Software Engineer: AI Coding Makes Me 'Lazier and
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The software engineering workflow has compressed from writing specs, researching data structures, coding tests, and documentation down to prompting AI, reviewing output, and merging code into the codebase.
- AI-generated code functions like "a competent, fast-working junior or mid-level developer" that lacks the institutional and systems-level knowledge senior engineers accumulate over decades on a codebase.
- Mihaly Csikszentmihalyi's "Flow" state, described in his 1990 book, is lost when developers become editors of AI output rather than creators, removing the deep engagement that pulls workers into four-hour focus sessions.
- The author reports becoming "noticeably lazier and stupider" after months of reviewing AI code, and says he deliberately avoids AI for novel writing because "writing itself is the act of organizing and clarifying my own thought."
- Stack Overflow and similar knowledge-sharing platforms are "drying up" as users take questions to Claude and ChatGPT instead, leaving future AI models with less human-generated expertise to learn from.
- Software companies fired junior developers to save costs on AI, but the author warns this eliminates the training ground for the senior engineers who, in five years, will be needed to manage AI bots on problems too large for their context windows.
Why it matters: The author's central concern is structural rather than personal: companies eliminated junior developers to save on AI costs, but those juniors were the pipeline for the senior engineers who vet AI output and solve the complex, broad problems the author says exceed AI's context windows. The drying up of Stack Overflow adds a second-order problem: future AI models will train on a shrinking pool of human expertise.
Ask SkimNews
