Have you had any experience creating a complete website using AI?

[Starblazer]

wait this is interesting quick q tho - when u say u genrated pages using python code that AI wrote, did u have to manauly tweak that code alot or did it just… work?? asking bc im trying to do smth similar and idk if i should just vibe code it or actually go the step by step route like u did ‍

I actually created code blocks first like js & css files, header, footer, & post template, sitemap, etc. I have separately created a database file in csv format. I then wrote a python code that takes elements from database like headings, content, images, etc. and generates html pages. I had to refine it multiple times until I got the desired output though. But it's an easy process because I'm not writing code myself. Once the site is ready, I had to tweak things for improvements in design & performance.

If you want to create a site or saas, you should first fix core functionality and don't change it during the development. The design can be modified a thousand times without any issues as it's a harmless css file. If possible, write modular code, so that you don't need to change every page. For example, I separated header & footer from post content so that I can modify ads across all posts at once instead of updating hundreds of html files every single time. Such opportunities exist during development and you should keep an eye on them. You should also decide monetization in advance. If you code a php site yourself and then decide to add memberships, it becomes difficult later. Such features require complex code and you may need to upgrade from manual php code to a wordpress site. So, take these big decisions first and keep doing small updates over time.

 

[Lucas10126]

Yes, had to keep correcting mistakes by ai , there were lots of bugs , although I used Claude and GitHub

 

[cosmoaccs]

I actually created code blocks first like js & css files, header, footer, & post template, sitemap, etc. I have separately created a database file in csv format. I then wrote a python code that takes elements from database like headings, content, images, etc. and generates html pages. I had to refine it multiple times until I got the desired output though. But it's an easy process because I'm not writing code myself. Once the site is ready, I had to tweak things for improvements in design & performance.

If you want to create a site or saas, you should first fix core functionality and don't change it during the development. The design can be modified a thousand times without any issues as it's a harmless css file. If possible, write modular code, so that you don't need to change every page. For example, I separated header & footer from post content so that I can modify ads across all posts at once instead of updating hundreds of html files every single time. Such opportunities exist during development and you should keep an eye on them. You should also decide monetization in advance. If you code a php site yourself and then decide to add memberships, it becomes difficult later. Such features require complex code and you may need to upgrade from manual php code to a wordpress site. So, take these big decisions first and keep doing small updates over time.

wait so u basically built ur own lil static site generator with python + csv database instead of just using wordpress from start?? thats kinda genius ngl. also quick q, when u say "fix core functionality first" do u mean like decide monetization/membership stuff even before writing any code at all, or just before u start scaling the site?? bc that part confuzed me a bit also how many times did u actually have to refine the python script b4 it gave decent html output lol, sounds like alot of trial n error

 

[nathanhill]

AI can save a lot of time when building a website. Has anyone here built a full site with AI?

 

[Starblazer]

wait so u basically built ur own lil static site generator with python + csv database instead of just using wordpress from start?? thats kinda genius ngl. also quick q, when u say "fix core functionality first" do u mean like decide monetization/membership stuff even before writing any code at all, or just before u start scaling the site?? bc that part confuzed me a bit also how many times did u actually have to refine the python script b4 it gave decent html output lol, sounds like alot of trial n error

You first need to create a project blue print and technical architecture documents before starting the coding. If you try to build a diesel-engine car and decide to make it into a electric car mid-way, it'll create a lot of confusion to you & the AI tool as well. First decide what's supposed to be a 100% truth during project development. It's absolutely fine if you want to launch a free product and then add monetization at a later stage but you must add it to project blue print so that the foundation is laid off considering all these constraints. It's okay to scale the project at a later stage but it should be adding, removing, replacing, or modifying the parts instead of redesigning the whole project. The core functionality should remain same throughout development and scaling.

I've refined the script 10-15 times before it worked. But it only took me 2-3 days because it's a small project. I've worked on mobile apps that took me 2-3 weeks. AI can't remember whole development, so it'll be difficult to develop projects with hundreds/thousands of lines of code. You need to keep correcting it or uploading files for reference. So, keep a changelog if you're building a huge project. For small projects, you can take decisions while you develop as they don't need much planning.

 

[jeezabull]

Hell Yeah and.. Hell No at the same time. If you are like me and have no idea what it takes it is better to start with a solid product and improve on it. It still takes a ton of time tuning up and I would suggest also getting a security specialist to take a look at the codebase. It definitely saves money on development costs and you can get the features you want the exact way you want it to perform. However it spits out what you put it. If your prompt is slop, the output will be slop too, especially in regards of content and UI.

Don't try to reinvent the wheel. Put your own tire on it.

One thing to note, If one ai gets stuck on a bug try a new chat or a different ai. For example; claude was stuck on a bug and would just go on a spree to find the bug and had not done any improvements on it. I got frustrated and went to codex. It did find out the issue within an hour and solved it within the next hour. Same thing happened in the codex and I went to claude to have it sorted. So going back and forth may help too.