Feeling extinct? Why programmers should ride the LLM wave instead of fearing it

Fabien Sanglard opens with the story of Phil Tippett, the stop-motion master Spielberg hired for Jurassic Park who declared ‘I feel extinct’ the moment ILM’s photorealistic CGI T. rex test made his go-motion craft obsolete overnight. Sanglard uses the anecdote as a lens on the anxiety programmers feel about LLMs. His argument, echoing a Hacker News commenter and John Carmack, is that every generation of developers has weathered a comparable upheaval — the web, computer graphics, mobile-first — and that survival means evolving rather than clinging to a craft. Coding was never the source of value; problem-solving is. LLMs are just the next tool, and refusing to use them (like the assembly diehards who wouldn’t move to C) means falling behind on sheer output.

The more concrete half of the piece is a working developer’s account of how the job actually changes. Sanglard no longer writes most lines by hand, but he still owns the output: full ‘vibe coding’ produces an indecipherable mess, so code quality, readable architecture, and iterating a PR back to hand-crafted standards still matter. He encodes his preferences (no magic numbers, early returns over deep nesting, enums instead of boolean params, respect for layering) into CLAUDE.md/GEMINI.md files so agents emulate his style, and flags context-switching across multiple simultaneous agents as a real source of mental fatigue and burnout.

Because authoring is now cheap, his expectations as a reviewer have risen sharply: well-formed commit messages, small reviewable PRs, unit and CI tests on every change, and an LLM self-critique pass before a human ever sees the code. He’s also more willing to refuse dependencies — asking an LLM to write a Levenshtein function rather than pulling in a package. The closing thought is that smaller teams can now accomplish far more, edging back toward the 1990s model where four people could ship professional software.