April 2, 2026
Smug today, automated tomorrow?
The Beginning of Programming as We'll Know It
Programmers say they still run the show; commenters yell 'copium' and The Sims jokes
TLDR: A veteran developer argues AI writes lots of code fast but still needs humans to guide and fix it, at least for now. Commenters split between calling AI “coding-adjacent,” timing how long humans stay in charge, and pitching a human+AI team-up—turning a tech post into a full-blown job-future brawl.
Developer Daniel Jalkut says the quiet part out loud: AI can spit out code fast, but humans still need to steer, review, and fix it before it counts as real work. He’s pro‑AI, but cautious: it’s powerful, it’ll take some jobs, and yet, for now, people who know what they’re doing win. The internet heard that and immediately turned into a town square with megaphones.
The sharpest jab came from one commenter who declared AI “coding‑adjacent,” comparing it to playing The Sims instead of actually homemaking. Another demanded a timeline: “Do I get to feel smug for 10 days or 10 years?” Cue the existential career panic. A chess fan chimed in with the classic analogy: just like grandmasters plus chess engines once beat engines alone, maybe the winning formula is coder + AI. Meanwhile, a resident pedant called foul on the psychology: “That’s not confirmation bias, that’s survivorship bias,” basically saying we only see the flashy success stories, not the flaming failures.
The vibe? Equal parts hype, hope, and heavy “copium.” Some users accuse devs of protecting the status quo; others say the human‑AI tag team is the future. Between “AI babysitter” jokes and smug timers counting down to replacement, this thread is a popcorn‑ready mix of panic and punchlines. Read Jalkut’s post on Bitsplitting, then bring a helmet.
Key Points
- •The author argues human programmers remain essential in the current transition, augmenting AI with judgment, taste, and caution.
- •Anecdotes of AI building apps from scratch are common but reflect selective reporting; failures are underreported.
- •The author uses AI daily, guiding and correcting outputs, and has one model critique another to improve quality.
- •Most AI-generated code is not high quality, but the speed and volume can be valuable under human supervision.
- •AI is expected to be broadly beneficial yet will displace jobs; AI code should not be considered “work” until reviewed and fixed by a human.