I Used AI to Help Build My Open Source Project and People Lost Their Minds — Here's Why They're Wrong
Source: Dev.to
My Experience with AI‑Assisted Development
Last week I posted my first open‑source project on Reddit. Within hours, someone accused me of writing “AI slop.” Another said my reasoning was “stupid.” A third said they were considering quitting Reddit altogether because of people like me. All because I used AI in my workflow.
The Project: ShopCore
I’m a solo developer from India. I spent months building ShopCore – a dual‑purpose e‑commerce and subscription‑delivery platform designed for the Indian market.
The problem I was solving is real
Daily essential services (milk, newspaper, water jars, bread delivery) still run on handwritten khatas, verbal agreements, and pure trust. There is no app, no tracking, no proper billing. Vendors lose money, customers have no receipts, and everything runs on paper.
ShopCore is an open backend infrastructure – think of it like UPI. NPCI built the open infrastructure, then Google Pay, PhonePe, and Paytm built businesses on top of it. ShopCore works the same way – any developer or startup can take the backend, build a simple frontend, and offer it to local vendors.
Scale of the MVP
- 120+ REST API endpoints
- 40+ entities, 30+ database tables
- Razorpay + Stripe payment integration
- Multi‑vendor with location‑based discovery
- Flexible subscription engine (daily delivery, monthly billing)
- Optional deposit and smart‑settlement system
- Built with ASP.NET Core 10, Entity Framework Core, MediatR, FluentValidation, SQL Server
I built all of this alone, without funding, and open‑sourced it for free.
How I Used AI
| What AI helped with | What AI did not do |
|---|---|
| • Writing the Reddit post – I had the idea and story; AI helped structure it clearly. • Drafting variable names – boring, repetitive work that AI handles in seconds. • Rubber‑duck debugging at 2 am – when no one else was around to talk through a problem. • Boilerplate code – scaffolding I would have written anyway, just faster. | • Identify the gap in India’s daily‑delivery market. • Design the dual‑purpose architecture. • Figure out how the deposit‑settlement system should work. • Decide to use MediatR and the CQRS pattern. • Choose to open‑source it instead of letting it sit on my hard drive. • Write 120+ API endpoints with actual business logic. |
The idea, problem, architecture, and judgment calls were all mine. AI merely accelerated the work.
Tools We Use Every Day (Without Thinking)
- IDE autocomplete – IntelliSense, Copilot suggestions, tab completion.
- Stack Overflow – You copy‑paste solutions you didn’t invent.
- Documentation & tutorials – Nobody memorizes everything.
- Linters & formatters – Automatically improve code quality.
- Google – Outsourcing memory to a search engine.
The only difference with modern AI is that it’s more capable and more visible. The principle is the same: use tools to move faster so you can focus on what truly requires judgment.
When AI Becomes a Problem
If you generate an entire codebase you don’t understand and ship it as your own without review, that’s a problem – not because you used AI, but because you can’t maintain or defend the code.
Conversely, using AI to speed up work you fully understand and own is simply being a smart developer.
My Takeaway
- The comment “such a long response, probably AI‑written” was partially right, but it doesn’t invalidate the 120+ API endpoints I designed.
- The criticism about post clarity was valid; I updated the post based on it.
- A blanket “AI = bad” argument is a feeling, not a technical opinion.
If you’re building something alone, without a team, funding, or a reviewer, use every tool available:
- Use AI
- Use Stack Overflow
- Use documentation
- Use shortcuts that let you spend more time on thinking, judgment, problem identification, and architecture.
The goal is to ship something valuable, not to prove you can suffer through building it the hardest possible way.
Open‑Source Availability
I open‑sourced ShopCore because I don’t have the capital to keep it private.
Maybe someone will build on top of it. Maybe it’ll help a developer skip months of backend work. Maybe it’ll eventually reach the vendors it was designed for.
ShopCore is open source at https://github.com/your-repo/ShopCore. Stars, issues, and PRs are very welcome.
If you’re a developer building for the Indian market or interested in subscription‑based delivery systems, I’d love to connect.