Inherited an AI-built Next.js/Supabase app - is it usually fixable, or do you end up rebuilding?
Unanswered
Finnish Lapphund posted this in #help-forum
Finnish LapphundOP
Looking for a sanity check from people who've been here before.
We have a Next.js/Vercel app with Supabase and Clerk, built over about two months with heavy AI assistance and no experienced engineer overseeing the work. Around $30k spent, and it's running $1-2k/month in API costs before we've even launched.
We're now deciding whether to keep building on it or start over with someone who actually knows what they're doing.
For those who've inherited an AI-built codebase: in your experience, is the usual outcome that it's fixable, or that you end up rebuilding anyway? And is $1-2k/month pre-launch a normal running cost for something this size, or a red flag on its own?
Not looking to hire anyone here, just trying to get a sense of what's normal before we commission a proper review.
We have a Next.js/Vercel app with Supabase and Clerk, built over about two months with heavy AI assistance and no experienced engineer overseeing the work. Around $30k spent, and it's running $1-2k/month in API costs before we've even launched.
We're now deciding whether to keep building on it or start over with someone who actually knows what they're doing.
For those who've inherited an AI-built codebase: in your experience, is the usual outcome that it's fixable, or that you end up rebuilding anyway? And is $1-2k/month pre-launch a normal running cost for something this size, or a red flag on its own?
Not looking to hire anyone here, just trying to get a sense of what's normal before we commission a proper review.
2 Replies
@Finnish Lapphund Looking for a sanity check from people who've been here before.
We have a Next.js/Vercel app with Supabase and Clerk, built over about two months with heavy AI assistance and no experienced engineer overseeing the work. Around $30k spent, and it's running $1-2k/month in API costs before we've even launched.
We're now deciding whether to keep building on it or start over with someone who actually knows what they're doing.
For those who've inherited an AI-built codebase: in your experience, is the usual outcome that it's fixable, or that you end up rebuilding anyway? And is $1-2k/month pre-launch a normal running cost for something this size, or a red flag on its own?
Not looking to hire anyone here, just trying to get a sense of what's normal before we commission a proper review.
Sokoke
Hi! I've worked with Next.js, Vercel, Supabase, and AI-assisted codebases, and what you're describing is something I've seen more than once.
From my experience, an AI-built project isn't automatically a rebuild. The first step is always a proper technical audit. Sometimes the architecture is solid and only needs refactoring, optimization, and cleanup. Other times, if the foundation is poor, rebuilding specific modules ends up being faster and more cost-effective than patching everything.
As for the $1k–2k/month pre-launch costs, that would make me want to investigate. It isn't necessarily normal or abnormal on its own it depends on what services are generating those costs. Common causes include inefficient API usage, unnecessary AI calls, poor caching, background jobs, or infrastructure that's been overprovisioned. In many cases, those expenses can be reduced significantly after optimization.
Before making a rebuild decision, I'd recommend having an experienced engineer perform a full code and infrastructure review. That should cover the architecture, code quality, security, scalability, database design, API usage, deployment, and cost analysis. You'll have a much clearer picture of whether the current codebase is worth investing in or whether starting fresh is the better long-term decision.
Wishing you the best with the review hope it turns out to be more salvageable than it looks.
From my experience, an AI-built project isn't automatically a rebuild. The first step is always a proper technical audit. Sometimes the architecture is solid and only needs refactoring, optimization, and cleanup. Other times, if the foundation is poor, rebuilding specific modules ends up being faster and more cost-effective than patching everything.
As for the $1k–2k/month pre-launch costs, that would make me want to investigate. It isn't necessarily normal or abnormal on its own it depends on what services are generating those costs. Common causes include inefficient API usage, unnecessary AI calls, poor caching, background jobs, or infrastructure that's been overprovisioned. In many cases, those expenses can be reduced significantly after optimization.
Before making a rebuild decision, I'd recommend having an experienced engineer perform a full code and infrastructure review. That should cover the architecture, code quality, security, scalability, database design, API usage, deployment, and cost analysis. You'll have a much clearer picture of whether the current codebase is worth investing in or whether starting fresh is the better long-term decision.
Wishing you the best with the review hope it turns out to be more salvageable than it looks.