HN user
Hacker, developer, entrepreneur.
@guiambros on twitter.
nospamHN at wrgms.com.
Comments are my own, and do not represent my employer or anyone else.
--
[ my public key: https://keybase.io/guiambros; my proof: https://keybase.io/guiambros/sigs/XWO9Wth2q7MSNMQLcZ83jqGh-kXYcfINRphBE6NxZ6s ]
Both things can be true. SO was universally hated because they let (or actively incentivized) bullying run rampant and destroy the culture.
And yes, it's possible AI would have killed it anyway, despite its best efforts. But in this version of parallel universes, AI was only a catalyst. Its fate was sealed well before 2020.
No, of course not coincidence. It's just that people didn't have an alternative until 2021-ish. Users hated SO for many years prior, but had nowhere to go other than small niche communities. When LLMs came into play, it was just the obvious choice.
In an alternate universe, where LLMs didn't exist, I bet you SO would be equally dead by this decade. Someone better, with healthier values and a more welcoming community, would come up and steal their lunch.
Big fan of SW:KOTOR series. Would love to test Vestiges when ready.
You should consider creating the game on Steam, so you can start building your audience.
You missed the point (and I should have clarified better). It's not that 1 Gbps is bad; it's pretty good, actually. And I do think Verizon is one of the least bad options.
The problem is that there's zero competition. And while many other countries have 2, 5, 8Gbps available, there's zero reason for Verizon to offer it, despite being a densely populated area.
And 1 Gbps has been the max available for 10 years. It's very likely in 2036 we'll be in the exact same situation.
I wish we had a model like Switzerland, where the fiber is centrally managed, and providers compete just to connect you to the network.
I live in NYC, one of the most densely populated cities in the world, and yet Verizon Fios 1 Gbps is my only option. I tried to upgrade to Fios 2 Gbps, but it's not available. Spectrum only goes to 200Mbps; no other providers in my area.
I have no idea if Switzerland is any better, but the US situation in 2026 is appalling. If we're this bad in NYC, imagine what someone in rural America goes through.
Dec 2022:
Articulating ideas: https://x.com/GuiAmbros/status/1598897735955988481
Dave Jones didn't spare words [1] on how insane it was to have a jellybean component changing specs so significantly, particularly the input voltage from 22V to 18V, the removal of offset trim, and more.
Direct link, to save others a search: https://store.steampowered.com/app/4252730/Cosmic_Gold_Rush/
Well done! Reminds me of Atari's Asteroids, with a modern take.
It doesn't seem the case; the study was made purely on behavioral conditioning. And mosquitoes don't live very long, so it's unlikely they'd learn this in practice, outside of a very controlled environment.
But there's a natural selective pressure, and it's plausible that mosquitoes would eventually evolve their sensors to become attracted to DEET, over multiple generations. And with each generation lasting only 20-30 days and a single female mosquito laying 300-500 eggs in total, they can evolve orders of magnitude faster than us.
It works well on Android. Just zoom in and click the number, and you can breakdown per state. Click on any state number and it breaks down per city.
Pretty functional design.
It's really unfortunate that FPGA development is still stuck in the 90s. The incentives between IP owners and hobbyists are so misaligned that I don't see the possibility of this ever improving.
The market is full of dark patterns, and vendors like AMD/Xilinx can pull shitty moves like what OP highlighted, knowing there is no decent alternative (Altera is another disaster). Lattice had the opportunity to fully embrace opensource toolchain and try to disrupt from the bottom, but they seem stuck in the middle, not wanting to commit one way or another.
I'm grateful to SymbiFlow, and IceStorm and others, even though they obviously lack support for proprietary hardware features.
Plus now you get access to Safari books, and you also have their online library, so virtually any books you may need are accessible for free.
(That's for the CS graduate program; not sure about others)
Well, it certainly wasn't for lack of warning about the glaring risks...
What a horrible title. This has nothing to do with PyTorch (other than the relationship you mentioned).
A better title would be the Github repo title: "Computational Refutation of Quantum Superactivation"
This is awesome news. Thank you for the great work, and being so open to suggestions from the community. That's what makes Obsidian a world apart from all its competitors and predecessors.
This is fantastic news. Just a few days ago I mentioned [1] the Obsidian Community Plugins model was broken and needed an overhaul. This is a step in the right direction.
If I may, two suggestions:
1) Allow the user to filter for plugins based on the desired level of strictness (manually reviewed, safety rating, etc).
2) The Disclosures seems a bit too lenient. For example, the popular Templater plugin [2] gets a 92 rating, with Excellent Health and Satisfactory review. But the disclosures are pretty concerning: dynamic code execution, network calls, wasm blobs, malware scan not available, etc.
I know it's tricky to boil this down to a single numerical score that works for everyone, but I think the bar needs to be higher than this. And Plugin developers should be held to a higher standard (e.g. don't use eval()) or at least thoroughly document why you need it.
[1] https://news.ycombinator.com/item?id=48089793
[2] https://community.obsidian.md/plugins/templater-obsidian
Also:
- Think Python
- Think Data Structures
- Think Java
- Think Perl6 (!)
- Modeling and Simulation in Python
- Probably Overthinking It
And more [1]. He's a prolific writer, and very generous for offering many of them for free. I read several of them online or through O'Reilly, and bought printed copies just to appreciate his work. Really enjoyed Think DSP, Think Complexity, Think Bayes, etc.
[1] https://www.amazon.com/stores/Allen-Downey/author/B001O8NBPS
Yes, in this specific case.
Obsidian Plugins are still incredibly vulnerable. A compromised plugin will essentially take over your machine. There's no sandboxing of any kind. It's even more insecure than browser extensions (that could steal your auth tokens, but at least don't have unfettered access to your filesystem).
This is really unfortunate. I love Obsidian and am a paid subscriber for many years, but the community plugins needs a security overhaul asap, before someone gets hurt.
Not sure why you were downvoted. From the last paragraph:
"I spent a total of 35 days here. The first arrest was 3 days of processing, the initial 10 days followed by the 10 days extension for a total of 23 days before my case was dropped. But the same time my case was dropped my accusers found a another reason to issue a second arrest keeping me there for an additional 12 days!
Both cases were ultimately dropped and the second arrest was essentially tied to the first and shouldn’t have even been possible."
The answer is in your question:
"...through a vulnerable WordPress plugin, a web shell, weak SSH credentials, or a compromised container"
DirtyFrag alone doesn't help an attacker; they need to get in first. But the blast radius is much wider now. A wordpress flaw, or a prompt injection in your OpenClaw skills, or a supply chain compromise in npm librarires means they now have full root access to your system.
Also hilarious to see Drew Houston responding a bit later on the same thread:
we're in a similar space -- http://www.getdropbox.com (and part of the yc summer 07 program) basically, sync and backup done right (but for windows and os x). i had the same frustrations as you with existing solutions.
let me know if it's something you're interested in, or if you want to chat about it sometime.
drew (at getdropbox.com)
Oh bummer, I was really enjoying the series. Hope they bring it back.
Posting this to remind folks of Linus' and the kernel team's longstanding stance on security vulnerabilities, given the recent CopyFail discussion [1].
The researchers followed the standard disclosure process of 90+30, but distros were not notified. The kernel had a bug, but kernel developers did not (and will not) notify downstream distros.
The real discussion we should be having is: what should be the responsible disclosure process, and who should be accountable for contacting the downstream projects?
And should the Linux kernel be treated differently than other opensource projects? And if yes, where do we draw the line? If, for example, I find a bug in OpenSSL, is it reasonable to expect that I contact every single operating system, device maker, or library developer that packages openssl in their gizmos?
Linus? You mean, the same Linus who thinks "security people are f*cking morons", and "security bugs are just bugs"?
Linus is the reason why kernel team doesn't talk to distros. For them bugs are bugs, security related or not.
https://lkml.iu.edu/hypermail/linux/kernel/1711.2/01701.html...
I still use it daily, mostly for managing information consumption. It reads my twitter feed and scans HN twice per day, and sends me a digest of the discussions on Discord.
The best part is that it reads the comments too, and sends me a quick blurb. For example, this is what it sent me earlier, commenting on [1]:
TL;DR: A classic essay arguing that compiler construction isn't as hard as thick textbooks suggest, pointing to Jack Crenshaw's accessible "Let's Build a Compiler" series as the real starting point.
The Vibe: HN agrees most CS textbooks are overcomplicated — developers sharing their own minimal compiler projects and alternative learning resources.
It runs in an isolated vm and doesn't have access to anything other than my X account, so I'm not too worried about prompt injection. I also don't have any skills installed other the ones I developed or carefully vetted.
I like the combined suggestions of three other commenters:
1) Allow transfers during a very short window (e.g. 24h before the event)
2) Allow full refunds up to x days before the event
3) Release a small batch of tickets 24h before the event, as a way of reducing the chance for scalpers to make money, and giving real fans a last chance without paying exorbitant prices
All three together offer a reasonable tradeoff. The tickets will go (mostly) to real fans, yet still giving you flexibility in case your plans change (work, sick, etc). And if you know well in advance, you can get a full refund, without having to worry with reselling, paying commission, etc.
Also prohibit secondary markets entirely. Similar to airlines, there's no reselling of tickets.
Of course, this is just wishful thinking. Too many intermediaries benefit from screwing showgoers, so this will never be implemented.
Dupe from a few days ago [1], but glad this is gaining some attention. Truly remarkable work.
Yes, cat and mouse, except that each user who gets caught will have their account and credit card banned for life. Good luck dealing with the hassle of losing your account history.
Look, I would love to pay a fixed amount per month ($20, $200, whatever) and have an all-you-can-eat token buffet. But that is (today) a commercial impossibility.
Maybe one day inference will become so cheap that we'll all have a single subscription, and assisted by powerful local models. But today the token consumption is outpacing the reduction in inference cost, so we're nowhere near this becoming a reality.
I understand you want really hard to believe in what you're saying. But you should believe them when they say "no third party applications using OAuth".
It's irrelevant if you're paying them. They say clearly 1P apps are ok; 3P apps are not. Yours is 3P, so you're breaking the ToS.
Will they care? Probably not. At least not until you become popular enough to show up in their dashboards. So you should be fine for a whole, until you cross some threshold and they decide it's time to close the loophole.
Clever, but still against Anthropic's ToS [1]:
Use of third-party tools that misrepresent their identity to Anthropic’s servers, attempt to route third-party traffic against subscription limits, or otherwise violate applicable terms or policies is prohibited and such use may be enforced against.
[1] https://support.claude.com/en/articles/13189465-logging-in-t...