Agentic coding is shifting the economics of software abstraction. Using a real freelance project as a case study, the author ran a head-to-head experiment building custom charts with D3 primitives versus Recharts. The D3 approach matched the design system exactly without workarounds, while Recharts hit a wall at the last 20% of customization. The core argument: high-level libraries were never free — they traded flexibility for implementation labor. When AI agents collapse the cost of writing custom code, the rational choice moves back down the abstraction ladder toward unopinionated primitives your team owns. The post also argues against itself: this bet only makes sense if your team has an agentic workflow, understands the output, and your UI requirements genuinely diverge from a library's happy path. For commodity UIs or deep accessibility needs, mature libraries still win.
Table of contents
The Experiment: D3 Primitives vs. RechartsWhat an Abstraction Actually IsTwenty Years Up the LadderThe Swing Back DownPrimitives Are Not Zero DependenciesWhen the High-Level Library Still WinsWhere the Crossover Sits30.1K Impressions1 Comment