December 28, 2025
From zero to infinity
Ask HN: By what percentage has AI changed your output as a software engineer?
Coders argue AI made them faster… or slower, or infinite
TLDR: A viral thread asked how much AI boosts coding, with the poster saying about 2x overall but wildly variable. Comments split between huge gains (10–20x, even 300%), skeptical pushback about self-reports, and warnings that AI code needs more review—showing AI’s impact is real but messy and uneven.
On Hacker News, a simple question turned into a productivity circus: by what percentage has AI changed your work? The original poster claims roughly 2x, sometimes 10x when they know the problem well, but warns of the heartbreak of “a month’s work in a day” followed by weeks of refactoring when the prompts (instructions to the AI) are vague. They also gush over tiny tool tweaks that feel like magic. Then the comments detonated. One user flexed “300%,” another deadpanned “it’s between negative gains and infinity,” and a third threw a meme grenade: “what’s 0 times how fast the LLM can do it.” LLMs (large language models) are those chatty code robots everyone’s poking. The vibe‑coding crowd bragged 10–20x after changing their whole workflow, shipping side projects they’d never finish alone. Meanwhile, office realists shrugged: AI is “only 2x” on grungy maintenance tasks. The drama? Trust. Skeptics say self‑reported percentages are fantasy and demand studies; pragmatists warn AI code needs more review; optimists wave new open‑source tools like trophies. The community mood swings wildly between superhero cape and cleanup crew—and somehow that chaos is the most honest answer of all.
Key Points
- •The author estimates an overall 2x productivity increase since adopting AI coding tools and LLMs.
- •In well-understood domains and familiar stacks, the author reports up to ~10x speedups with similar or better code quality.
- •In unfamiliar domains, AI may produce quick but low-quality code that requires weeks of refactoring, though still faster than pre-AI overall.
- •Working in unfamiliar stacks increases risk because mistakes from AI or prompts are harder to detect, reducing net productivity.
- •About 10–15% of the productivity gain comes from rapid development environment tweaks (e.g., .zshrc, .vimrc) automated by AI.