Long-running coding agents shift software development from prompt-driven to systems-thinking¶
Insight: When coding agents can run long-horizon tasks in parallel for days or weeks, software stops being prompt-driven and becomes systems-thinking-driven engineering work. Two concurrent challenges emerge: continuity (keeping agents coherent across context resets, per Anthropic's research) and concurrency (coordinating swarms of agents on the same codebase, per Cursor's experiments). This raises the bar on "builder literacy" — practitioners need systems engineering skills, not just prompt engineering.