10 days of no-code product MVP development (idea to production)

Using Cursor with Claude 3.7 for development -- Azure for hosting ChatGPT, storage and DB -- Fal.ai APIs for image-to-video models

Mar 20, 2025

10 days ago, I decided to try out a new experiment: Building a product MVP (mostly a technical POC) from scratch without any coding, only using AI tools.

The idea for the experiment is to generate short videos with smart, auto-generated prompts from an image.

The final result is very promising! With a total of approximately 125 prompts, I was able to create the application, test the functionality and deploy it.

Summary of integrations:
(without writing a single line of code)

The MVP version includes all the functionality (except payment):

* Mongo DB integration
* Google OAuth integration for logging in
* Self hosted ChatGPT service on Azure and integration
* Uploading an image from local or with URL
* Image storage on Azure
* Running analysis with ChatGPT on the image
* Integration with Fal.ai APIs (creation of the video & status checks)
* Credit system
* Code redeem system
* Library for generated videos

1-) Ideation & Brain Storming (ChatGPT Plus)

Lately, I am very interested in generative AI models. I started brainstorming with ChatGPT to find an idea related to that for developing a product MVP, without any coding. I explained what I would like to work on in general and also my will to keep the scope limited and simple.

After reviewing different suggestions, I chose to start with a generic image to video application.

2-) Application to the Microsoft for Startups Program

After deciding on the initial plan, I created short descriptions of the product and what the future would look like.

Also, I created a summary of my background and what I want to do now as a solopreneur in detail.

Then I searched for some startup programs to apply and hopefully get some perks and credits for a real, working prototype.

I chose Azure, as they offer tiered sponsorship and they are the only major cloud provide to accept early level individuals who haven’t created their company yet!

I applied to the program with my Linkedin account, together with all the details I created with ChatGPT in the first phase, highlighting the project will be using “AI” specifically!

And, in 24-48 hours, voilà, I got the Microsoft Azure Sponsorship with $1000 credits to use in Azure Cloud!

4-) Choosing the Tech Stack (ChatGPT Plus)

I turned to ChatGPT again, explaining the acceptance of my Azure startup program, asked it to create a tech stack so that I can take advantage of the sponsorship credits as much as possible.

I ended up with these choices:

NextJS for both FE and BE of the application
Azure OpenAI: To host ChatGPT models and use the APIs from the app
Azure Cosmos DB for MongoDB (so that I can switch to plain MongoDB later if I can get sponsorship from them later!)
Azure Storage Service: For storing the images
Github private repo to keep the code
Netlify to build and continuously deploy the app through the Github repo automatically

5-) Web Development with AI (Cursor Pro)

First, I bough the monthly membership for Cursor to be able to use high quality and fast models (500 fast prompts per month).

I started using “Claude-3.7-Sonnet” with the agent mode and was very satisfied with it till the end. For very simple tasks or questions,

I used the free “gpt-4o-mini” model with ask or edit mode, which has unlimited usage. In the agent mode, it does not work well. I also don’t recommend to use it for advanced tasks in the edit mode. Just use it for simple questions and tasks that is not worth to use the premium models.

The initial prompt is important. I tried to explain the purpose and the usage of the website in detail, while getting help from ChatGPT.

The initial codebase was created easily, without a major discrepancy.

For all integration points, before implementing them, I asked for guidance on how to create the service in cloud and and how to configure it. I made it write MD documentations as well, explaining in detail. Surprisingly, this stage works really good as well. It guided me through Azure cloud console menus and specific steps, which was quite nice!

During the whole development process, I might have used around 100 premium prompts. For most of the functionality, I guess I might have used 30 prompts max. The rest is tweaking, fixing bugs, improving stuff etc. That is where you learn how to work with AI during coding!

5-) Deployment to Netlify (free tier)

Once I have tested the happy path locally, I decided to make it public.

I signed up to Netlify and connected my Github account. It immediately started trying to build the project but it failed.

I have a few fixes and tweaks to make it work with Netlify (max 5 prompts) and then finally it was live!

Netlify gave me a subdomain and I was able to see the application there.

After that, whenever I made a change and push it to my Github repo, in 1-2 minutes it was LIVE. A free, out-of-the-box CI/CD system, I loved it!

The Result: DiCoLabs (Digital Content Labs)

TBH, I am very satisfied with the result!

Within just 10 days, without writing a single line of code, it is available live with MVP features.

To make it clear: this is not the final product, it is more like a technical POC for the future product. The POC is a generic image to video generator but the product will be focused on a niche, solving a single problem in an efficient way!