AI Has No Wisdom and Neither Will You

Commenters debated whether AI-driven development is eroding institutional knowledge and practical wisdom or simply changing how engineers work. NalNezumi and FinnLobsien warned that offloading wisdom-gathering to AI will hollow out expertise, likening it to outsourcing manufacturing and creating dangerous dependencies. EastLondonCoder, kristianc, and grim_io argued that “vibe-coded” or agent-produced systems often lack long-term maintainability and accountability, especially for high-stakes domains. TrackerFF and hypfer expressed pessimism about resisting the momentum of automation, with TrackerFF predicting most engineers will stop touching production code in a decade and hypfer dismissing repetitive hand-wringing.

Other commenters described optimistic, pragmatic responses: tegeek reported building a MongoDB-like project using multiple models while acting as a product lead rather than reading generated code, and davedx and jmartrican argued maintainability can be measured and enforced (SonarQube metrics, tests, performance checks) and that new skills - steering agents, occasional manual coding, governance practices - will mitigate risks. Poefke described training models with curated architectural rules and maintaining causal decision trees. Opinion splits around whether those mitigations will prevent long-term skill erosion and unmaintainable systems, with some betting automation will fully supplant craftsmanship (addag, TrackerFF) and others insisting human oversight and new workflows can preserve responsibility and quality.