Hacker News Reader

Comment by lubujackson | Hacker News Reader

This misses the best use of AI in my opinion, which is to gain understanding. Whst is this bit of code doing? Is there a risk of data leaking here? Are permissions enforced downstream of this function?

Code review can go much deeper now if you use AI to aggressively attack a PR combined with your human insight. Same for planning a feature:

Can I consolidate this logic to a shared function? Does the error surface to the user and are there any gaps? What preexisting functionality is affected by this PR? Can this query be made more efficient?

LLMs are great with focused questions, up and down abstraction layers and across all kinds of concerns. Stack up these focused concerns into a rich understanding of what you are doing or writing.

Understanding is the real output, code is the byproduct.

View on HN

Agree with the spirit of this, but it assumes a human is still looking at the code, and human inspection of code is vanishing faster than people are willing to admit (especially when coding in a language said human doesn’t know).

I have always found "not knowing" a programming language to be such a fascinating failure mode for professional software engineers. Especially as someone who learned to program with all programming documentation available online.

Like, sure, there are some wildly different paradigms that can be difficult to switch between, especially if you don't have any prior exposure to that paradigm. But there are way more programming languages than there are paradigms, so odds are high that using another language will primarily use a paradigm you're familiar with. So the main difficulties are in using a new standard library (where "using a new library" is a skill software engineers exercise every time they pull in a new library as a dependency), and some syntax quirks (of which tend to be extremely minor differences in the scope of languages used for industry).

It's hard for me to see "I'm not an X language engineer" as anything other than directly rejecting the task that using X language would enable. Especially now, when we can directly ask "what does this line / block of code do?" and get a detailed breakdown of exactly what's going on, on demand.

> I have always found "not knowing" a programming language to be such a fascinating failure mode for professional software engineers

This is how you wind up with code riddled with vulnerabilities and anti-patterns, with extremely bad performance imo

It's so easy to say "I know Java and C# is basically the same" and then wind up using built ins that are deprecated, using libraries that contain vulnerabilities, or whatever other mistakes

Maybe that's "good enough" for your usecases, but I've never found that it's a good idea to just assume I can switch languages for anything reasonably complex without some growing pains. Sometimes very big ones

Do you need that analysis every time from an LLM? Those are things I’d expect anyone to grow out of pretty quickly once you get used to understanding code. Especially someone with Senior in their title.

Doing this may give you understanding but won’t really teach you how to understand. For students (as the article is focusing on them), the latter is equally important.

> LLMs are great with focused questions, up and down abstraction layers and across all kinds of concerns. Stack up these focused concerns into a rich understanding of what you are doing or writing.

I agree. This is much better advice than what the article tries to do, which is to give best practices around how to use or not use AI to code.

Closely related: one of the biggest AI superpowers I've found is testing. "Please implement a comprehensive unit testing suite for this, making sure to test ..."

You still have to check the work and shouldn't trust it blindly, but in my experience it's pretty damn good.

Agreed — I also always append to my agents.md — "If there are any technical choices available, please surface them to me so I can understand them and choose"

If your harness has an “AskUserQuestion”-like tool, you can prompt it to use that, or, as I have done, add a stop hook that uses a smaller model to check if there was any similar prose that was output toward the end of the previous response.

Happy to share the implementation or package it up if anyone is interested.

Exactly, i so wish the narratives around LLMs in most fields were around how they can enable us to be better ourselves...

Wish I could upvote twice or something

> This misses the best use of AI in my opinion, which is to gain understanding. Whst is this bit of code doing? Is there a risk of data leaking here? Are permissions enforced downstream of this function?

Do I understand when I just let AI regurgitate the function and behavior of the code?

SOMEONE didn't read the article

> You code and you have the AI check your work. Treat it like a reviewer. That’s my advice in a sentence.

Bro you're giving the same advice as the article

>Preferably, don’t give the AI access to the codebase. Copy-paste snippets of your code in the web interface.

Well, I can't ask it for an overview of a system it can't see.

> Preferably, don’t give the AI access to the codebase. [...] If the AI does have access to the codebase, this access is read-only.

also

> I don’t follow all these religiously.

So if you need an overview of a system, you can and should provide access to the entire codebase. Even the rule itself says "preferably", not "don't ever". If you have different needs, you can always just look at it as a useful guideline. It's just a blog post by some rando on the internet, not the epiphany of AI coding's Ten Commandments.

Using AI to review our own comments before posting could be a good idea

Quite different advice, actually. I do not trust an LLM to review or to code by itself, but it can enlighten or expose blindspots, without requiring me to dig through the codebase, especially around unrelated downstream impacts. The point is to use LLMs as a research and concept navigation tool rather than a reviewer or "decider". Broad asks like "is this code good" are vague and lossy, narrow asks allow much less room for unknown unknowns.

> Whst is this bit of code doing?

This only applies to bad code. A formal programmic language beats an informal language anytime. Why on earth would you translate an unambiguous formal language to English?

> Is there a risk of data leaking here?

> Are permissions enforced downstream of this function?

If you trust AI to give you that answer, you are going to run into serious issues.

> Why on earth would you translate an unambiguous formal language to English?

Why on earth are all the books about programming, science and math written in English (and other natural languages) instead of a formal one then?

Why on earth did comments even got invented?

A formal programmic language beats an informal language anytime. Why on earth would you write a comment in an ambiguous informal language when you could translate it to code?

programmer looks at formal programmic language which has so beaten the informal language that one cannot tell the reason why this stupid bit of code exists in the silly form it does is to prevent an edge condition in a particular browser while still working fine for other browsers.

Luckily genius programmer who understands formal programmic language and does not need weak comments to clarify why the world is not perfect, fixes the stupid looking code to be beautiful, opening company up to some really super neat, seemingly random bugs in one browser that nobody will understand for several months.

tldr: one important reason comments exist is to clarify why code you might expect to look another way looks the way it does, of course if the only reason it looks the way it does is this is the way the AI wrote it I guess I agree the comment is unnecessary.

I think you missed the joke.

I wasn’t talking about code comments, but why this commenter who professed the superiority of formal languages could only get his argument across in an informal language.

Oh, sorry, I'm used to that exact argument that self-describing code is superior to comments, and it drives me up the wall.

Reply on news.ycombinator.com