Open-source Zig book

I’m not sure how much value is to be had here, and it’s unfortunate the author wasn’t honest about how it was created.

I wish I wouldn’t have submitted this so quickly but I was excited about the new resource and the chapters I dug into looked good and accurate.

I worry about whether this will be maintained, if there are hallucinations, and if it’s worth investing time into.

> The Zigbook intentionally contains no AI-generated content—it is hand-written, carefully curated, and continuously updated to reflect the latest language features and best practices.

The author could of course be lying. But why would you use AI and then very explicitly call out that you’re not using AI?

Plenty. I assumed that the code examples had been cleaned up manually, so instead I looked at a few random "Caveats, alternatives, edge cases" sections. These contain errors typically made by LLMs, such as suggesting to use features that doesn't exist (std.mem.terminated), are non-public (argvToScriptCommandLineWindows) or removed (std.BoundedArray). These sections also surfaces irrelevant stdlib and compiler implementation details.

There are too many things off about the origin and author to not be suspicious of it. I’m not sure what the motivation was, but it seems likely. I do think they used the Zig source code heavily, and put together a pipeline of some sort feeding relevant context into the LLM, or maybe just codex or w/e instructed to read in the source.

It seems like it had to take quite a bit of effort to make, and is interesting on its own. And I would trust it more if I knew how it was made (LLMs or not).

As another suspicious data point see this issue by the author: https://github.com/microsoft/vscode/issues/272725

Edit: https://news.ycombinator.com/item?id=45952581 found some concrete issues

1. There is no evidence this is AI generated. The author claims it wasn't, and on the specific issue you cite, he explains why he's struggling with understanding it, even if the answer is "obvious" to most people here.

2. Even if it were AI generated, that does not automatically make it worthless. In fact, this looks pretty decent as a resource. Producing learning material is one of the few areas we can likely be confident that AI can add value, if the tools are used carefully - it's a lot better at that than producing working software, because synthesising knowledge seen elsewhere and moving it into a new relatable paradigm (which is what LLMs do, and excel at), is the job of teaching.

3. If it's maintained or not is neither here nor there - can it provide value to somebody right now, today? If yes, it's worth sharing today. It might not be in 6 months.

4. If there are hallucinations, we'll figure them out and prove the claim it is AI generated one way or another, and decide the overall value. If there is one hallucination per paragraph, it's a problem. If it's one every 5 chapters, it might be, but probably isn't. If it's one in 62 chapters, it's beating the error rate of human writers quite some way.

Yes, the GitHub history looks "off", but maybe they didn't want to develop in public and just wanted to get a clean v1.0 out there. Maybe it was all AI generated and they're hiding. I'm not sure it matters, to be honest.

But I do find it grating that every time somebody even suspects an LLM was involved, there is a rush of upvotes for "calling it out". This isn't rational thinking. It's not using data to make decisions, its not logical to assume all LLM-assisted writing is slop (even if some of it is), and it's actually not helpful in this case to somebody who is keen to learn zig to decide if this resource is useful or not: there are many programming tutorials written by human experts that are utterly useless, this might be a lot better.

Even if it was totally legitimate, the "landing page" (its design) and the headline ("Learning Zig is not just about adding a language to your resume. It is about fundamentally changing how you think about software."?????) should discredit it immediately.

There is, actually, You may copy the introduction to Pangram and it will say 100% AI generated.

We're used to errata and fixing up stuff produced by humans, so if we can fix this resource, it might actually be valuable and more useful than anything that existed before it. Maybe.

One of my things with AI is that if we assume it is there to replace humans, we are always going to find it disappointing. If we use it as a tool to augment, we might find it very useful.

A colleague used to describe it (long before GenAI, when we were talking about technology automation more generally) as following: "we're not trying to build a super intelligent killer robot to replace Deidre in accounts. Deidre knows things. We just want to give her better tools".

So, it seems like this needs some editing, but it still has value if we want it to have value. I'd rather this was fixed than thrown away (I'm biased, I want to learn systems programming in zig and want a good resource to do so), and yes the author should have been more upfront about it, and asked for reviewers, but we have it now. What to do?

It does make it automatically worthless if the author claims it's hand made. How am I supposed to trust this author if they just lie about things upfront? What worth does learning material have if it's written by a liar? How can I be sure the author isn't just lying with lots of information throughout the book?

They should not have lied about. That's not someone I would want to trust and support. There's probably a good reason why they decided to stay anonymous.

Personally, I would want no involvement in a project where the maintainer is this manipulative and I would find it a tragedy if any people contributed to their project.

    That didn't happen.
    And if it did, it wasn't that bad.
    And if it was, that's not a big deal.
    And if it is, that's not my fault.
    And if it was, I didn't mean it.
    And if I did, you deserved it.

See the first screenshot in the exchange with newlines breaking up the label and block - and maybe this for proper formatting https://zig.guide/language-basics/labelled-blocks/

Some text in the book itself is odd, but I'll be a guinea pig and try to learn zig from this book and see how far I get.

https://news.ycombinator.com/item?id=45948608

From the readme.

https://github.com/zigbook/zigbook/issues/25

"AI ALLEGATION" and "RE**ED COMPLAINT"