dminik

I'm sure that if I go to a court after a hit and run, the judge will accept me saying "There was lots of paying attention, just not where he came from. You might not understand how many petabytes of data these eyes are processing a day, so monitoring the face doesnt really tell you anything."

If OpenAI is unable to contain their own software, they should not be permitted to operate it. Instead, they're bragging.

So is a bomb. Doesn't make it safe.

And how is that not incompetence? They've set up an autonomous system with apparently zero monitoring and let it run wild. Yolo?

I had the same issue. Linux + Firefox resulted in zero movement. Chrome worked fine.

Did OpenAI not communicate with Hugging Face? The incompetence here is staggering.

Well, hacking is a crime, so surely someone will go to jail for this, right?

You can't say that. It would be a "slippery slope fallacy" or something. You're just supposed to wait until it's too late.

I'm having a hard time squaring together two of the claims Andrew makes here:

- No source changes - https://codeberg.org/ziglang/zig/issues/36237#issuecomment-1...

- No escape hatches - main post

Surely there are already projects out there that rely on behaviours prohibited by Fil-C. One example that immediately comes to mind is reliance on external/C libraries.

Overall, I think adding it as a build target/ABI to the main compiler is pretty neat, but I do agree that it seems quite against the spirit of Zig (no hidden allocations, no hidden control flow). Combined with the combative title, I wonder how much of this is just "if I give them this, they will shut up about Rust."

"golang" is literally "Go language". Doesn't matter that it's one word.

Did you have a $3000 phone?

Funnily enough, no. When I read that comment, I didn't even know that phones can go for such a price. But, apparently some triple folding phones do.

I read that as a "buy a budget phone and use it until it's gone, don't buy an expensive phone," which to be fair was exactly what I was doing up until this point.

And having a phone you like more is surely better than having to return your phone and buy a different one, no?

Yes. But I do include functionality when evaluating how much I like a phone. I liked the phone at first, but then the issues moved it to a dislike. Considering I only had it for a week, what further issues awaited? I didn't want to risk that.

Searching for a different phone sucks, but at least I get my money back.

This might come as a bit of a shock to you, but not everywhere has terrible customer protection. I was less interested in the (free) 2 year warranty than the 14 day free return mandated by the EU.

I wasn't really risking being denied, as long as I didn't break any rules.

Thanks for calling me an idiot I guess ...

I mean, I don't particularly understand how "caring about warranty" goes against what you've written after that. Replacing something for free is surely better than doing so for $300 dollars, no?

Are you saying I should have installed GrapheneOS on the phone, possibly discovered that the phone has hardware issues and then go out to buy another phone because I have an emergency fund? Or stick with a new phone that had issues?

Or maybe I have made a mistake by buying a phone more expensive than $300? I can see this one actually, but I was going for something that didn't have ads in every menu as the cheap Chinese phones I was using up until this point.

Outside of the used market, which I tend to ignore due to battery/performance degradation, there's no way for me to buy a Pixel for less than $300 anyways.

I work both on a pretty much bog-standard web (GraphQL) backend and the frontend that uses it. We switched over from Apollo on node to async-graphql on Rust.

The runtime performance is much better, but the compiler time performance is terrible. To be fair, this is mostly the fault of async-graphql, but that doesn't really matter all that much. For example, it's not uncommon for a single character SQL query change to trigger over a minute long incremental rebuild.

The rust compiler is just choking on the number of generics and codegenned functions.

I've personally looked at how to improve this, but short of breaking up the type graph using federation, nothing can help. Not even cranelift makes a noticeable dent.

Additionally, the team started off composed by a bunch of TypeScript/React/Node developers, so mistakes were made along the way.

Honestly, I would have recommended to just use C#.

That's not to say that I don't think Rust can work for web development. We have some (GraphQL-less) services where Rust is a great fit. Just maybe shouldn't have been the default. That or give up graphql ...

Bit hard to do with a phone I no longer have in my hand. Additionally, I wasn't sure if that would void my warranty.

It did cross my mind, and I did buy it for the possibility to do that in the future if necessary, but I just wasn't in the position to actually do so.

I have had a mixed experience. Bought my Mom a Pixel 9a. It seems to be running fine with no issues.

Bought a Pixel 10 Pro XL for myself and had to return it. Connectivity issues (WiFi connected, but no internet), screen losing colors (white would turn gray), ghosting issues (scrolled/hidden content would stay on screen for a period of time).

While I'm a rust enthusiast, I do agree that certain languages lend themselves well to particular domains. So a rewrite from Rust to something better suited is fine by me. In fact, while I do work on a rust project, I would not have and still would not recommend it as the choice for that particular project.

That being said, I had to do some double takes while reading this.

https://rtfeldman.com/rust-to-zig#memory-safety-post-rewrite

I feel that it's a bit weird to compare a rather well tested 7 (?) year old rust implementation with a brand new not yet released less than a year old Zig implementation. Without that context, this looks like a bad comparison for rust, when it is in fact the complete opposite.

https://rtfeldman.com/rust-to-zig#build-times

The swiftness of the Zig compilere here is insane, and would would very much shift my recommendation of Rust if it got to similar speeds.

That being said, I do find it funny that currently, the compilation speed is actually worse on Zig than Rust, despite Zig (anonymous commenters at least tbf) claiming the opposite for years.

How did you eventually discover the 35 ms figure for Roc? Did you have to temporarily update the codebase to 0.17?

https://rtfeldman.com/rust-to-zig#memory-control-zero-parse-...

Nothing negative here. I did play around with implementing a scripting language in this DOD-ish, index-based paradigm and yeah, it is neat.

I was thinking that it might be possible to do resumable computation across the network like this (in the context of frontend frameworks "resuming" UIs), but ultimately I have no use for this so just the experience itself was enough.

One note here is that it does tend to break completely if non-pointer-free data is introduced. It seems like it's either all or nothing.

https://rtfeldman.com/rust-to-zig#ecosystem-relevance

This is more of an LLVM thing, which is fair, but I find it funny that "LLVM unstable bad" while "Zig unstable whatever".

Overall though, this was an interesting read. And if the folks contributing to roc like zig then more power to them.

Last thing, the link here is broken (points to a TODO):

Zig's compiler itself is another

The agent does have to pull some data into the context. The way it usually works is that the LLM will output a tool call, which is just some structured text, that the harness, a software managing the LLM running on your own PC, then processes. The most common tool calls are read, write, update and execute (usually bash).

For example, the LLM might request to read /some/path/to/file.js at lines 10 to 50. The harness then sends the result of that tool call to the LLM which causes it to generate further text and possibly more tool calls.

Crucially though, since it is the harness processing requests from the LLM, it can do stuff like deny access, prompt the user for permission and various other things.

What's weird is that no other harness really does this for regular usage. I know some providers now offer a cloud based environment for their agent to run in independently, but as far as I know this is something you have to opt in to.

It's also not really necessary to do this. The input processing/token generation process dwarfs any gains you could make from moving the project closer to the metal running the LLM.

Really, the "negligence" here is that there was no validation for uploads. Even a simple "is this the home directory" check could have prevented much of the backlash.

That being said, I believe that this was mainly done to get clean training data for Grok. If you're just working off of file traces/snippets from regular agent usage, your training data is incomplete. Why not just get the whole project to train on ...

I do use old reddit on my PC, but unless I'm programming or gaming, most of my browsing is done on mobile.

There are extensions to make old reddit mobile friendly, but by default it is a terrible experience. Another example for "you can't use extensions on apps".

I kind of get this, but quite a lot of the websites/webapps I use are better off on the web.

Imagine that from tomorrow, hackernews is an app. You can't do stuff like open 10 different tabs of it, put it in a tab group, expect it to restore where you left off after shutting down your device. It's interaction with other pages is also wrong. It would have to either open a webview or a separate browser app, both of which make navigations more complex. Extensions no longer work, you can't change stylesheets or use user scripts and so on.

This is why I don't like Reddit pushing an app and breaking/limiting the mobile website. It's just not how I want to use it at all.

Also, after a quick look at a few files, the rust version appears to be much more commented. Not sure if that makes up the extra several thousand lines, but surely counts accounts for some of that.

According to this breakdown: https://ghloc.vercel.app/Poseidon-fan/linux-0.11-rs?branch=m...

It's about 15k lines of code for the kernel and the rest is various utilities, libraries and programs that can run on the kernel.

That's not how any other coding harness/agent works. Why confidently comment on stuff you know nothing about?

It will have to be sent through the context again. That's how LLMs work.

The only reason to do this is so that Musk has clean training data for his next model. Project setup, popular libraries, CI workflows, etc.

Apparently prompting LLMs is now "survival mode".

If you're relying on asking the LLM "pwease don't delete" then you're already in trouble. This kind of stuff doesn't work with people either and they generally exhibit actual signs of intelligence.

Oh, you can actually. But you have to bring receipts, otherwise it very well could be libel.

My issue with the paragraph is that it doesn't really connect into anything meaningful. It's just saying that Jarred was a beginner.

The only connection it has is as a segue into calling Jarred a terrible manager here:

It was at this point - when he suddenly became a manager - that this "beginner energy" started to hit differently for me.

Note how "beginner energy" is considered good in the first paragraph, but suddenly terrible when applied to a different thing here. Terrible work culture aside, and the fact that it seemingly worked out for those who joined aside, Jarred would obviously have beginner energy in management as well considering he's not done it before. Why is it suddenly bad here?

Honest question, is there any record of Andrew actually saying Jarred had "beginner energy," or was this invented for the post as well?

It's not really titled "My Thoughts on the scumbag Jarred Sumner" though, is it? Nor does Jarred's post even mention Andrew by name.

Should Jarred's post have mentioned how Andrew kept bikeshedding Bun code style and how he always felt superior about others?

Well, first off, it's literally the first sentence of a post which is ~62% about Jared, ~26% about the rewrite itself and 10% closing thoughts. It also sets up Jared as a beginner who apparently never learned.

Again, this post never needed to say anything about Jared. It looks weird to pull him in in literally the first sentence. It really shows what you're actually writing about.

Also notice how I didn't have to bring up Andrew soiling his Pampers when he was a wee beginner.