Skip to content
TopicTracker
From HackerNewsView original
TranslationTranslation

Golden Paths Weren't Built for Agents

The article argues that traditional "Golden Paths" (predefined infrastructure templates) are inadequate for AI agents and LLM-driven automation, which require more dynamic, adaptable workflows rather than rigid, opinionated scaffolding.

Background

- **Golden Paths** (also called paved roads or opinionated defaults) are pre-approved, templated infrastructure/setup paths that organizations offer developers to reduce friction — the "blessed way" to deploy code. - **Massdriver** is a cloud infrastructure platform (Y Combinator–backed) focused on self-service internal developer platforms (IDPs). The blog post argues that golden paths were designed for *human* developers making manual choices, not for autonomous AI agents running unattended. - As AI coding agents (e.g., Cursor, Devin, GitHub Copilot) now generate infrastructure and deployment requests automatically, the assumptions behind golden paths break down: agents can't navigate branching decision trees, don't ask clarifying questions, and need strict guardrails or API-based enforcement rather than documentation-based guidance. - This is part of a broader industry conversation about how internal platforms must evolve when AI agents — not just human developers — become the primary "consumers" of platform APIs and deploy pipelines.

Related stories