Who is behind this

What this service is

vibeharbor.io takes software that was largely written by an AI coding assistant and makes it safe to put in front of real users. You connect a GitHub repository, an automated scan sizes it, and you get a fixed price before anyone touches the code. Once you pay, one named engineer hardens it for security, stability, performance and deployment, and hands the work back as a pull request you review like any other.

The pull request is the whole deliverable. There is no separate findings report, because a list of problems you still have to fix yourself is not the outcome you were trying to buy.

How engagements are assigned

One engineer per engagement, matched to the stack the scan detects, named to you before work begins. The engineer works on a dedicated branch and never commits to your default branch. If no engineer with the right stack has capacity, you are told the actual wait before you pay, rather than being left in a queue you cannot see.

The engineer bench is still being built out. That is a real constraint today, not a marketing position, and it is the reason capacity is confirmed per engagement instead of advertised as a headcount.

What this site will not claim

Every statistic in the articles here is traced to a named study, survey or report, and the sources are listed at the foot of each piece. If a number cannot be sourced, it does not get published.

The same rule applies to claims about the business itself. This site does not display a live engagement counter, a headcount, a delivery-success percentage or a compliance certification, because none of those would currently be measurements of anything. The figures on the homepage are the inputs to the pricing formula, which are published in full precisely so the quote can be checked rather than trusted.

Writing

The research blog covers what the evidence actually shows about AI-generated code: the security studies, the maintainability data, what a code audit costs and when hiring an engineer is premature. It is written for founders making a call about their own codebase, not for other engineers.