MechanismArticle

Coding models are trained in ways that make codebases worse over time, because they're optimized to win SWE-bench-style benchmarks that reward known fixes but can't measure bad architecture.

Dex argues LLM-driven codebase degradation stems from benchmark incentives: SWE-bench-style evals reward reproducing known fixes but have no way to penalize poor architectural decisions. ✦ AI generated

Dex Horthy · The Pragmatic Engineer · 2026-07-15 · original ↗

Share card ↓Export .md Export .txt

Dex believes that the reason we see LLMs “degrade” existing codebases is because they are optimized to do well on SWE-bench-style benchmarks. These benchmarks reward reproducing a known fix in codebases like Django, but cannot measure poor architecture decisions.

Read full article ↗excerpt · fair-use quotation

✳ Django ✳ SWE-bench

Around this claim

This moment responds to

explains mechanism → AI makes technical debt a bigger problem than ever because it's so easy to pile on new code and the AI resists removing code even when explicitly asked, while many of its biggest boosters have never had to maintain an old, long-running codebase and don't grasp the consequences.Wes Bos · Syntax