March 15, 2026
AI intern or office villain?
Ask HN: How is AI-assisted coding going for you professionally?
Coders divided: helpful sidekick, office feud, or just hype
TLDR: A Hacker News thread asks how AI tools work on the job; results say they help but won’t replace developers. The comments bring heat: anti‑AI review wars, “AI‑first” confusion, and cautious users treating AI like a supervised intern—useful, but never trusted alone.
On Hacker News, a “Ask HN” call for real talk on AI coding tools turned into a workplace soap opera. The usual doom vs. denial split gets replaced by a messy middle: small wins, big confusion, and a lot of office politics. One “AI‑First” team, says mc-0, shipped fast but admits they don’t understand parts of their setup, like a Next.js frontend (a website toolkit), while the backend is “a gigantic monorepo” (think one giant folder jammed with many projects). AI‑first… but first confused? The drama heats up with lazystar, whose anti‑AI teammates are allegedly stonewalling reviews, forcing them to delay a launch—while someone else’s quick fix, with zero tests, gets merged instantly. Cue snark: “AI means Always Ignored,” and the meme of “let the robot write the tests,” met with collective eye‑rolls. Not all doom: drrob, a veteran since 1999, uses Copilot with Claude and finds suggestions right ~40% of the time—enough for about 10 minutes saved a day. jantb loves AI for bug hunts in a huge codebase, calling it a needle-in-a-haystack finder. And daringrain keeps AI on a short leash: handy for reviews and brainstorming, but absolutely nothing ships without human approval. The vibe? AI is a flashlight—not a self-driving car.
Key Points
- •The post invites professionals to share concrete experiences using AI tools for coding.
- •Contributors are asked to detail which tools were used and why they were effective.
- •The call requests descriptions of challenges encountered and how they were resolved.
- •Respondents should provide context: stack, project type, team size, and experience level.
- •The goal is to build a grounded picture of AI-assisted development as of March 2026, avoiding polarized takes.