Most active commenters

    ←back to thread

    Open-source Zig book

    (www.zigbook.net)
    692 points rudedogg | 13 comments | 16 Nov 25 19:44 UTC | HN request time: 0.88s | source | bottom
    Show context
    shuraman7 ◴[16 Nov 25 21:21 UTC] No.45948508[source]▶
    It's really hard to believe this isn't AI generated, but today I was trying to use the HTTP server from std after the 0.15 changes, couldn't figure out how it's supposed to work until I've searched repos in Github. LLM's couldn't figure it out as well, they were stuck in a loop of changing/breaking things even further until they arrived at the solution of using the deprecated way. so I guess this is actually handwritten which is amazing because it looks like the best resource I've seen up until now for Zig
    replies(2): >>45948572 #>>45948933 #
    blks ◴[16 Nov 25 22:14 UTC] No.45948933[source]▶
    > It's really hard to believe this isn't AI generated

    Case of a person who is relying on LLMs so much he cannot imagine doing something big by themselves.

    replies(1): >>45948985 #
    1. shuraman7 ◴[16 Nov 25 22:21 UTC] No.45948985[source]▶
    it's not only the size - it was pushed all at once, anonymously, using text that highly resembles that of an AI. I still think that some of the text is AI generated. perhaps not the code, but the wording of the text just reeks of AI
    replies(2): >>45949190 #>>45950261 #
    2. BlackjackCF ◴[16 Nov 25 22:52 UTC] No.45949190[source]▶
    Can you provide some examples where the text reeks of AI?
    replies(3): >>45949233 #>>45950095 #>>45950142 #
    3. wild_egg ◴[16 Nov 25 22:59 UTC] No.45949233[source]▶
    Literally the heading as soon as you click the submitted link

    > Learning Zig is not just about adding a language to your resume. It is about fundamentally changing how you think about software.

    The "it's not X, it's Y" phrasing screams LLM these days

    replies(2): >>45949396 #>>45950193 #
    4. ropable ◴[16 Nov 25 23:24 UTC] No.45949396{3}[source]▶
    It's almost as though the LLMs were trained on all the writing conventions which are used by humans and are parroting those, instead of generating novel outputs themselves.
    replies(1): >>45949674 #
    5. fifhtbtbf ◴[17 Nov 25 00:00 UTC] No.45949674{4}[source]▶
    They haven’t picked up any one human writing style, they’ve converged on a weird amalgamation of expressions and styles that taken together don’t resemble any real humans writing and begin to feel quite unnatural.
    replies(1): >>45950087 #
    6. wild_egg ◴[17 Nov 25 01:25 UTC] No.45950087{5}[source]▶
    The Uncanny Valley of prose.
    7. dilap ◴[17 Nov 25 01:28 UTC] No.45950095[source]▶
    I read the first few paragraphs. Very much reads like LLM slop to me...

    E.g., "Zig takes a different path. It reveals complexity—and then gives you the tools to master it."

    If we had a reliable oracle, I would happily bet a $K on significant LLM authorship.

    replies(1): >>45951635 #
    8. xeonmc ◴[17 Nov 25 01:41 UTC] No.45950142[source]▶
    https://www.zigbook.net/chapters/45__text-formatting-and-uni...

    The repetitiveness of the shell commands (and using zig build-exe instead of zig run when the samples consist of short snippets), the filler bullet points and section organization that fail to convey any actual conceptual structure. And ultimately throughout the book the general style of thought processes lacks any of the zig community’s cultural anachronisms.

    If you take a look at the repository you’ll also notice baffling tech choices not justified by the author that runs counter against the zig ethos.

    (Edit: the build system chapter is an even worse offender in meaningless cognitively-cluttering headings and flowcharts, it’s almost certainly entirely hallucinated, there is just an absurd degree of unziglikeness everywhere: https://www.zigbook.net/chapters/26__build-system-advanced-t... -- What’s with the completely irrelevant flowchart of building the zig compliler? What even is the point of module-graph.txt? And icing on the cake in the “Vendoring vs Registry Dependencies” section.)

    replies(1): >>45952518 #
    9. anon7000 ◴[17 Nov 25 01:50 UTC] No.45950193{3}[source]▶
    Plenty of people use “it’s not X, it’s Y”

    As someone who uses em-dashes a lot, I’m getting pretty tired of hearing something “screams AI” about extremely simple (and common) human constructs. Yeah, the author does use that convention a number of times. But that makes sense, if that’s a tool in your writing toolbox, you’ll pull it out pretty frequently. It’s not signal by itself, it’s noise. (does that make me an AI!?) We really need to be considering a lot more than that.

    Reading through the first article, it appears to be compelling writing and a pretty high quality presentation. That’s all that matters, tbh. People get upset about AI slop because it’s utterly worthless and exceptionally low quality.

    10. trenchpilgrim ◴[17 Nov 25 02:03 UTC] No.45950261[source]▶
    > it was pushed all at once

    For some of my projects I develop against my own private git server, then when I'm ready to go public, create a new git repo with a fully squashed history. My early commits are basically all `git commit -m "added stuff"`

    11. sgt ◴[17 Nov 25 07:50 UTC] No.45951635{3}[source]▶
    Yeah and then why would they explicitly deny it? Maybe the AI was instructed not to reveal its origin. It's painful to enjoy this book if I know it's likely made by an LLM.
    replies(1): >>45954407 #
    12. cowsandmilk ◴[17 Nov 25 10:59 UTC] No.45952518{3}[source]▶
    The repetitiveness suggests copying and pasting, not an LLM. I actually find LLMs unlikely to do this.
    13. dilap ◴[17 Nov 25 15:28 UTC] No.45954407{4}[source]▶
    If you find it useful no harm in enjoying it! The main problem with AI content is it's just not good enough...yet. It'll get there. The LLMs just need more real-world feedback incorporated, rather than being the ultimate has-read-everything,-actually-knows-nothing dweeb (a lot of humans are like this too). (You can see the first signs of overcoming this w/ latest models coding skills, which are stronger via RL, I believe.) (Not first hand knowledge tho -- pot kettle black situation there.)