Hacker News Reader

Comment by tobyhinloopen | Hacker News Reader

AI also loves to explain what code does with stuff like this:

    // Check if sensor is available
    if (sensor.isAvailable()) {
      // Read current angle
      double angle = sensor.readAngle();

Which, to me, is a clear indicator of the author (being AI, I hope) not caring about code readability.

View on HN

Unfortunately staff engineers at my company will demand annoying comments like this so I don't think it's necessarily a sign of vibe coding

I get worse. I write a test

    TEST(SubSystemABCTest,CheckIfAIsNegativeItThrows)
    {
      ....
    }

And I get told all tests need a comment so I copy that line and add spaces between the words. WTF! >:(

I’ve run into this as well, it’s so obnoxious. Now I just go into multiple cursors mode and type the same exact stuff multiple times haha.

One would hope that the principal engineers above them have more sense…

What if a human wrote all of the comments in the function first then wrote the code? Should they then delete the comments?

Yes 100%

I need to create an AI agent that comments on any PR when it finds useless comments

Oddly enough I find senior engineers at my company demand ridiculous comments like this oddly enough. I've found that a 1:3 comment:code ratio (basically a comment every 2-3 lines with a new line separating each block) is the only thing that will prevent some annoying code review comments asking for clarifications.

Give them a copy of Robert C Martin’s Clean Code hah

I'm fighting this shit so hard at work every day, every PR review. The LLM writes comments like someone who can't read code is going to read it.

Bold of you to assume the majority of developers can read code.

> Bold of you to assume the majority of developers can read code.

Haha. Bold indeed.

Yeah, it's typical for LLMs. A comment anti-pattern. I wonder if LLM explainability can deliver and answer on "why" they are doing it.

This one is hallmark of Arduino sample code, which consists a majority of materials found on the internet regarding sensors and code.

I think the original human intent is to inflate mental share on hardware and/or time domain issues as those are just routine occurences in Arduino projects.

I wouldn't be surprised if it improves the language model's ability to write code. By inserting useless comments like that, its context will contain a very recent "prompt" of what it's trying to do when it actually writes the code.

Reply on news.ycombinator.com