AI Ruined Programming, So I Fixed It (2025) (lightcapai.medium.com)

A provocative take argues that generative AI has fundamentally changed programming: machines now write most of the code, shrinking the craft of implementation and shifting value upstream to problem choice, systems thinking and “taste.” The author cites large-scale bibliometric work (Park et al., Nature 2023) showing declining disruption in science and patents, and recent productivity studies (Noy & Zhang 2023; Doshi et al. 2024; Brynjolfsson et al. 2025) that show generative AI dramatically speeds knowledge work—especially prototyping—while also increasing homogeneity unless steered by humans. The result: coding feels less rewarding, but the opportunity is clear: human judgment matters more than ever. The practical fix is a five‑phase “developer_evolution”: (1) identify broken systems, (2) validate with real users, (3) prototype with AI (claimed 10x speed), (4) stress-test edge cases where AI assumptions fail, and (5) ship, iterate and measure to earn user trust. Key technical implications: treat AI as a rapid implementation engine, not an oracle; invest heavily in realistic inputs and adversarial/edge testing; instrument releases for metrics and feedback; prioritize system design and problem selection over syntax mastery. Bottom line: AI handles code, humans handle decisions—teams that master that collaboration will outpace those that don’t.