Background
I started a new project not too long ago. I've briefly mentioned it in other posts; it's a mobile-first social app for automotive enthusiasts. The stack is Node/TypeScript/GraphQL/Supabase on the backend, and React Native/TypeScript/Expo/GraphQL/Supabase on the frontend. This is a personal project that I've wanted to do for nearly 20 years, and I'm finally making the time for it.
I had initially considered it a hobby project, and another opportunity to explore vibe coding more deeply. But my viewpoint has shifted. This has gone from being a hobby project, to something that I want to use myself, and hope others find useful. It's something I want to launch.
Repo is Private
I don't realistically believe this will turn into a money maker for me. But it's not beyond the realm of possibility. Considering this project may have some commercial viability, I decided to make the repo private. The primary justification is to protect any potential intellectual property.
There are some downsides to the repo being private. One is obviously personal marketability/visibility; I have little public presence on github, and if this repo were public it would be something I could share with confidence, enthusiasm, and pride. Also, I'm between jobs at the moment. It would be hugely beneficial if I could include this as part of my portfolio, but I've still chosen to keep the repo private.
Lack of AI Assisted Governance
The biggest downside, however, is lack of overall context from AI assistants. I talked to ChatGPT for at least a month before I wrote the first line of code. I've used it to help me through some technical decisions, how to model certain interactions/activities, to explaining monetization options and SaaS offerings in spaces where I have little experience. ChatGPT has my complete thought history of this project from inception to its current form.
It is a frequent occurrence that I will ask ChatGPT for some seed structure (sql staetments, graphql changes) based on our chats, but what it generates is not directly consumable because it's missing the underlying source. Since the repo is private, ChatGPT can't actually see the source code. So it's generating content that doesn't use the same symbols as in the code.
Likewise, I would like to ask ChatGPT to analyze certain slices of the repo, and give me an updated entity diagram, or tell me where what I've implemented isn't in agreement with a previously discussed design. That's not possible, because the repo is private.
Cursor & Limited Context Windows
On the other hand, Cursor has access to the entire repo (or individual folders for different tiers, in my case). From a code standpoint, it has all of the context ... of what's been implemented. It is missing all of the context of the historical discussion on the project. When I implement a given feature, I find that I have to repeat portions of previous discussions; that I have to explain certain parts of the domain model again, for example.
To put it another way, I can deploy Cursor tactically, but would like to leverage an LLM on the codebase strategically. And since the repo is private, I cannot.
Is a Public Repo Worth The Risk?
I am extremely tempted to temporarily make the repo public, and share it with ChatGPT (or Claude). That would give me a huge step up. But that comes with risk. Sure, the vast majority of repos totally escape notice. But once it's public, it's going to be crawled by some bot, and that can never be deleted. Knowledge of the domain model, the physical database structure, backend providers, etc, are all possible attack vectors even if I changed the repo back to private.
I don't have many readers, but I would genuinely appreciate some feedback. Have you successfully launched a product, your own intellectual property, from a public repo that *was not an open source project?
Where Do You Keep Your Holistic Context?
For others that are leaning heavily into AI assistants outside of the code, where is your full historical context? Are you leaving it in an LLM? Do you keep it in some kind of knowledge repo? Do you document things within your repo itself? If so and your repo is private, how are you bridging that gap?
Top comments (1)