I built an entire SaaS to prove that Cursor is the only engineer you need

For context, I’m a Product Designer by trade, but I’m always enjoyed getting into the mechanics of product building.

After seeing all this cursor talk back in October, I decided to take the plunge and built an entire saas product by myself with AI generated code

Was a very fun experience that really made me realise that a combination of the University of YouTube + AI will get you very very far in making cool things, and eventually cool things that make you money online

Full article here
https://medium.com/@sakkyb/how-i-built-a-saas-product-with-100-ai-generated-code-5728e0c97c8d

---

Comments

rkaw92 • 41 points • 2025-03-05

I built vehicleexpirytracker.com without any engineers
[…]
If I couldn’t crack the issue with Cursor in 4 goes, I’d drop a note to one of my many engineering friends

So… two weeks with two different AIs, supported by several experienced software engineer friends at no cost? To build a reminder for car insurance that you have to pay for? Yeah, sounds realistic.

I’ll be reserving my judgement on Cursor being “the only engineer you need”, though.

eyyyyminor • -40 points • 2025-03-05

I did also state that I could solve it with cursor if I really wanted to, but the engineer friends were faster on certain parts.

And to start with yes it is the only engineer you need. But the key is - to start with.

Me asking my engineer friends is no different to you checking stack overflow or Reddit to ask strangers that are engineers. Sometimes it’s just a bit faster.

digitalwankster • 22 points • 2025-03-05

”Me asking my engineer friends is no different to you checking stack overflow or Reddit to ask strangers that are engineers”

So, still requiring engineers other than Cursor which is the exact opposite of your post title.

trevorthewebdev • 8 points • 2025-03-05

you bet starting paying those engineer friends then … this whole post is shit. To start with? Yeah it’s fucking easy to start a coding project. What is hard is to finish it and take it to market or anything like that.

le_christmas • 2 points • 2025-03-05

LOL OOP needs a $20/month subscription and an entire freaking AI just to run create-react-app 🤣

trevorthewebdev • 2 points • 2025-03-05

not even going to vite, rookie movie

edit: thought i would add that I was spending too much time reading your message trying to think of why object orientated programming would need money here, then got it … haha, I’m so burnt out man

le_christmas • 2 points • 2025-03-05

Oh 100% guaranteed it runs `create-react-app` and then suggests ejecting instead of using a tool that generates actually good build infra

3141521 • 1 points • 2025-03-05

I just want to say your comment was hilarious and my morning better haha

LayerComprehensive21 • 1 points • 2025-03-05

Yeah lmao the functionality is so basic I could implement this in any calendar app 🤣

besneprasiatko • 18 points • 2025-03-05

Good for you. I have the exact opposite experience. I fell for posts about game-changer updates and the newest Claude model, and bought PRO. I gave Cursor task to create django app, create some cron jobs, set up backups, and to dockerize everything. After 4 days of endless fixes, it still doesn’t work. I used Claude 3.7 with reasoning. I tried to use specific cursorrules to improve code quality, but it didn’t help much. I would be much better if I just cloned some boilerplate and made it run in 15 minutes.

TroubledEmo • 3 points • 2025-03-05

Feel you. I thought I could rely on Cursor/Windsurf/Copilot for the TUI of a small CLI application I wanted to do as I suck with graphical interfaces. Yeah. I took days and they always somehow destroyed their own work.

So I ditched the TUI and just started doing a library which I could use later if I want to do the TUI or other could use the library if they want to do stuff like this. Either way it seems like doing everything by my own would have saved money and… nerves.

FrostBerserk • 3 points • 2025-03-05

The other thing people dont understand yet is prompt creation to get the AI to do what you want.

You need to be able to take complex tasks, break it down like you’re talking to a 5 year old and then you need to break that up in prompts, slowly building it.

People want to type in diatribes and think it’ll just work.

Folks need to learn how to use the AI before complaining it “doesn’t work”. Which what does that even mean?

besneprasiatko • 1 points • 2025-03-05

After 4 days and explaining the task in hundred different ways, generated code still contains trivial logical errors. I tolerated it just from some morbid curiosity and to see its performance. Yes, I second guessed myself, implemented cursorrules, break tasks in smaller part, but after all this I can sincerely say it just sucks. I will try some different tasks, but I dont think I will extend my subscription.

FrostBerserk • 1 points • 2025-03-06

You probably needed to start over or there’s something missing in how you’re doing it.

What you’re essentially saying is “I tried to use this paint sprayer but the nozzle got jammed, the paint spilled out of the container and when I sprayed it was all uneven. This thing sucks! It didn’t pain the house by itself!”

It’s operator error and it’s still leaps and bounds ahead of where it was months ago.

In it’s current state, it can allow for total laymen to program in ways they never could before.

It is far from “just sucks”.

This is more of operator error than anything else.

Though if you aren’t able to use it, then you should either figure out how to use it or just give up.

I always found out when I give up, things usually just magically happen for me in life.

Why bother trying ya know?

besneprasiatko • 1 points • 2025-03-06

No, it just sucks and models are underperforming. Without Cursor I was able to get much better results with OpenAI o1, even if it didn’t understand the whole context. I will probably return to Copilot. As a senior dev I don’t give up on coding, just on Cursor.

eyyyyminor • -3 points • 2025-03-05

How much composer did you use for it. I’ve found asking it to clarifying things before going off on one is key to getting it to build in the right way.

Also asking another AI when Cursor runs into trouble helped.

Turbulent-Dance3867 • 6 points • 2025-03-05

Sooo… You didn’t actually prove anything

digitalwankster • 6 points • 2025-03-05

They literally proved the opposite of the post title lmao

eyyyyminor • 1 points • 2025-03-05

Sorry I meant when Claude runs into trouble (my default model for cursor) I ask another AI model

infinished • 5 points • 2025-03-05

You never used git ?

eyyyyminor • 3 points • 2025-03-05

Not before I built this not really

V4UncleRicosVan • 3 points • 2025-03-05

Have an up vote OP, not sure why someone would downvote you for learning to use Git.

EstonBeg • 5 points • 2025-03-05

What’s the plan when you find a bug you can’t fix? What’s the plan when you find a security vulnerability? What’s the plan when you have to scale the system to fit more users?

You don’t use PayPal, you have your own little modal to type the the credit card information. I don’t trust that this is secure.

I set up my company name on your site as \“. You have 0 fronted or backend validation, I hope u didn’t store that in mongo as json. I’m not that hopeful about that because I couldn’t log in for a second time.

You don’t verify customer ids on the payment screen, iI can change the customer ref and it doesn’t complain at all. It even let me use a string for the key.

I don’t trust AI to build anything secure ever. I am not a security guy at all and in 5 minutes I found two big problems in your site. If you want an actual quality product you have to hire real engineers. 

eyyyyminor • -8 points • 2025-03-05

lol… my own little modal? It’s a Stripe payment link…

And if you’d read my article, you’d see this was never meant to be a B2C product with a full self serve onboarding, it was a test, and like every test you learn about things to improve (even tiny things such as suggesting not using special characters for organisation name)

And… I used supabase not mongo

Thanks for the engagement though, I hope you channel your energy into something more useful to your life in the future:)

CaptainCactus124 • 10 points • 2025-03-05

OP the title of your article says that cursor is the only engineer you ever need. I think it’s completely understandable to get this sort of response with such bull shit like that.

EstonBeg • 3 points • 2025-03-05

It wasn’t about supabase not mongo.  It was about your db got broken from my phone. This energy channelling was me bored on my phone. If any real security engineer looked at your site it probably wouldn’t stay up for long. 

loyalekoinu88 • 3 points • 2025-03-05

What about security? What have you implemented to curb abuse, etc? Scaling?

eyyyyminor • 2 points • 2025-03-05

We can get to that when we have some humans paying money for this shit!

loyalekoinu88 • 7 points • 2025-03-05

You think that security is an option? Yikes.

V4UncleRicosVan • 3 points • 2025-03-05

Hey man, congrats on your app, it’s exciting to see your progress and it’s exciting to see what’s possible for folks without a dev background. Pretty cool for 2 weeks of work. Ignore the haters, keep building, and remember that this is the least powerful these models will be going forward.

West-Code4642 • 7 points • 2025-03-05

congrats for shipping and solving a real business need and the nice blog post

i think being able to write good PRDs is a skill many engineers lack, i’m going to have to experiment with that. i’ve noticed AI tools do respond well when requirements are written up well from the get go

eyyyyminor • 1 points • 2025-03-05

Yeah, having designed products for many years PRD writing is one of the things that comes easier to me now

Drifting_Grifter • 2 points • 2025-03-05

Now come back when you need to keep adding features

elrosegod • 2 points • 2025-03-05

I’ll get back to you guys on this, i am doing this as well from product/business analyst/pm role with 6 years of self learning. I think my issue is i tried to go cloud agnostic/ local first which is crushing me right now

prophitsmind • 0 points • 2025-03-05

Phenomenal post. Added wayyyy more value than as a whole than just how awesome Cursor is. (complainers will complain before trying a few things out / viewing it as simply a great tool on the market anyway).

eyyyyminor • 0 points • 2025-03-05

Glad you found it valuable!

TheFIREnanceGuy • -3 points • 2025-03-05

I stopped reading as soon as you said product designer. Lol so you build a SaaS with no understanding of coding? Yeah would love to see how you can scale when you don’t know what you dont know.

Cute_Commission2790 • 2 points • 2025-03-05

The only engineer you need might be a little far fetched, but every piece of software doesn’t need to be built for scale and handle million concurrent users. I agree with points on security but even then assuming this was built for friends then a lot of the things that come with software en masse are not necessary

sharpfork • 2 points • 2025-03-05

Having to scale is an excellent problem to have! One might be able to scale a decent chunk by leaning on their chosen hyperscaler instead of optimizing. They may need to bring in a human architect at some point but that isn’t a bad problem either.

eyyyyminor • 1 points • 2025-03-05

Exactly, if it scales then bring in more people who are best placed to execute on what needs to be executed at a high level

charlesmwray • 1 points • 2025-03-05

No one believes this

well_wiz • 1 points • 2025-03-05

Well, I hate to tell you but you just prove that they still need us engineers.

East-Text-266 • 1 points • 2025-03-06

Is it a todo SaaS? Cursor is getting worse everyday,

Big-Funny1807 • 1 points • 2025-03-06

AI Cannot replace us yet, and for those who can’t get a working product with the help of Cursor you are just not doing it right, you need to learn using AI exactly like anything else you learn in your life.