I'm not a "real developer."
I built this whole platform with AI anyway.
Hi, I'm Benjamin. I don't have a computer science degree. For years I thought that meant the things in my head would stay there. Then AI changed the rules, and I built this entire platform, feature by feature, with it.
So I built the room I wished existed, a place for the rest of us: people turning ideas into real products with AI, shipping in the open, without anyone looking down on how we build.
People call it "AI slop." What they miss is that AI hands you nothing. You still bring the idea, the taste, the thousand small decisions. The tool is leverage. What you make with it is yours.
Free forever, no ads. Browse everything without an account.
What is DevConnect?
DevConnect is a free community for people who build real products with AI. One place to share what you're shipping, meet builders who get it, and sell what you make. No gatekeeping about how you built it. Free forever, no ads.
Your projects, your stack, what you've shipped. One link that proves what you can build, not what school you went to.
Post what you're building with AI and get curiosity, not contempt. Real feedback from people building the same way.
Rooms around Cursor, Claude, a model or a stack. Set the rules, run it your way. You own it.
Also here: voice rooms in every community, a marketplace for what you make, remote dev jobs from junior up, and an app-tester exchange for store releases.
A certificate that someone can check
A free track assessed on the work you submit rather than on hours attended. Every certificate carries a number and a public page anyone can look up without an account.
Studying, and want to bring this to your college? The Campus Ambassador role is open now. Part-time, remote and unpaid: experience and a reference rather than a salary.
Talk, don't type
Some things take twenty messages, or two minutes of talking. Every community has voice rooms now, and you can call anyone you both follow — straight from their profile or a chat.
Your mic stays off. Nobody can hear you.
You can speak. Your camera stays off.
Camera allowed, off until you turn it on.
Recent from the feed
Shipped: A guide for the next contributor
Documentation Requirements – Contributor Guide 1. A CONTRIBUTING.md file has been added to the repository. 2. The project structure is explained, including the purpose of the main folders and files. 3. The documentation explains where new features should be added. 4. Guidance is provided on which existing components may need to be updated when adding a feature. 5. Instructions are included for installing dependencies and running the project locally. 6. The commands for building and linting the project are documented. 7. Expected results of successful build and lint checks are explained. 8. Fragile parts of the code are identified, including ingredient calculations, serving controls, accessibility announcements, and responsive layout. 9. The README includes information about the project, its main features, technologies, and how to run it locally. 10. The documentation provides clear guidance to help future contributors understand and extend the project without breaking existing functionality. Repository: https://github.com/humaeidmansur/accessible-recipe.git Live: https://humaeid-internship-reciepe-project.netlify.app/ Built for the Verified Frontend Internship · Documentation
RAG
For a while now, I've been experimenting with Retrieval-Augmented Generation (RAG), exploring how to make LLMs answer questions using information retrieved from actual documents. Today's experiment with Raggy took me through another complete RAG pipeline. Document ingestion: Processed a research paper in PDF format. Text chunking: Split the document into 82 chunks. Vector embeddings: Used nomic-embed-text to convert text into 768-dimensional vector representations. Semantic search: Retrieved the top 5 chunks based on similarity to the user's query. LLM inference: Used a locally running qwen3:4b model to generate an answer from the retrieved context. Retrieval tracing: Inspected chunk rankings, similarity scores, source pages, token usage, and execution timings. Then I asked: “What was the IoT adoption rate across the companies?” Raggy returned 74.4%, grounded in the retrieved document context. But here's where it gets interesting. The trace exposed duplicate chunks in the retrieval results. The generation stage also took significantly longer than embedding and retrieval in this run. These are exactly the kinds of details I want to investigate. Because building RAG isn't simply about connecting a vector database to an LLM. It's about understanding document parsing, chunking strategies, embedding models, vector similarity, top-k retrieval, context construction, grounding, and inference latency. A system can return the right answer and still have room for improvement. My next focus is making retrieval more reliable, reducing redundant context, and improving response time without sacrificing answer quality. Still experimenting. Still debugging. Still learning by building. The goal isn't just to make AI work. It's to understand the engineering behind making it work well. RAG AIEngineering LLM Python MachineLearning softwareengineering
Shipped: The data model, explained
Repository: https://github.com/garima-2006/ticket-sales-api Built for the Verified Backend Internship · Documentation
Questions we answered properly
Shipping, hiring, reviewing AI-written code, finding people to build with: the honest answers are scattered and half of them are out of date. Every source on these pages was fetched and checked before it went live, and nothing goes up that fails that check.
Yes, if the remote role reports to a New York supervisor, office, or work site, New York requires a pay range. If none of those New York connections exist, state law does not cover the posting.
Yes, if the employer is covered by New Jersey’s pay transparency law, remote jobs that do business, employ people, or take applications in New Jersey must include salary and a general benefits description in the posting.
Yes. New personal Google Play developer accounts created after November 13, 2023 must complete a closed test with 12 opted-in testers for 14 continuous days before applying for production access.
Yes. If the remote role reports to a supervisor, office, or worksite in New York, the posting must include pay or a pay range, even when the work is done from home elsewhere.
Project of the Week
Post what you built this week with the showcase tag. Until Wednesday night the community votes and, more importantly, replies: comments count double. The winner takes the featured spot at the top of the feed for three days.
How to build with AI without shipping slop
The difference between a real product and "AI slop" isn't the tool. It's the decisions you make around it. Six things that actually matter:
- Own the idea before you prompt
AI is great at "how," useless at "what" and "why." Know what you're building and who it's for. That part is still yours.
- Read every line it writes
If you can't explain what your code does, you don't own it, you're renting it. Ask the AI to explain until you actually understand.
- Ship small, ship real
One feature that works beats ten that half-work. Slop is what happens when you generate faster than you can check.
- Test it like a user, not a demo
AI writes the happy path. The bugs live in the edge cases you have to go find yourself.
- Keep the taste human
Naming, flow, what to leave out: that's judgment, not generation. It's why two people with the same tools ship completely different things.
- Learn the fundamentals as you go
You don't need a degree, but understand what you keep reaching for. Every "why did that work?" you chase makes the next build more yours.
A small team. A lot of AI. No VC, no roadmap deck.
DevConnect is built and run by me, Benjamin, with a small team who own big parts of it with me. Not a room full of engineers, just a few people who learned to build with AI and refused to wait for permission. If something breaks, you're talking to the person who built it.
"People kept telling me I'm not a real developer. Maybe not. But I built the thing you're looking at, and I built it for everyone who's been told the same. That's who this is for."
Frequently asked questions
Come build in the open.
It's free. It always will be, and there will never be ads.