Why are good developers so wary of agentic development?
Experienced developers often display caution toward agentic development because building software successfully is not the same as ensuring it functions safely and stably in production. An internal experiment at Thoughtworks, where agents built an application entirely from code without human typing, revealed that while development speed increased, the build was frequently red and unstable. The author argues that this hesitation stems from hard-earned professional judgment and the need for proper guardrails, rather than mere resistance to change. Open article →
Notable comments
The discussion is split but high-signal: many worry about broken accountability/trust and missing system-level understanding, while others argue good devs already know how to review/pause/guardrail agents—or that skepticism is mostly cultural and context-dependent.
HN · gopalv Draws a direct analogy to summer interns: the same accountability/reliability gap that forces humans to micro-control risky changes applies to agents—“if an intern can bring down production…”
HN · curt15 Argues that if humans stop internalizing the codebase, source code becomes performative; the agent could be treated as producing binaries that only “pass tests,” raising readiness/ownership concerns.
HN · datadrivenangel First-hand incident framing: major outages/security failures often start with a long chain of ‘minor’ misunderstandings/parsing/flag mistakes—highlighting why partial automation can’t reliably prevent worst-case combinations.