Raw capture — raw/ai-slop-starts-with-the-codebase-itself.md
url: https://thetruthasiseeitnow.com/ai-slop-starts-with-the-codebase-itself/ title: AI slop starts with the codebase itself. author: Unknown (originally published on LinkedIn) date_fetched: 2026-07-11 date_published: 2026-07-09
My view on software rewrites has changed because of AI. AI makes coding faster, but it also changes the answer to a fundamental question: how does it know what to do next?
The more popular your tech stack, the better AI will be with it. What the model already knows from training data matters — and less-discussed, so does the context you're providing through the codebase itself.
I've noticed a pattern when moving between codebases. In a clean, well-structured system, the model understands the patterns from reading a few files and gives me an implementation with minimal extra prompt. In an inconsistent codebase I designed two years ago, it fumbles unless I provide examples of similar existing code and additional documentation on how things should be structured. The difference isn't subtle — the former workflow feels like using AI to solve the problem, while the latter feels like trying to get AI to learn your language first.
That time spent teaching AI your proprietary patterns and inconsistent conventions — that's your competitors' advantage. Not because they prompt better, but because their codebase speaks a dialect the models already understand.
A rewrite isn't just a technology modernization opportunity anymore. It's a chance to rebuild around patterns that align with AI's strengths. This changes the economics of the "should we rewrite?" question — the AI-productivity dividend is a factor that compounding already favors, but now the baseline shifts too.
Every minute your AI spends trying to figure out what you meant instead of building what you need is a minute your competitors are shipping.