bakugo

Oh man, this is a blast from the past. Played TF2 with him a few times, he hasn't been online on Steam for a long time, but apparently he just left AI behind before it turned into the dystopian techbro circlejerk it is now, and still continues to work on other things he likes today, which is great to hear.

It does feel like the "hacker" part of Hacker News is all but gone at this point, the only thing people are trying to "hack" anymore is their bank account balance. I will remember 15.ai as a landmark from the tail-end of the "people doing cool things just for the sake of doing them and showing it to others" era of the internet.

> I actually can't read this.

I don't think anyone can. At this point, I'm almost starting to believe it's an intentional move to discourage reviewing AI-generated commits by making them extremely unpleasant to read, in favor of just pushing straight to master without question.

Recently, I was browsing through commits on a tool I use, to evaluate how AI-slopped it had become, and came across this absolutely amazing sentence that took me several passes to interpret, courtesy of Fable:

> The wrapper is the try/finally seam future entry-condition changes need without re-indenting the loop.

So this is definitely not an Opus-specific issue, as some people seem to believe.

https://github.com/phpstan/phpstan-src/commit/934432a1b5007f...

The prompt was probably "respond to this issue" (assuming it was even a manual prompt and not fully automated).

I'd say they read the room perfectly. The entire issue is AI slop, I don't see why their response shouldn't be, too.

They have absolutely no reason to take such complaints seriously when the complainers are so dependent on their product, they can't even complain without it. Think reading AI slop is unpleasant? Great, just stop generating more AI slop, it's easy. But that's not going to happen. These people probably need to ask Claude how to tie their shoelaces every morning.

- Three different instances of "The X is Y" in a single short comment

- Relatively new account, "AI Engineer building agentic systems"

- Most past comments contain em-dashes, more "The X is Y", etc.

AI slop article based on a flawed premise. It's 2026, the process of correctly securing an SSH server has exactly two steps:

1. Disable password auth, only public key auth should be enabled

2. Block public access to SSH entirely, use a VPN instead (Tailscale & co. make this trivial)

And 2 is entirely optional for most people reading SSH guides who just want a server to host their hobby project. Let's be real, you're probably not reading the auth logs anyway so they don't need to be clean, and if someone discovered an OpenSSH public key auth bypass vulnerability, they absolutely wouldn't waste it on you. Just let those dumb scanners go at it all day, they're not getting in.

And a little tangent: fail2ban is 100% placebo and does nothing except clean up the logs a bit. I don't understand why it's still a common recommendation for beginners, it's a relic from the past when bruteforcing was still a concern because people used password auth.

> which maybe means you can spot distillations a mile away.

I doubt this is the reason. The fact that Chinese labs are all distilling Claude/GPT/etc isn't exactly a well kept secret, they don't even bother removing the name "Claude" from the training data, so the models randomly refer to themselves as "Claude" all the time.

I think it's far more likely to be a side effect of how much synthetic data is being fed back into the models to make them better at coding. The degradation of Claude's prose has been gradual but steady ever since they shifted towards focusing only on code with Opus 4.5.

The obvious next step is to remove the code editor entirely.

Ironically, this article is rage-bait.

It still only blocks "well-behaved" bots that have proper User-Agents and respect robots.txt, so it's largely pointless.

The problematic bots are all disguising themselves as Chrome and sending requests from millions of residential proxy IPs, and the only real solution to those is some sort of captcha or PoW page on first visit.

> Name a game whose settings screen added to the gameplay in any meaningful way. Hell, name a game from memory that didn't have the same settings screen with minor styling variations as every other game you played.

This mentality is exactly why most modern AAA games are so generic and uninteresting. Great games are more than just their gameplay; they're complete, coherent experiences assembled by people who care about every part of them, not just the parts they deem to be "the important ones". Someone who doesn't care at all what their game's settings menu looks like probably doesn't truly care about any other part of the game, either - it suggests their primary concern is not the quality of the game, but how quickly and cheaply it can be made.

Games having menus and HUD interfaces that fit the rest of their aesthetic was once the standard, before the current trend of minimalism took over. Why do you think that is? Was it just because the developers of yesteryear were all idiots wasting time and budget on things that didn't matter instead of maximizing shareholder value? Or was it because they were focused on making a high quality, coherent game first, and making it "efficiently" second?

> Do you also rail against game engines which provide all of this stuff in a generic way?

Yes, every other game being made in Unreal Engine nowadays is one of the biggest contributing factors to them all feeling so samey (and running badly). Developers who don't care about the quality of the game and just want to get it out the door with the minimum effort possible gravitate towards Unreal Engine, because it's made for them, and this is reflected in the end product.

> I can keep coming up with ways that AI can make a game better.

No, you can come up with ways that AI can make the game easier and cheaper to make at the cost of craftsmanship. It's not the same thing. Adding more side quests won't make the game better if they're not made by people who care about the game as a whole and all its little details. Whatever improvements or optimizations your AI can come up with, a skilled human developer could come up with, too, the difference being that the human demands a living wage in exchange for his work, and that's just awful, isn't it? It's simply not possible for the entirety of the game to be crafted with care without going over budget (let's ignore all the past examples of games that did exactly this, with much smaller budgets, before unappealing ultra realistic graphics became the standard).

He made... a Wordpress SEO plugin? Is that supposed to make him an authority on web standards? If anything, this just gives more credence to the idea that it's just the usual SEO-optimized slop.

Actually, it gets better: He originally submitted the website to HN himself[0], and his post was flagged/removed, likely due to being AI generated and indistinguishable from a lot of the AI spam that clogs up /newest.

[0] https://news.ycombinator.com/item?id=48323406

I have to disagree. Someone who fully embraces and perpetuates sensationalist AI hype marketing like this would be far more likely to pay $10/mo to be fed more marketing than someone who questions and doubts it.

I'm sure that's very easy to say when you financially benefit from it.

How do you know they weren't? Have you analyzed the prompt they used?

There's no shortage of examples of unmaintainable spaghetti AI code, Claude Code is just one of many. If you have examples of good codebases maintained by "prompting experts", I'd love to see them.

> “Code quality” is mostly “human readability”

That's an extremely narrow view of programming, and shows a complete lack of experience.

Are these experts in the room with us right now? Because if even the creators of Claude seemingly can't prompt non spaghetti code (see: Claude Code leak), I'd like to know who can.

> Something suspiciously absent from this article is addressing whether some of these are in fact human visitors, but humans issuing chatgpt or similar queries instead of going directly.

That's not absent from the article, it's right there in the section titled "The Claude ratio". ChatGPT, Claude, etc. use different user-agents for scraping vs user-initiated requests, and the author notes that user-initiated requests were an absolutely miniscule fraction of the total traffic.

Someone willing to take the 2 minutes to ask Claude to change the logo can probably also ask it to fix any bugs they find, or add any new features they might want.

This marketing stunt must've been really successful in their eyes, because they're milking it as much as they can.

The social contract for putting a website on the public internet was built on the assumption that browsers would be used to access those websites, and that most browsers would generally behave in a similar way (click on page, display page, pretty simple).

LLMs are not browsers, nor are they people, so I don't see why they should be part of this equation at all. They do not display the page for a real person to read, they merely ingest its contents and then spit out something completely different.

It's open source, just ask your AI agent to change the logo.

API pricing is almost definitely profitable, but at this point I assume it's a small minority of their inference traffic compared to subscription usage, and unlikely to make up for the rest of their expenses on its own.

If you go into the "Specs" tab on the right, it explicitly lists the original N64 as compatible.

I agree with dwroberts here. Flashtoo's comment is indeed a better faith engagement on paper, but you should consider the question whether the root comment was deserving of such good faith engagement in the first place.

Going into a post where someone talks about their cool programming hobby project that they do for fun to say "I just made this with AI instead, it was so easy and I didn't even have to do anything myself" is effectively also dunking on the author and disrespecting/devaluing the effort they put into whatever they submitted, just in a slightly more indirect way. Personally, if I read such a comment under something I created, I'd feel way more insulted than if they just called me a slur or something.

These sorts of comments are everywhere on HN now, every single post not directly related to AI has someone in the comments mentioning [latest AI product] and how they used it in relation to the topic of the post somehow, to the point that I'm starting to believe it's straight up astroturfing. I'd argue that such comments deserve to be dunked on, downvoted, or even flagged, unless you want all non-AI discussion on HN to die off.

Except "Claude, take this code and change the variable names up a bit" is not lifting ideas, and you would not get away with doing that to software belonging to a company with a competent legal department.

Here's some annoyances of my own that honestly baffle me:

- The command to exit the program is /quit, not /exit like literally everything else. If you type /exit, it'll get sent as a message to the AI. Apparently, /exit was an alias of /quit at some point, but was removed because it was simply too intuitive.

- It has terminal subcommands, such as `pi install` to install an extension. Any reasonable person would expect that `pi help` would give you a list of such commands. Nope! That actually starts it up and sends the message "help" to the AI. Amazing.

It's a common theme that most AI-generated or AI-adjacent software is full of nonsensical little decisions like this, that almost make it feel like it's not intended to be used by sane humans.

The point is to signal to investors that they're all-in on the current fad, thus making the stock price go up.

Sounds like everything is proceeding according to plan.

https://www.youtube.com/watch?v=-gGLvg0n-uY