## Where Architecture Should Happen When Coding Gets Faster Based on current discussions in the software development community (Hacker News, Medium, LinkedIn), here's the consensus on where architecture should happen when AI-assisted coding accelerates development: ### **1. Shift Architecture Left (Earlier in the Process)** The most common recommendation is to **move architecture decisions earlier** in the development lifecycle: - **Requirements Phase**: Use AI to validate requirements before coding begins - **Design Phase**: Evaluate architectural decisions before implementation - **Pre-Coding**: Generate architecture diagrams, sequence diagrams, and data models from product docs **Why**: When coding becomes cheap, the cost of architectural mistakes becomes disproportionately expensive. Fixing a bad architecture after implementation is much harder than getting it right upfront. ### **2. Continuous Architecture Review (Post-Merge)** If you can't shift left completely, implement **continuous architecture feedback loops**: - Review what changed at the system level after merge - Check for reuse opportunities and duplicated logic - Reassess security assumptions - Schedule refactors explicitly (not as wishes) - Keep features behind flags for easy rollback **Key insight from Hacker News**: "Architecture should already be continuous, but faster code creation may require stronger post-merge mechanisms." ### **3. Tiered Approach Based on Risk** One practical pattern from the discussions: - **P0 (Critical)**: Human-written, AI-verified (business logic, security, core systems) - **P1 (Important)**: AI-assisted, human-reviewed (connecting logic, integrations) - **P2 (Low-risk)**: AI-generated, minimal verification (UI, boilerplate, non-critical features) ### **4. Architecture as a Bottleneck** The Medium article highlights that **architecture has become the new bottleneck** because: - AI can generate code faster than humans can evaluate architectural implications - Architectural judgment hasn't become equally cheap - Teams need explicit processes for architectural review, not just "refactor later" ### **5. Practical Recommendations** **Before Merge:** - Use AI tools to generate architecture diagrams and validate designs - Implement architectural checklists in PR templates - Use AI to flag potential architectural issues during code review **After Merge:** - Schedule regular architecture review sessions - Use automated tools to detect architectural debt - Maintain a "refactor backlog" with explicit prioritization **Continuous:** - Embed architectural standards in CI/CD pipelines - Use AI to suggest architectural improvements based on code patterns - Regularly reassess architecture as the system evolves ### **Bottom Line** When coding gets faster, architecture should happen **both earlier (shift left) and continuously (post-merge)**. The key is making architectural feedback loops explicit and actionable, not relying on "we'll fix it later." The teams that succeed will be those that treat architecture as a continuous process, not a one-time upfront activity.