The AI-and-programming-identity crisis sharpens
Two threads ran in parallel today: one on how to keep enjoying programming in a world of LLMs, and one from a teacher who redesigned their entire homework structure after AI did all the assignments. Both threads circled the same question from different angles, what is the actual skill being developed when an AI can produce the artifact?
The teacher thread landed on 'grade the process, not the result' and 'a simple explain-why question exposes the gap.' The programming thread split between people who feel fulfilled delegating boring tasks to LLMs and people who feel like they got an A on a test by cheating. One commenter noted: 'I'll happily let it generate the CSV parser and still want to hand-write the interesting parts.' Another said using Claude Code for two days felt deeply unfulfilling.
A separate thread from a 60-year-old engineer saying Claude Code re-ignited their passion for programming added a useful counterpoint: the people who feel liberated tend to be those who had the skill first and are now unblocked on execution. The people who feel hollow may be those using it to skip the learning.
So what?
If you're hiring engineers, the interview process needs to probe for genuine understanding, not just output. The 'explain your solution' bar is going to become the standard filter. For founders building dev tools, the products that will win are probably the ones that keep the human in the loop on decisions, not just generation.