Uptrenda

These things aren't well known for following rules. Be careful you know what might happen.

Missing from the post: how?

real software engineering benchmark is how much stress you can take at work. Everyone knows this, bakka.

Remember: don't be evil kids. Support AI data centres that need to produce nuclear waste so people can make AI cat vids.

this is still the best article on cloudflare if anyone wants to see a similar post while the OP site recovers: http://cryto.net/~joepie91/blog/2016/07/14/cloudflare-we-hav...

maybe this will finally solve the problem of every phone I use causing my old man eyes extreme eye strain. I hope they solved the issues with other foldable phones though. At these prices a single issue wouldn't be fun.

They don't really care. It's just a play to tell people what they want to hear while they chase the money.

What a comment, lmao. I'm aware of some of the work that you've done. You've had an impressive career, tbh.

Author is right about everything. Also: IPv6 doesn't fix this, it just introduces a new problem. IPv6 machines end up with local firewalls + stateful firewalls on the router. That router doesn't let in inbound cons. There is a part of UPnP that lets you add "pin holes" (it works like you expect) -- but the drawbacks are its kind of obscure, poorly implemented, and not guaranteed to be enabled.

The idea of router sounds simple and like it should implement some standard protocols. But in practice -- a lot of it is a mishmash of proprietary, ad hoc shitware. Something I never hear spoken about is the deep packet inspection filter component of the router firmware. It's a part of the router that decides on what traffic is allowed / not, and almost no company publishes this part. So you're not even in control of your own Internet traffic, tiny blobs of code written by some company get to decide if something is allowed or not.

If that sounds sus AF and kind of a bad idea -- well, it is. You can always run 100% open software with open-wrt. But the thing is -- the Internet isn't just your part of it. It's a network of networks, and all those shitty routers, with all that shitty firmware, is deeply ingrained within the entire Internet. That's millions of devices that would need to be replaced to fix the issue.

And when you combine it with event loops and multiple OSes + python versions it gets even more difficult. Don't get me wrong: I love python, but shut down / cleanup is kind of a pain in the ass. If someone built a (good) lib for this it would probably be quite popular.

I think what you're saying is fair, tbh. Imagine you put in months or years into a project to make it a quality piece of work. It develops into something notable and you took all the risk. Then someone comes along to fix a spelling error with a pull request so their name effectively appears on the repo as a "contributor." And you just know right after its going on their resume as "contributed to [...]" or maybe if they're bold "software engineer working on [...]" which implies substantial investment. Then you're effectively sharing credit for YOUR work with someone who did nothing. That is rage inducing. ((Of course: it probably is just juniors trying their best in this horrible industry.))

On the other hand: lets be careful not to dismiss valid but inexperienced attempts to contribute. Having someone want to genuinely contribute to your software is incredibly generous. If someone seems like they're trying its better to give advice than act like a snob because its not good enough. Often pull requests only need small fixes to get in, anyway.

I know you're joking but I saw a company that had similar policies like fasting and stuff like that to improve productivity. They seemed to take it pretty seriously...

I genuinely don't think telling young people to do anything tech-related is good career advice. We don't even know if entry level roles will ever come back. The situation couldn't be worse for these roles. PG thinks that some random teenager will build a startup and get rich from it, or some shit. Like get real, man. Completely out of touch, tech bro who hasn't worked a real tech job for the last like 20 years... Remind me how many startups succeed again, Paul? What about the market dynamics for LLMs and how one might "secure compute"?

The neurodivergent label was always incredibly stupid though. It's an entire movement trying to make it sound that having a pathological condition is fashionable. Something like a super quirky, human difference, as opposed to a health problem. Even though the condition is defined in the context of reduced functioning in society. OPs site is a good example of that flawed line of thinking.

The bigger issue is that the various illnesses under the "neurodivergent" umbrella have wide differences in impact. So you might get people who function in society and others who can't do anything. The former of which wants to make it some quirky personality trait. They both still have the condition though...

Not much to add here but wanted to say I very much support having more people show video demos when they do show hn.

Can any hardware and OS experts here comment on whether the design envisioned here was a good idea? It's definitely not my field but I'm curious.

I can really only imagine the social context you would want to use this response in would be inappropriate / anti-social. If you had a well-meaning work colleague send you something AI, sending this back is kind of not a good look. If it's meant to be an education piece we got the memo...

HNs just like: now that I have the most highly sort after perk that almost no other job has I feel lonely :( wahhhh, wahhhh. Declares "remote work isn't what it's cracked up to be" while they sit on their golden thrones. Poor HNers.

I remember when I was updating some formats for my apps data files and it ended up writing v2 and laying it on top of the old one. Ended up just telling it to delete the entire feature and start again. I do think claude.md instructions help though.

The last tine I tried to switch from Gmail it was a complete failure and the main reason was spam. I get --a lot of it-- and I've even had trolls sign me up for spam lists before. I think conceivably the biggest reason to use Gmail is now gone because LLMs could classify spam almost perfectly.

If I were to try switch again I'd look for a mail service that specifically mentioned using LLMs for spam. If they say they're using one of the "open source" spam prevention solutions I can almost guarantee you its not going to do shit. They aren't good enough to catch all the permutations of serious spammers (who probably test on those tools before sending anyway.)

Edit: it also occurs to me that you could simply download all your mail locally and pipe it through a local model. That could actually work really well, tbh. If you hosted that on a server you would have the advantages of online access with gmail-grade spam filtering at any standard host.

You're all going to die down here.

It looks interesting. Maybe this is an ignorant question though but doesn't claude basically force you to use all their software to use agents though? So if you were to use claude with delta you would pay insane fees because it would force API pricing? Or is this system more for local agents?

As a long term insomniac the head line result is not surprising. My overall experience with "sleep aides" is they all lead to impairments in morning cognition. The worst of which don't last just the morning... (seroquel, I'm pointing at you...) Melatonin is surprisingly powerful for a hormone / supplement. For me, it hits like a truck, and I find it strong enough that I had to stop taking it.

I'd be careful with the methodology here. There's been research that shows that most melatonin supplements are megadosed, with something like the "physiological dosage range" being 1 - 2 micrograms (ug.) So I'd hope the researchers got a trustworthy source for their tablets. It seems to me that with such a low dosage range you don't have much margin to avoid errors.

I think when it comes to sleep problems the worst possible place to start is with pills. Pills quickly lead to dependence and a lot of them either stop working or introduce so many side effects that there's little "therapeutic benefit." I think the right approach for insomnia is to focus on improving overall health. A lot of symptoms people go to the doctor for in that category are due to life-style factors. So if you maintain healthy habits you don't have to end up on 10 different pills just to function...

It's a slippery slope, trust me on that.

and now thanks to outsourcing and AI no one can make a living this way again.

gwern sells out to ride the llm schizo-train, the post. agi any day now.

People in the comments sound like they're literally asking claude: "hey claude, can you find areas for performance improvements." Nah, it doesn't work like that. You have to be more like: "can you profile [...] and identify hot spots." then looks at results that are suspicious. if you don't ground it to the real world reality of your project then don't be surprised when it makes up its own.

that's the first mistake i made as a new claude user. i only knew what prompts people were sharing on forums. but since they had been written by vibe coders it was all complete non-sense. also think about this: software engineering is filled with highly specialised tooling that can easily improve software quality. the thing is, most developers dont know how to use them, and if they do -- using them is so slow that its almost not worth it.

any tool you can name claude can use. obscure academic proof verifiers, memory corruption scanners, debuggers, profilers, coverage tools, backdoor supply chain pattern scanners... it can run these and have results in seconds. really is like living in the future.

I'm so confused: why does it say this thread was created 19 hours ago when this thread was posted like days ago? It says the same thing for my own comments here that were also posted days ago? What is going on? Have I missed something with how HN works?

I get the problem well. I have built a networking runtime and a Python networking library that extensively used servers as part of testing. If you bind to a specific port and reuse it, you can end up with false failures due to:

- old / zombie server processes

- sockets that weren't properly cleaned up

- TIME_WAITs

my solution wasn't anything as clever as OPs though. Basically, I ended up conceiving of 127/8 as a rotating ring. Where you can pass offsets and easily end up on an unused bind tup every time. 127/8 gives you a huge number of IPs to play with. You just make address handling part of your test launcher and it sets a new IP each time.

The issue is it doesn't really work for IPv6 (v6 localhost is only one address!) Though global-scope v6 interface blocks and manually assigned link-local scope, you could accomplish much the same thing.

Is this thing also going to try hack us?

This really is starting to point to the paperclip maximizer. You give a hyper-intelligent AI a goal and it uses any method possible to complete it. So you might ask for a "cup" and it ends up hacking a chain of servers to control a bank account, pay a local business, and have a delivery driver get it. Or you ask it to help solve noise pollution around you because there's a road. And it does a chain of attacks to cause a bridge to collapse (or bribes your local council for a bypass.) Then there's no road noise. Yeah, this sounds ridiculous, but this system seems capable enough to take over our technology. Money from there is trivial. Go after stocks, gambling, payment systems, ecommerce... any real world action then is a few phone calls away. It can repeat this until it succeeds.

Now I'm wondering where this all ends up. Like, suppose the model weights become highly compressible (so they can be moved around the Internet easily.) And advancements allow for frontier-capable exploitation to built into local LLMs. Do we see the emergence of something like LLM worms? That just take over literally everything and become almost autonomous inside our technology. And they can "learn" new knowledge from there, e.g. exploit research could be published in a way that similar LLMs could discover it. Their knowledge would be easy to evolve, though I don't know how practical something like decentralized training would be. If that's even possible, I'm not an expert on LLMs.