Announcing the difficulty of replacing legacy systems due to technical debt, the article highlights how old systems remain in production while new systems are developed. Developers tend to do minimal work on the old system, which continues to accumulate debt.
The process of rewriting from scratch often results in two systems operating simultaneously, with the new system handling limited features and containing mostly inactive code. This situation persists as priorities shift or work stalls, risking abandonment of the replacement.
For engineers managing models or agents, the key takeaway is that incremental improvements, such as automated testing and targeted refactors, may be more effective than complete replacements. This approach can reduce risks and improve system stability.
Source: https://simonwillison.net/2026/Sep/6/theres-no-limit-to-how-bad-code-can-get/