Hacker News is a community-driven platform for sharing and discussing technology news, startups, and programming-related topics. Through user submissions and comments, Hacker News offers insights into emerging technology trends, industry developments, and entrepreneurial ventures. Readers can participate in discussions, share their insights, and stay informed about the latest advancements in technology and innovation.

Hacker News

AI coding agents have dramatically reduced the cost of writing code, disrupting long-held engineering intuitions about trade-offs in planning, refactoring, testing, and documentation. While generating code is now nearly free, producing *good* code—correct, tested, documented, maintainable, and secure—still carries significant cost and requires developer judgment. The piece argues that engineers need to develop new personal and organizational habits to adapt, including reconsidering previously dismissed features or improvements by simply firing off an async agent prompt to see if it's worth pursuing.

Simon Willison’s Weblog

Alexey Zerkalenkov

### Writing code is cheap now

Code has always been expensive. Producing a few hundred lines of clean, tested code takes most software developers a full day or more. Many of our engineering habits <...> are built around this core constraint...

### Writing code is cheap now

Code has always been expensive. Producing a few hundred lines of clean, tested code takes most software developers a full day or more. Many of our engineering habits &lt;...&gt; are built around this core constraint...

Please explain me one thing here. How this will works in 2-3 years when entire AI industry will collapse? No AI is profitable and never will be. It simply requires too much power and expensive hardware to run.
This is new .com and dark cables, but worse.

Code has never been expensive.
I can write thousands of lines of code that compile, it’s not the flex people think it is.
What is expensive, and still is, are ideas, architecture and design decisions.
Implementation has always been cake the second we got good language features, coding was already easy as hell.
This is why we are barely seeing any productivity gains, because fast code doesn’t translate to good ideas nor architecture.
Not to mention the time you have to spend debubbing code you’ve never seen because an agent made it (and fails to fix).
Debugging code you’ve architectured is the easiest thing in the world, because you understand exactly how everything works.
I see my coworkers vibe coding the most useless stuff I’ve ever that I then have to rewrite almost entirely, not because AI sucked, but because their architecture choices sucked, agents just did as they were told.
You pay programmers to spend 90% of their time thinking how to solve a problem, 5% drinking coffee, 5% actually writing code.