Benchmarking Coding Agents on New vs Legacy Codebases — Denys Linkov, Wisedocs
Legacy codebases with 10+ repos significantly impede AI coding agent effectiveness during refactoring.
“because it's a legacy code base, or actually more than 10 repos, nobody actually wants to touch the code. It's not a fun experience.”
Wisedocs engineer Denys Linkov shares a real-world case study on deploying AI coding agents against a sprawling legacy medical-claims pipeline to evaluate a 6-month refactor decision. The talk frames technical debt as compounding financial debt and sets up benchmarking results comparing agent performance on legacy vs. greenfield codebases. The content is cut off before the actual benchmark findings are revealed, limiting its immediate signal value.