Codebase quality is declining, and most engineering managers do not care. That is the signal buried inside 900+ survey responses collected by The Pragmatic Engineer from its subscribers in 2026. The engineers who still understand the codebase are absorbing the maintenance burden as AI-generated code accumulates. Code ownership is eroding. Team collaboration is becoming less important. These are not predictions: they are reported conditions on the ground right now.
The survey surfaces three other findings worth your attention. Junior engineers are getting less value from AI tools and running up higher token costs than their senior peers, which inverts the assumption that AI is a great equalizer. AI agent usage is described by respondents as feeling like a slot machine, with pricing structures that reward prompting more. And org-level AI adoption is failing not because the tools are bad, but because the engineering culture that existed before AI arrived determines how much value teams actually extract.
What makes this worth reading in full is not the conclusions but the mechanism. The Pragmatic Engineer breaks down how sentiment has shifted since 2024, fewer engineers are negative but positivity has not risen to match, and why better models alone are not closing that gap. The data on less experienced engineers and token spend is specific enough to change how you think about onboarding and tooling budgets. Read the original for the breakdown by company size, engineering culture type, and the verbatim responses on what slot-machine prompting actually looks like in practice.
[READ ORIGINAL →]