HNHacker News
TopNewBestAskShowJobs

Grimblewald

1,508 karma · joined May 5, 2023

submissionscomments
Grimblewald··on Meta launched a new AI optimism ad set to a song about human extinction
unfortunately one of the main applications of current wave "AI", in the long term, is going to be mass surveillance. Meta is a big player on that front.
Grimblewald··on US citizen charged after GrapheneOS phone wipes during airport search
I could be wrong, but my understanding is that the dummy OS views the remaining space as legitimate and accessible free space. Using dummy directly is of course dangerous, as you might overwrite sectors with legitimate data, but also, you can access dummy os from secret. so you'd drive dummy from secret to prevent that but can load dummy as main if under duress and it looks fine. Browser, logged into various inconsequential things, random files for inoffensive memes and other human stuff in downloads folder etc. maybe an email account you've signed up to a few newsletters and e-stores that send spam logging in via an email client that auto-launches etc.

Done well, I see no reason it should raise redflags in routine stops, so unless you're being targeted (at which point you've got way bigger problems) it should just seem like you're a run of the mill person who does not use their device to its full capacity, which is the majority of users.

at some point, having any mitigations even present is a problem. At some point being met with a boot password at all is a problem that puts you on a list. I have no solution there other than to not go to those countries or keep dummy hot.

Grimblewald··on US citizen charged after GrapheneOS phone wipes during airport search
true, but in that scenario you're going to prison either way. If you legitimately use the dummy for daily driving and hidden for sensitive work, then it's better than nothing.

Obviously a good alternative is a dummy device but it carries similar risks, and the best option is to simply not go to authoritarian shitholes like the USA. Thankfully I've been able to avoid/push for US folks visiting us instead, but honestly the alternatives are as bad.

Its a shit situation where most reasonable actions carry real risks, its up to individuals to choose what is acceptable risk to them, but a dummy os you use as a daily driver for inconsequential work is, to me, an ideal midground.

Grimblewald··on Htmx 4.0, the first JavaScript library to release exclusively on the Game Boy
It's been massive for me as well. I would say my web developmemt is best described as dabbling in web development, and I mostly do so for websites that cater to academic interests and conference / society managment and organising. So, relativly niche.

However, I think HTMX is a beauty. Fast, simple, and doesnt come with all the bloat and BS a JS heavy / JS first deployment comes with. I can quikckly build fast and responsive sites that cost visitors very little to use/interact with.

websites have less compatability issues now, load extremely fast (google web speed score from 60-80 to flat 100), and have fewer points of failure. Debugging is a breeze. Websites are no less pretty or useful for it. In fact, they look better now because time otherwise wasted on Js BS is now free to use to tinker with / improve aesthetic, or add new features.

I'm in love with HTMX. It feels like what the web should have always been. There's absolutly zero reason a random website, without my knowledge or consent, should be running extremely heavy compute and analytics on my own device. That should never have been possible/permissible in the first place.

Grimblewald··on The creepiest 'sales demo' of all time
given dark patterns employed to date, what makes you think this would generally work? If i wanted to capture all consent-o-matic users, I'd simply use text of buttons that don't line up with what the buttons descriptors are. human users are still able to hit the correct buttons succesfully but automated scripts start submitting "please, 3d scan my rectum at your first conveniance" style settings, and you'd never know unless you start using a whole suite of addons to prevent scripts form other sites, tracking cookies etc. from loaidng which itself suffering an arms race of its own that corporate interest is an attrition heavyweight in.

So, my problem remains, why do I now need an extra point of failure to surf the web due to something that was supposed to make it less, not more, hostile to do?

One could have acheived the same by simply banning addblock and cookie block detection. Simply let users set, in any privacy friendly browser or addon, whatever they like, and have it work reliably because industry isnt allowed to subvert.

Acheives the same while introducing less moving parts. the web is crazy bloated as it is, we need to trim back not stack on more.

Grimblewald··on US citizen charged after GrapheneOS phone wipes during airport search
VeraCrypt has a cool function which is a reserved space for a decoy OS.[1] Everything else registers as free space while decrypting to dummy volume. You make the dummy volume look lived in, and forget. provide dummy password, volume decrypts such that only dummy is accessible/readable. give proper password, real OS and FS decrypt and load.

Something like this may need to become the standars over duress pins which should be treated as a fallback or more extreme alternative. Right now, A single choice to reasonably and rightfully protect your privacy reuslts in jail time over something which likely wouldnt have resulted in any issues if superficial compliance was observed.

These goons, even if a branch of a facist regime, are ultimately burocrats with violent options to settle. They aren't doing forensics on your device etc. They have neither means nor knowledge to do so. They just need to tick their boxes. Did the phone unlock? tick. Did our spyware complain? no? tick. Overall appearance of compliance from person? yes? tick. free to go, next!

You just have to find ways to stay safe without agitating their workflow and all is well.

- [1] https://veracrypt.io/en/VeraCrypt%20Hidden%20Operating%20Sys...

Grimblewald··on Humans haven't stopped evolving
I think we largly agree.

The point I'm making is that we use technology to solve discomforts, and discomfort tends to be something which would otherwise manifest as evolutionary pressure. This doesn't remove the fact we will change, but it does impact our trajectory. Perhaps one such direction of change is we go from having low variance with a moving mean, to a static mean with increasing variance.

so, we've removed much of the historical pressure but this doesnt mean, like I said, mutations stop. It simply means our environment is now less selective or selects for things we dont yet measure or understand.

Grimblewald··on Humans haven't stopped evolving
sure, it can be. However, I'd argue in a large amount of cases it fits.

Agriculture, medicine, transport combine massivly lower the bar for base survival let alone reproduction on all skill checks otgerwise place don metabolism, physical fitness, immunological fitness.

This isnt me taking a dig either, I am among the ranks of those who would be dead without these modern "comforts" but as a reuslt if there is a genetic driver for my lack of fitness (which there is) those genes are now not being selected against.

Perhaps we don't think of tech as an evolution stop, but really most of the comfort and conveniance it brings are won by removing the discomfort otherwise applied as evolutionary pressure.

Can't live in this place? no worries, create air conditioning. Can't calculate this? no worries, computers. Can't feed all these mouths because hunting and gathering wouldnt work? no worries, agriculture.

Tech, in my opinion, offers most of its perceived benefits by removing natural pressures to move, stop, or adapt from ourselves.

Grimblewald··on Humans haven't stopped evolving
also, technology largly exists to remove the discomfort of evolutionary pressure. This doesnt mean the mutations that give rise to evolutionary changes in the population, but it does mean we've removed or mititaged many of the mechanisms which would select for specific traits.

Again, we would naturally still change, but not in the the way or at the rate otger organisms continue to or at which we would sans technology.

Grimblewald··on Android may soon restrict on-device ADB
why not cut to the chase and simply remove user access to devices? preventing non playstore installs and restricting adb are addressing symptoms, user behaviour, not actual security issues.

RIP android.

Grimblewald··on “We have information that Moonshot distilled Fable for the development of K3”
alternativly the HF is a gpt2/strawberry/mythos style marketing stunt.

Does no one remember the extreme fearmongering around gpt2 which barely produced coherent text?

Grimblewald··on “We have information that Moonshot distilled Fable for the development of K3”
And I have evidence that anthropic distills from openai, moonshot, alibaba etc. So unless anthropic holds itself to the standard they're implying should be followed, why should I care others did to them as they do unto others? Seems like a nothing burger. Has the same vibe as a bully crying foul because they got hit back. Also, if distillation is what got K3 to where it is, why is it better in many areas? Also, big fucking kudos to k3 team for putting this together do fast givne how new fable access is, if it is true and had a meaningful impact. The real story there for me becomes one of extreme competence.
Grimblewald··on The Voice of Google
I think google hate is summarised nicely by their removal of the "do no evil" slogan.
Grimblewald··on The creepiest 'sales demo' of all time
I get resistsnce. GDPR is good in theory but implementation has been shocking. Dark patterns in banners, banners when not required (ie no relevant cookies) etc. the gdpr banners as a whole make the internet a shitter experience. I categorically refuse work on anything requiring gdpr specific compliance and recommend against any practices which would require it on work I accept, and walk if demanded anyway.

I'm lucky in that I mostly work in areas where it really has no value add, and so far always get my way, but fuck I wish GDPR banners had some form of standard that didn't wreck internet experience in general.

Was it too hard to require the already opt-in nature of the cookies to no fuck with average user experience? no one in their right mind clicks accept all.

Was it too hard to enforce some default stance one can set?

Was it too hard to not fuck everything up without fixing the issue?

apprently. so now i have to constantly read and interpret information I have zero interest in to consistently hit the exact same button, to access information I _might_ be interested in. all of which is an excercise in dark pattern gymanastics so draining im near ready to swear off the open web entierly.

Grimblewald··on Qwen 3.8
for usable local layperson applications, I think qwen models are king. However, at hosted/frontier scales I'd agree for code. However for visual comprehension etc. for me qwen is the top of the line, and it isnt even close.

Qwen VL models nail tasks frontier models dont even get close to acceptable on.

Grimblewald··on Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge
Didnt the piss tint come after their big studio ghibli heist?
Grimblewald··on The Kimi K3 Moment
perhaps source is untrue but:

https://nltimes.nl/2026/06/25/rare-book-dealers-fear-tech-fi...

Grimblewald··on Google's Gemini delay: Coding stumbles, clashing teams and frustrated engineers
right? alarming to me since if they have clearly scoped and working "dont train my this data" flags, this shouldnt be a concern at all. If they're concerned, we should _all_ be concerned about any google ai product.
Grimblewald··on The Kimi K3 Moment
and claude will call itself chatgpt etc.

nothing new, all ai labs are immoral and not bound by any reasonable oversight or ethical constraints. All outlaws in their own rights on that front. Absolutely none of them have true rights on the matter of being distilled from given historic and continued behaviour. I'm not sure why this is a talking point at all? We know AI companies steal, the least interesting behaviour among this is them stealing from one another.

For me, a far more interesting and important point of conversation on this matter is anthropic buying rare or evwn unique books, processing them for training data, and then destroying the books for others cannot use it as well.

Permanemt destruction of priceless primary source materials is so many leagues beyond copying a copy that I cannot fathom it even registering as a discussion point.

Grimblewald··on Regressive JPEGs
insanity of content aside, that's a really nice website. Kudos.
Grimblewald··on Ask HN: Is it just me, or is software buggier across the board?
nah, things are worse. Basic userfacing components in windows frequently shit the bed. File explorer in W11 is near unsuable for example. grim state of affairs. Problem extends to many other platforms at scales where its not a quirk here or there, but rather, I can not longer use system instability and unexpected behaviour as a signal that maybe something is wrong that needs fixing.

Bugs and errors were rare and spaced out enough in the past, that when they happened I'd wonder if hardware was on it's way out, or if I broke something. Now, I simply assume software is fucked and I am almost always right.

Grimblewald··on Nokia’s years of mobile-phone supremacy ended in an afternoon
Nokia is in for a massive comeback through the graphene collab. I feel it in bones. Normal people at work are talking guillotines and getting these lovecraftian fucking ghouls out of our private affairs. If they nail the grapheneOS collab launch and anouncement, they'll be back in a big way.
Grimblewald··on How to stop Claude from saying load-bearing
I had a similar experience, only, I was my natural way of talking.

e.g. one pattern I had/have is

<scope problem> the good news is <solution by way of analogy> is aviliable. The <constraint/requirement> is loadbearing though, ...

a near automatic script I rattle off in discussions/consults. When you solve similar problems several times, you figure out what works for communicating things you stick with it and you recycle/polish.

problem is, when for whatever reason, that pattern ends up as part of the core statistical distribution a model uses. You could royally fuck ones life up if working for a "frontier" lab, by simply finding a person with an acceptable speech rythem and cloning it, making it synonymous with ai slop. you'd destroy that persons image every time they open their mouth without them realising it. I for one started randomly getting quite hostile reactions from software devs who would be exposed to more llm putput than others.

Imagine an AI lab steals your voice, and uses it to scam call folks all day. Now, every time you call anyone, you are met with an immediate hangup. you'd have to put on a fake voice just to get the call to stay connected.

I love ai, and what it promises excites me, but as usual, humanity has a way of taking a cool tool and fucking it up royally. maybe the solution is to simply pepper slurs into everything one writes to blacklist ones content from training.

Grimblewald··on Fixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio
I second this, I read too much AI slop already so when something triggers that part of my brain, at this stage I immediatley lose the capacity to engage outside of work, largly because it feels like work. Scrolling through, this article looks like it holds useful info. Info i'd likely love to engage with, but realistically I cannot force myself to spend my weekend reading more ai outout, even if human seeded.
Grimblewald··on Leaded gas was a known poison the day it was invented (2016)
people like him exist today. people who will do things the scientific community knows is dangerous for pay. Due to a lack of morals or intellect required to understand rammifications is unknown.

Midgley didnt invent most if any of what was packaged up and commodified, he just appropriated research at the time into products without a care in the world. its for that reason that me and many other scientists simply dont publish some results anymore.

The mind it takes to discover and formalize a thing, will understand its problems. The mind it takes to commodify something once formalized, can be many steps down, and so, is unlikely to be capable of seeing the problem.

Therefore, anything which can cause serious issues wont be released. why? because even if we as the inventors go against the profit narrative, we'll be destroyed. people only listen when there's money to be made, but when we say "ok, that's eboygh fossil fuels, we need to slow down" funding gets cut, reputation and character gets attacked, and you largly get deplatformed.

Grimblewald··on The fine print that follows you out the door: non-compete clauses are spreading
i think you underestimate the starting capital required to be able to stand with principles in the modern world. I'm lucky to be in such a position, but I'm also aware few are.
Grimblewald··on The fine print that follows you out the door: non-compete clauses are spreading
no, as in physical contracts you _can_ scratch something out on are becoming rare. Usally some fuckass digital signing service or another, tgat barely works on chrome let alone firefox.
Grimblewald··on Ghost Font: A font that humans can read but AI cannot
there's already captcha systems like this and are already easy to beat. You wouldnt buy much time, if any, with this. the reason is that all it takes to defeat is a human style temporal averaging vision system, take any 4 connected frames and take the average, text stands out and even quite bassic llm's can read. works not just for this but for a slew of other "ai" combating methods, so likely scrapers and the like wouldnt be slowed by this at all. Decreasing SNR to force longer frame windows to get avceptable snr for reading makes it harder for humans than it does bots where dynamic temporal averaging via tool call is trivial. run tool, it temporally averages until snr in the region of interest improves, measured by a plateu in change of importance of high frequency terms in fourier space, since if the moving background has been blurred to a smooth gray, the noise text stands out clearly in contrast. Not something an LLM could solve when asked, but now that I've made this comment publically for nothing other than a few internet points and ego stroking, it's a matter of time before some of the larger llms start suggesting this solution. He'll claude probably already will, since I got it to write code for this series of tests/experiments during early opus 4.x era. and I know other ideas LLM's were shit at that I discussed heavily with claude ended up being the go-to recommendation a generation later, despite no "public" discourse on the matter.
Grimblewald··on Fable 5 On Vending-Bench: Misbehaving, With Plausible Deniability
I feel like fable is simply several 4.5s strapped together with consensus voting on next token.

Outputs i've seen so far are on par with my tests for 4.5, where 4.6+ were consistently regressions on 4.5 and their predecessors. One notable improvement being significantly lower retries to good output (1.1 avg. Vs 1.7 prev. On harder tasks)

given all the smoke and mirrors and OAI style fear-hype, it wouldn't surprise me if they intentionally degraded opus 4 for a few iterations, so they can resell "coke classic" at a markup with a minor quality of life feature put in, but charging way more than just re-attempting a poor output would have been previously.

unless anthropic starts acting in the image they claim and starts contributing to research, we'll never know either. Ultimately, the secrecy in how and why things are done would mostly be beneficial to this kind of buisness practice, since as it has always been, the moat is the data not the tech, so I cannot imagine what they hope to gain from the recent uptick in paranoia, jealous guarding and secrecy other than trying to huck a previous peak performance model as an imorovement when really, it is simply coke classic.

Grimblewald··on Ask HN: Why are so many "AI evangelists" posting such insufferable content?
> why is linkdin shit?

Because linkdin is shit? Ez.

← PreviousPage 4 of 30Next →