
14:33want to be Greg. Now, this is how
14:35winners do it. And this is with the chat
14:37approach. Now, I know it's a it's a bit
14:38abstract, but this is what the kids are
14:40doing these days, so uh I got to ride
14:41the wave here. But Chad, he got super
14:44super clear on the industry pain points
14:46that his SAS actually solves. He had
14:48just enough substance to get proper
14:50useful feedback. He spent the morning
14:52preparing, afternoon doing, cold called
14:53cold called cold called cold called cold
14:53called cold called cold called cold
14:53called cold called cold called cold
14:54called cold called cold called cold
14:54called cold called cold called cold
14:54called cold called cold called cold
14:54called cold cold walked in Chad was
14:55oozing charisma now personally I took
14:57the chat approach not because I am a
14:59Chad unfortunately I'm not as handsome
15:01as this gentleman here but because in
15:03the past I have taken the Greg approach
15:05now Liam I'm sure you're the same most
15:06people are when they're starting their
15:07first businesses they take the Greg
15:09approach so you guys can be you guys can
15:11be wise and learn from the mistakes of
15:12others instead of starting with the Greg
15:15approach okay so here I've actually I
15:18started with cold calling my dentist and
15:19to be honest that was such an awkward
15:20Ward conversation trying to explain to
15:22the receptionist. I wanted to make an
15:24appointment, but I didn't want to get my
15:26teeth cleaned. I wanted to talk to the
15:28owner about my new software startup.
15:29Took a while to her to get the memo
15:31about that, but uh friend of a friend as
15:33well, dad's college mates, like
15:34literally any professionals, um I could,
15:36you know, get in touch with and set up
15:38meetings.
15:38>> What was the promise? Was it like, hey,
15:41you're not really going to get anything
15:42out of this, but can I just talk to you
15:44or like how did you frame that?
15:46>> Yeah, pretty much. So I I know it's a
15:48bit of a scary thing to do, but I
15:50genuinely think that if you call people
15:52and say something to the extent of,
15:54"Hey, I'm building a software product. I
15:56would love 5 to 10 minutes of your time
15:58to get some feedback on if my product
16:00solves X problem for your industry."
16:02Now, the success of this actually
16:05working and setting up a meeting or just
16:07getting treated like every other cold
16:08caller does and being told to basically
16:10get lost is going to be is your problem
16:13actually something that resonates with
16:14this person? because you can have all of
16:16the Chad charisma that you want and all
16:19of the sort of going with the approach
16:20of market feedback and testing things
16:21and blah blah blah, but if that's not a
16:23problem that actually resonates with
16:25them, they're going to, as I said, treat
16:26you like every other cold caller.
16:28>> So, you're kind of like subtly
16:30flattering them, but also like peing a
16:32curiosity around like this person's
16:35called me and said that he's interested
16:36in just like hearing my thoughts and
16:39from my position of expertise on this
16:42thing. And it's also appears to be
16:44something that I struggle with right
16:45now. So maybe [ __ ] like what's the worst
16:47that can happen? I take this call.
16:49>> Exactly. If you're a business owner
16:50who's interested in what genative AI can
16:52do for your business, you can get in
16:53touch with me and my team at Morningai
16:55in one of the links in the description
16:56below and we can start your entire AI
16:58transformation process going all the way
17:00from the education and training of your
17:01staff to the identification of the best
17:03AI use cases for your company all the
17:05way through to development and beyond.
17:06We've worked with some of the world's
17:07biggest sports teams and also publicly
17:09traded companies. So rest assured you
17:11are in good hands.
17:12>> Now for me at least in dentistry and you
17:14could probably extrapolate this out into
17:16your own market in dentistry since the
17:19beginning of the technological age in
17:21like actually implemented in dentistry
17:23the technical and the software aspects
17:25have been dominated by these prehistoric
17:27slowmoving corporate giants. So for
17:29someone like a young person, you know,
17:31who speaks well, well presented, for
17:32them to come in and kind of have a bit
17:34of an an innovative approach, um that's
17:36another thing that's actually going to
17:37be very attractive to most um owners in
17:40most industries anyway. So another thing
17:42to keep in mind there. Now, if you do
17:44get desperate, you don't have enough
17:44warm network professionals to get by,
17:47literally Google search, dentists,
17:49chiropractors, whoever near me, cold,
17:51call them, set up the meeting. Right? I
17:53know it's a little scary. the people the
17:55archetype is going to be drawn toward
17:56starting a SAS probably isn't going to
17:58be too excited about doing this but uh
18:01it's what it takes. So step four is time
18:03to build the front end. So now you've
18:04actually validated that your output
18:06solves the front end. You're going to
18:07build the front end. Now for those of
18:08you who don't know the front end is
18:10basically the interface that users
18:12interact with buttons, screens, forms,
18:14things of that nature. So we're going to
18:16head into our old friend chat GPT or
18:18your preferred LLM to do some
18:20brainstorming. So, we need to figure out
18:22the specific UI sections that you
18:24actually need. So, I'm talking um do you
18:27need a upload, a processing, a results,
18:30a dashboard, a settings, all of those
18:31things. You need to get clear on what
18:33components you need. And then you need
18:34to create a lovable UI generation
18:36prompt. Now, I know you're an enthusiast
18:38of uh Lovable, Liam. Is that right?
18:40>> I've used Lovable and Bolt here and
18:42there and like we've done a course on on
18:44Bolt. Um but because it's my team
18:46actually uses a lot more than me. Um, we
18:48do it for like rapid MVPs to show
18:50clients like on when we're doing our
18:52inerson audits, we'll build literally
18:53build a prototype right in in front of
18:55them and just show them what it would
18:56look like. So, we use it more for like
18:57visualizations, but I'm definitely more
18:59in the uh in the cloud code and and uh
19:02and VS Code sort of arena now.
19:04>> Gotcha. Yeah. Well, it's an awesome
19:05platform. For those who don't know, it
19:07basically vibe codes uh VI codes the
19:09front end for you, the UI. So once
19:11you've got clear on the actual
19:12interfaces that you need, you're going
19:14to get chat GBT to create a lovable UI
19:16generation prompt for my SAS dashboard,
19:18my settings, my whatever it is. So now
19:20you're going to have a separate prompt
19:22for each UI for me. As I said, upload an
19:24X-ray AI results report view and send
19:26dashboard. Pretty simple stuff. Now
19:28you're going to paste all of these
19:29individual prompts into Lovable. And bit
19:33of a pro tip here. You get five free
19:35credits per day with Lovable. Meaning
19:38you can generate five UA screens for
19:40free in one day. If you were to tell any
19:43developer that 5 years ago, they'd lose
19:45their mind. I know I sound like a broken
19:46record here, but it's truly insane
19:47stuff. Now, once you've got those front
19:48ends generated, click the code button,
19:51click download, and you will now have
19:52the downloaded code for each front-end
19:54component on your computer. Now, quick
19:56pro tip here. If it gives you something
19:58slightly off, don't worry about it. You
20:00can fix it later. things, spacing,
20:01alignment, colors, anything that's
20:03drastically wrong, obviously regenerate.
20:05So, this was my UI here. Um, nothing too
20:07crazy. So, that is the AI detections
20:10page. And I generated this um ages ago
20:14when lovable was a lot worse basically.
20:16So, this is okay, but you guys could
20:18definitely generate something much much
20:19better. Uh, now moving on to step number
20:22five, the fun stuff. So, you have two
20:24things now. You have your output from
20:26step two, the back end, kind of parts of
20:28the back end, and the front end. So,
20:30it's time to stitch these two components
20:32together to vibe code. Now, before we
20:34get into the vibe coding juicy stuff, I
20:37do want to talk about the Dunning
20:38Krueger effect because coding, the
20:41experience of coding hasn't really
20:42changed, but the emotional roller
20:44coaster of it has been compressed into
20:46higher highs, lower lows, and into a
20:48shorter time period. So, for those of
20:50you familiar with the Dunning Crruder
20:52effect, you got confidence on one scale
20:54and competence on the other. So when you
20:56start vibe coding, you actually you
20:58prompt and then you deploy it and you
21:00see the feature has been added. You are
21:02going to instantly scale mount stupid
21:04peak confidence. Okay? You're going to
21:05be telling your homie, "Bro, I'm
21:07basically a full stack engineer now."
21:09But then something is going to break.
21:11It's called the value of despair.
21:12Probably going to be looking like this
21:13right here. And you might think to
21:15yourself, "This SAS thing isn't for me.
21:17This is too hard." Then if you thug it
21:19out and you stick through it and you
21:21have a little bit of persistence, you
21:23start to get a little bit more
21:24comfortable. You'll fix some bugs,
21:25understand the folder structure, and you
21:27will slowly start climbing here.
21:29Increase confidence, increase confidence
21:30until you arrive at this point here.
21:33Curs is going to feel like your
21:34co-founder, and all of the pieces are
21:36going to start falling into place. Now,
21:38I know it seems like a bit of a weird
21:39thing to take the time to explain this
21:41to you guys, but it is very important
21:42for you to understand the emotional
21:44journey that is ahead because if you
21:46don't know and you aren't optimistic
21:48that things will eventually work out,
21:50you are likely going to quit here and
21:51end up becoming this guy. So just
21:53understand that. Now, vibe coding for
21:55beginners, the juicy stuff you've all
21:56been waiting for. So this is the tech
21:58stack that we're going to use. Now, as I
22:00go through these platforms, there are
22:01alternatives for each of them. I'm just
22:03going to give you the recommendation
22:04based on my own personal experience
22:06here. So first one here is cursor. Now
22:09what cursor is, as I said before, is the
22:11interface on your computer that allows
22:13you to leverage these different models
22:15to actually edit the coding files on
22:16your computer. Now this is not the
22:17model. Um the models you will still be
22:19able to select, you know, anthropic or
22:21open AAI's models. Now, two quick tips
22:23here. Number one is keep an eye out for
22:25new model promotion week. So, what's
22:26happened over the last month or so as
22:28Opus 4.5 and GPT 5.1 codecs have come
22:31out, they actually let you use these new
22:33models for free. So, this isn't just a
22:35one-off thing that happens rarely. Like
22:3760 to 70% of the time you are going to
22:40be using the new, which of course is the
22:42best most competent models completely
22:44for free. So, cursor is already cheap,
22:46but this makes it that much cheaper.
22:48Also, use the dictate feature, this one
22:50here, for faster prompting. Okay, moving
22:52on to GitHub. Now, GitHub, you can kind
22:54of think of it as the storage for your
22:55code. So, you write, edit the code with
22:57cursor on your computer and then you
22:59send it to GitHub to be stored. Okay,
23:01two pro tips here. Create a private
23:03repository on day one and connect it to
23:05your cursor project. Now, that
23:06connection is going to be made much
23:08easier if you do this step here, which
23:10is before you create your cursor
23:12account, create your GitHub account and
23:14then when you're creating this one, sign
23:15in with GitHub. That'll make the
23:17connection so much e easier and save you
23:19a lot of headaches. Yeah, GitHub can be
23:21pretty confusing for the for the
23:23beginner and it it took me a very long
23:24time to properly wrap my head around
23:27like all the different things uh in
23:29there. So, uh I would definitely uh make
23:32sure you've got that signed in. And chat
23:34between your best friend and
23:35understanding what each of the different
23:36comm command means um because it can get
23:38can get very difficult for a non I'd say
23:41this is one of the more difficult parts
23:42of getting into it's like understanding
23:45how to how the the terminal works, how
23:46like you're installing packages and
23:48things like this. It's it's al it's more
23:50so like the environment of setting up uh
23:52your computer in the coding environment
23:53properly which I'm sure cursor helps out
23:55with a lot now but um that has
23:57traditionally been like a very sticky
23:59point for beginners getting into
24:00alongside GitHub.
24:01>> Yeah, absolutely. And if you're
24:02struggling there just know you're
24:03definitely in the valley of despair.
24:05Just push through. So number three is
24:07render. So this is where you actually
24:09deploy your code. This is actually the
24:11sort of makes it visitable by users.
24:12Gives them a URL to interact with. And
24:15my pro tip here is the backend logs will
24:17be your best friend for troubleshooting.
24:18So you can copy and paste them into
24:20cursor. I'm actually going to give you
24:21guys a hack on how to do this
24:22automatically um in a second. And then
24:25finally is uh Superbase. So Superbase is
24:28a data storage platform. So it will
24:30store user information, login details,
24:32things of that nature and it will talk
24:34to your back end. Now you also don't
24:36have to do anything in Superbase. So
24:38because cursor knows the exact storage
24:41structure that you need in this
24:42platform. You can ask it and some models
24:44will do this automatically uh to
24:45generate an SQL uh file for you. SQL is
24:48just a data language. You can then paste
24:50that in superbase and it is basically
24:51like a prompt to superbase. So it will
24:54generate all the tables and columns and
24:56stuff that you need for you. Makes
24:57things super super easy. So this is the
25:00tech stack. Now as I said this is this
25:02is the perceived barrier to entry. The
25:04barriers to entry have slowly slowly
25:06been dropping. They've now fallen off a
25:08cliff and this is kind of what remains.
25:09people like okay I want to build a SAS I
25:11want to get an application what do I
25:13actually need I don't understand the
25:14components or or what goes into it so I
25:16promise if you guys just create a
25:18account with these four platforms you
25:20connect all four of them that is
25:22basically as technically challenging as
25:24it's going to get for the most part
25:25there's obviously going to be hiccups
25:26along the way but getting started and
25:28getting over this original hump is where
25:30the biggest headache likely comes from
25:32I've just sold that for you now this is
25:34all less than 100 bucks a month I'm not
25:36going to go through the indiv individual
25:37prices of these but like this is
25:39insanely cheap. And again, there's
25:40things like the promotion models that
25:42make this even cheaper. So, all of that
25:44to say, uh there's really no excuses. It
25:46is unbelievably accessible. And then
25:48finally, for my pro tip on fixing bugs.
25:51So, normally the process for fixing this
25:53is you go into render your backend logs,
25:55copy and paste the errors um from the
25:57logs here, and then paste them into
25:59cursor and say, "Fix this." Now, what
26:01you can actually do is use render API to
26:03fetch the latest logs for you. So, you
26:05don't have to copy and paste them. save
26:07those logs somewhere in a file and
26:09basically tell cursor, hey, when this
26:10script runs, um, read the error and fix
26:12the issue. Now, for those of you who are
26:14beginners, not familiar with this stuff,
26:15probably sounds like I just spoke a
26:17foreign language to you, but my
26:18recommendation would just be screenshot
26:20this and then when you come up to this
26:21point and you're confused, uh, put this
26:23in chat JPT or something, get it to
26:25explain this to you and run you through
26:27how to set it up. So, that takes me back
26:29to where we started, which was the big
26:31bang of vibe coding. So again, this the
26:35barriers to entry have been slowly
26:37slowly slowly dropping and they've just
26:39fallen off a cliff. This is the first
26:41time that a complete beginner with no
26:43coding experience, not much money, no
26:45developer can make an impact. You can
26:47build something significant. So I
26:50encourage all of you to don't be Greg.
26:53Take the chat approach and really apply
26:54yourself and go and build something
26:55incredible.
26:56>> Yep. Couldn't agree more, man. And thank
26:58you for breaking that down. I think
26:59that's going to demystify a lot of the
27:01process for the for the beginner and
27:03you've communicated it really clearly
27:05for uh for people as well. So if you
27:07guys were skeptical before um what's
27:09that that quote that I really like
27:10recently uh pessimists sound smart
27:13optimists make money and uh this is
27:15truly one of those eras where the
27:17pessimist about vibe coding and about
27:19like being able to build software as a
27:21nontechnical person. You may sound smart
27:23now but there's absolutely no way that
27:25you're going to make any money off of
27:26that that position. And there's people
27:28like yourself out there, and like I
27:29said, many, many, many more who are
27:31following the same path and going, "Fuck
27:33it." I back myself to be intelligent
27:35enough to figure this out. And I tell
27:37you, if you're watching this channel,
27:38you probably are, cuz uh I've I've met
27:41so many of you people in person now.
27:42I've met people here through these
27:44podcasts, and I'm always blown away by
27:46the the quality of people who are
27:47watching these videos. So, if you're
27:48watching this, take uh permission from
27:51me and permission from Ben to [ __ ]
27:53shoot on this stuff next year, guys.
27:54Like there is you will regret this to
27:58the freaking day that you die if you
27:59don't do something about this now. And
28:01it's the decisions that are you're
28:02probably like stressed about it and
28:04stuck on the fence. If you can just make
28:05a decision and say I promise you like it
28:08is going to be easier than you think.
28:09Like for for example for me getting back
28:11into the engineering recently getting
28:12into learning this like a gentech
28:14engineering workflow and claude code and
28:16how to like it felt out of depth even
28:18for me and within literally like two
28:21days I'm like oh okay I kind of get this
28:23and you start getting the like
28:26essentially the the dopamine of like the
28:27the small wins and once you're getting
28:29small wins you are on the way and the
28:32way that uh Ben's laid it out here is
28:34exactly how you should be approaching it
28:35as a beginner. So really appreciate your
28:37time, man. And uh I'm so happy to to
28:39hear about your success and I'm sure
28:40you're going to go from strength to
28:41strength. Um is there anything else
28:43you'd like to to wrap up with then? Um
28:45how can people get in touch with you if
28:46they want to?
28:47>> Yeah, so school.comarketingmastery.
28:50So I've got um vibe coding stuff um AI
28:53lead gen models, ton of ton of just AI
28:54general content in there. So um if
28:56you're interested in connecting with me,
28:58um shoot me a message on there.
28:59>> Sweet, Ben, mate. It's been great to
29:01meet you.
29:02>> Pleasure.
29:02>> Congrats on your success and uh
29:03hopefully I'll be back on here soon.
29:05>> Yeah. Happy new year to you, you land.
29:06Have a good one.
29:07>> Likewise. So, that is all for this
29:08episode of the podcast, guys. If you
29:09want to see something similar that I
29:11really think you'd like, you can click
29:12up here to watch another one. And
29:13remember, if you think you have a story
29:14worth telling and some valuable insight
29:16you can share with the community, you
29:17can fill out my podcast application form
29:19in the description below. I'd love to
29:20have a chat with you and get some
29:21exposure for your business. Aside from
29:23that, guys, that's all for the video.
29:24Thank you so much for watching and I'll
29:25see you in the next

About this transcript

This page contains the full transcript of How I Built a $100K AI SaaS Without Knowing How to Code (5 Easy Steps) by Liam Ottley, generated from the public captions YouTube serves with the video. The transcript has 6,811 words across 959 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.