Shkuratov outlines four agency corrosion types

Sergey Shkuratov posted a taxonomy of agency corrosion on dev.to, arguing that activity can continue while the underlying agency silently degrades [dev.to]. The article enumerates four concrete failure modes. The first is an unclear contract: code-level constructs such as include act as a macro rather than a language feature, and static analysis may hide incomplete diagnostics, letting work proceed on a half-defined interface [dev.to]. The second is a background priority: Shkuratov's own job-search schedule illustrates a task that is nominally central but lives only as a “don’t forget” reminder, causing other activities to consume attention without a real ordering. The third is a false position: a mobile-development write-up pretended to speak from an experienced perspective while the author’s actual expertise lay elsewhere, leading the text to collapse under its own pretense [dev.to]. The fourth is missing stop conditions: a paused Binance experiment remains open because no explicit termination rule exists, so the track continues to drain resources despite lacking purpose. According to Shkuratov, unclear contracts can inflate effort by 15-30% per iteration, while background tasks that lack formal gating can cause teams to spend up to 40% of sprint capacity on low-impact work [dev.to]. False positions can create a feedback loop where confidence is borrowed, not earned, making peer reviews less effective and increasing rework rates. Open-ended tracks without stop criteria continue to consume compute and human hours, a phenomenon Shkuratov likens to “rust that never stops eating metal” [dev.to]. Teams that adopt explicit contracts, enforce priority gates, maintain honest authorial stances, and define clear termination criteria can cut hidden overhead, according to internal case studies cited by Shkuratov [dev.to].