1dom

I think the issues you're describing are all modern internet software issues. I'd guess you're possibly a bit younger than me, or maybe forgotten how software worked before about 2010.

- All compiled software is static. New versions are released. I'm happy to pay for new versions/features if I want/need them. But if the version I have is working offline, it doesn't need security updates, and if it does what I need, it doesn't need new features. Same applies if I'm building software for others, if the library does what I need for my software for my customers and I need nothing more off the dev, why should I keep paying the dev?

- I'm struggling not be sarcastic here, sorry, but this might shock you to hear: most humans find the idea of being able to try whatever they want and get paid for it appealing. That's not really how money or life works though.

- I don't know how to respond to this point, it's a bit of a strawman. Yes, if the only other option is to pay $50k up front for a local model, then your point is valid.

As a kid, I spent literally hundreds of £ on the first Sims, and Sims expansion packs at £20 /pop. EA got loads of money off me, I got loads of new features I wanted. I moaned at the time, but on reflection, everyone was a winner. Sims is still going now (possibly as a subscription!)

Again, apologies for a bit of sass here (that's a me problem), but the number of people who seem to sincerely believe the points you've put forward is why I'm personally sad at the state of modern software economics.

This is a cool way of approaching AI models. I'm a big fan of local LLMs, local specific models like this should be even more powerful.

> Every model is free up to 100k monthly active devices. No tokens, no logins.

I dunno about the business model though. Cloud LLM billing makes sense: you're getting another computer to do work with each request, and using their compute via their gateway that they bill you.

These local models are like old school software. They're producing the weights, and then giving them to people. If I'm happy with the weights you've given me, and I'm not using your compute for inference and not wanting or needing any updates off you, why should you continue getting money off me and my customers?

The whole "but we need to keep it updated for your security" doesn't really work as well for software designed to run fully offline like these local models are.

I'm not saying they shouldn't get paid, but I guess I feel a personal sadness that it's less obvious how to successfully monetise such a sincerely useful and beneficial approach towards AI models.

Yeah, I totally agree. I think there's no 1-size-fit-all solution. I think the problem should shift onto "how many solutions do we need"?

At least 2 I'd guess: 1 is going to be incomprehensibly hard and painful compared to now, but there will be maybe only a few handfuls per country. Sure you have to sacrifice your first newborn, but it will come with the full weight of government PSAs because lots of normal and vulnerable citizens depend on it.

The other solution matters less so, because it's could all just be optional/nice to have things, and if people want them, they have to accept some amount of risk/competency in exchange for easy of access. The current solution getting only <50% spam/scam seems to be doing better than traditional mediums so is probably fine for that group.

Maybe there's some more measure solutions between the 2 extremes, but I really don't think there can or should be 1 fix for the DNS system given the range of use cases and customers.

Regardless, Dave down at the local is always going to panic click amazaoncom.paercel-delay.info regardless of what the local bus stop ad or government tell him to do. Such a Dave thing to do.

All good thoughts, thanks for the response!

Maybe the solution is in the offline world. A government funded public service announcement of tv/radio/print/busstop ads saying stuff like "never do government stuff not on .gov, never do banking stuff not on .bank". I think that would be highly effective for the sort of people who need the protection here.

It would obviously require banks and gov departments to get their ducks in a row first which is a whole other problem in itself, but it might still turn out to be an economically viable improvement.

I assume this is supposed to be somewhat tongue-in-cheek, because the argument that SoMeThInG mUsT bE dOnE aBoUt DnS because a whopping 10% of registrations were associated with spam/scams seems silly.

There are lots of large open systems that the average person interacts with daily that would love to see <30% spam/scam volume:

- email - paper mail - telephone numbers - SMS messages - basically any social media platform

> I don't want to live in a world where I have to show my passport and pay thousands of pounds to register a domain which is only available after being vetted by private interests. But I also don't want to live in a world where scammers have effectively no deterrent from abusing millions of people.

One solution would be to have recognised verified TLDs that require some verification on some, whilst allowing others to be more lax an accessibel. The issue is we have these, e.g. .gov.uk.

Another solution would be to dissuade people from using less well known gtlds in favour of ones that have some sort of limits/controls, such as recommending people avoid .mobi domains in favour of ones with more oversight like .com. (I say this as someone with a .fun personal domain!)

I don't know. I'm not saying this isn't a problem, I'm just saying that Terence kicking this nest seems like the start of some monkey paw meme or something.

From what I can understand from reading a few different, slightly conflicting, versions of these events: they weren't given write access. They found a zero day exploit that allowed them to create folders, and the folder names were initially used for agents to communicate.

I'm not sure artifactory was connected to the net. Some agent sandboxes had internet access and were able to communicate with ones without access via artifactory.

I think you're wrong to say "you decimated the thing they care about". The way I see it is those loathing people have a job to care about a thing. The people who gave those loathing people their job also gave another job to another person who's just doing their job too.

Everyone has a job with a remit and responsibilities. The role of business leadership and management is to make sure the people and remits operate harmoniously to efficiently reach a common goal.

If some people doing their job well decimates another persons job in the same company, that is 100% the fault of the people managing the jobs, not the people doing the jobs.

I learned about https://en.wikipedia.org/wiki/Assassination_market in the early 2010s. It wasn't phrased like a prediction market focusing on when a person would die, but more like a bounty. If the bounty/pot got big enough, eventually it would be worth it for someone to do it.

Even thinking/writing about this now, it still makes obvious sense and I'm surprised it hasn't been seriously made yet with crypto: if there's a pot of billions, contributed to by millions, the incentive becomes not just about the money, but the adoration, protection and power that would come with doing so many people a favour.

They're fair points. I do respect their ethos and what they're trying to do, that's why I hope they have more success than they're having at the moment. I do want to give equal weight and time to that. But at the end of the day, I'm buying a modern phone, it needs to behave like a modern phone.

FWIW, my understanding is the FP6 seems to be notably less successful and "just works" compared to earlier models.

It isn't really slow. It takes average pictures, and sometimes loses them. It regularly does annoying little freezes and bugs.

The FP6 has average spec, average performance, more bugs than average, worse/slower support than average (at least on their own forums) and a higher price than average.

I believe fair stuff done right should be at least as good as average, and I'm fine to pay more than average for that. When I pay more than average and get less than average, and looking like they're ignoring/avoiding problems in their own forums, that feels as though someone isn't being very fair towards their customers.

Sorry, I strongly disagree with what you're saying here, and I think you're part of the problem. I bought a fair phone because I believe treating people and things fairly genuinely leads to a better output over time - it should, and it hasn't with Fairphone.

I'm not trying to have my cake and eat it, to extend your analogy, we both had the cake, we've both eaten it - it's the same cake. I'm saying I appreciate the cake was fair, I appreciate the staff were paid well, I appreciate good fair ingredients were used, I'm glad I bought the cake, I know why I did, because it aligns with my ethos.

All the necessary things that went into an average cake - and more - have been put into this cake. My point here is I don't understand why it tastes worse than the average cake. Based on my view of the world, it should taste at least as good as a regular cake, because it has more money, at least similar quality materials, and should have higher quality, more passionate labour involved.

I feel like you don't understand my comment because you've taken personal offence to me providing critical feedback about the taste of your favourite cake. I like your cake vendor, and the cake vendor cause. This is why I want them to succeed, which is why I provide the honest feedback.

If you say you can do something well, it just needs money and fairness, and people give you money and fairness, and you don't do the thing well... that's not really fair.

This makes no sense to me - what's the opposite of AI astroturfing? Why are people complaining about it?

Ironically, your comment could easily be a failed comment from an astroturfing LLM tasked with criticising/disagreeing with people calling out LLMs.

FWIW, I checked your profile, I don't really think you're an LLM. The internet is hard now - sorry :(

Thanks for this. My main hope is that as PMOS become daily-driverable, the nature and number of irritated Fairphone users will quickly make PMOS a better alternative to any of the Androids available on FP6.

If that happens, the resultant hardware/software/freedom combo could be well worth the cost. I hope Fairphone make some big donations to PMOS if this happens - it would not be very fair for the people responsible for a buggy phone get all the money whilst the PMOS people who make it great, fair and open get nothing.

I feel the votes and excitement in this thread aren't from people with a Fairphone 6 or PMOS.

From the PMOS wiki: https://wiki.postmarketos.org/wiki/Fairphone_(Gen._6)_(fairp...:

> No internal speakers or microphones are currently working. Media sound (not calls) is working if connecting a Bluetooth or USB audio device.

I got a Fairphone 6 last year, after having a OnePlus 7 Pro which I bought in 2019 and still worked flawlessly, apart from not having received an update in several years.

The Fairphone 6 works. It's slower and more buggy than a 7 year old flagship though. I'll probably use it until it dies in however many years time, but my overall experience with Fairphone has been that I've paid a huge amount for a below standard, poorly supported, but apparently fair phone. The Fairphone forums are a trainwreck too.

I really, really wish Fairphone was better, but the entire company experience is so janky for how expensive the products are.

A prime example: I thought I lost a few eclipse pictures because they weren't showing in the gallery after I took them. I'm pretty certain it's this same bug from over a year ago: https://forum.fairphone.com/t/fairphone-6-camera-pictures-mi...

Me and my partner both got our oneplus 7 pros at the same time, and hers also lasted. She upgraded at same time and just wanted/needed a phone that works for work, so she opted for the Pixel 9a instead. It's cheaper, faster, more reliable and better supported.

Sorry to rant - it distresses me a lot when good intentions take a lot of money and deliver a relatively subpar product and service. I guess this is the price of principled tech. It's really not great, unfortunately.

Claude should modernise this. It'll come out looking like all the other results when you search an image engine for "AI landing page": dark, with purple, blue or other neon accents.

I provided you evidence but you ignored it, so I'll restate it.

In the article that we're all here to discuss, it talks about an unpublished novel that is unpublished specifically because expert publishers couldn't determine if it was AI. I'll quote from the source that the article references:

> In an email obtained by the outlet, Hodgman said that the “incredible” manuscript “dazzled” publishers around the world, but they failed to authenticate the evolution of the writing. The email adds that while they vetted the submission and were reassured that AI tools weren’t used, they could no longer substantiate any claims.

It's a dazzling manuscript according to publishers around the world. Seasoned publishers couldn't be certain if it was AI. This is literally an article about the exact thing you're arguing against - I can't be more specific with my evidence.

Ultimately, I feel you're not trying to come to a correct conclusion together like a healthy debate should. You spent half a sentence dismissing what I was saying and then paragraphs talking about how certain you are about your own opinions on AI writing, whilst providing no real argument or evidence beyond rhetorical questions.

That's not really what I'm here for, sorry.

This feels about 2 - 5 years out of date now. There isn't enough dark theme with slop-neon accents.

No need to worry, I just visited. Now you know.

Seems like the sort of site and person who should have a guestbook - I would have left an appreciative and encouraging comment there, but this will have to do:)

It's far quicker/easier/more convenient to sit in front of a screen and try find some buttons. Your issue seems to be that those buttons have got harder to find and understand - that's a fair point.

My point is that - overall - clicking some buttons and waiting a few hours is far more accessible, faster, easier and less disruptive than what my parents had to do 20 years ago:

It required far, far more effort and thinking and action to find some physical, expensive, modern hardware, digitise signatures, and the post things off.

"Faster" is very relevant in a discussion about convenience and effort. The important point is how convenient it is for the child and the adult. Now, a child has to sit for an hour watching you fumble around on a screen. I had to wait for days and days whilst my parents fumbled around with cables and when to the post box and other stuff.

> the amount of time actively spent dealing with these things.

My parents used to have to spend more time actively dealing with me and my things whilst I was waiting for the post.

We're not going to agree here. You're a parent, complaining about how much time you have to spend supervising your kids. I sympathise for you, the modern world has a whole lot more expectations of your time which my parents didn't have. This is one of the many reasons me and my partner chose not to have kids, sorry.

I wasn't making a comment about toupees, I was referring to the toupee fallacy.

Look, I don't know who you are, or what your background is, and without that, there's nothing I can do to convince you that AI is being used in writing and you can't identify it.

The fact is there's a well-known author saying from their experience that the writing produced by LLMs can be amazing, that's the article we're all here to discuss. I've then included a link to another article where a whole bunch of publishers couldn't tell if AI was being used.

This means either you're confidently know you're a more discerning reader/writer than a well known author and a bunch of publishers. Or, you know you're not, and you're choosing to believe your feeling that you know better then them, which I'm not really interested in discussing.

To be clear: I'm not saying AI doesn't produce slop, I'm saying it also produces stuff that the vast, vast majority, even of experts, cannot identify as AI produced. This means anyone who suggests they can easily identify it now, or "all AI stuff makes me vomit" is probably needlessly deluding themselves. I think this has only been the case for the last 6 - 12 months.

> If this is the case it should be quite easy for you to provide examples of excellent AI-fiction.

I feel that's the opposite of the toupee fallacy. The lack of labelled excellent AI fiction supports the idea that there's a lot of excellent fiction that was written by largely by AI but nobody was able to identify it as AI, because they were looking for writing that makes them vomit.

Regardless, how about this: https://deadline.com/2026/07/ai-questions-debut-novel-book-f...

I don't mean to be confrontational, but the key detail you thought I missed, about it being complicated, was a fairly crucial part of my point that you've overlooked:

My parents when I was young ended up having to figure out how to use digital cameras, scanners, printers, fax machines and then print, sign and sometimes even post off a letter for me to access one website which I wanted, which could take hours of their time in a process that took literally weeks. This was in the early 2000s, when scanners and digital cameras weren't exactly user friendly. They had to do this sometimes multiple times a month. Every website was different, there was no universal permission slip. But they did it. Eventually, they too got bored and gave up, or maybe I just grew out of it.

Point being, the tools have always existed, and always felt inappropriate or convoluted, but my parents at least were happy to do it for their kids 20+ years ago. From what you're telling me, they're now a lot faster and easier to access and use.

Even if the tools were made to take only a few minutes to setup, there would still be a vocal set of parents unintentionally arguing that everyone should suffer with overbearing ID tools so they can spend less time supervising their kids internet usage.

> AI writing makes me vomit.

I think this was fairly obvious and fair 1 - 2 years ago. Now though, I have to wonder if people who think this with such conviction are falling for the toupee fallacy.

You said it yourself: someone you consider a well known (and presumably accomplished?) author thinks AI is pretty good at writing.

I might be being overly cynical here, but I think part of the problem is parents expect to give their kids the internet so they don't have to spend so much time/attention with them.

I think this means any solution which involves active parent participation is going to get shunned by a vocal subset.

I'm in my mid 30's. When I was very young I remember my Dad semi-regularly having to email or even fax off permission slips to websites aimed at kids before they would create an account for me.

Sorry, it wasn't meant to come across as a complaint with copy and paste curl command, I agree with you it's convenient and I generally like it.

However, the sort of tools that generally provide it also generally have a GitHub link nearby to encourage reading/writing the source. That too is convenient and I generally like it. Seemed like a fair suggestion.

This is a really cool idea, and very appealing. Also very nice that it looks usable with just an SSH client. It would be useful if it was explained what the companion client is for?

Another small personal gripe: I think you should either remove the 1 line copy/paste sh install on the homepage, or add a link to a git repo on the homepage. I know it's possible to view the sh by visiting it, but a raw text file doesn't give any info about contributors or development process. Current form of no git repo and copy/paste bash pipe on homepage feels like a dark pattern.

It's possible to login and interact and use late.sh with just your SSH client - there's no need to download or run custom code. But you that did make me realise that they don't have a github link on that page anywhere.

Your tone in this comment is a bit out of place on HN. Why are you here if you think it's 1M dummies?

That's fair, I completely get that. I guess I was just surprised that there's an expectation that a website with content like Gwern's would be up front about their real identity. I'm trying to understand how widespread that expectation is now, because I'm concerned I'm becoming out of touch with modern web expectations!

I think you're inferring something that people aren't implying, and that inference leads to you eventually discussing how ATProto is better than mastodon. I think this doesn't look great, particularly at a time where there's quite a few posts of about unproductive conflict between open social network contributors. I don't think there's any bad intent or malice on your side: there is value in comparison between AT/AP to help understanding.

Your argument that there are no instances in ATProto isn't flawed if you redefine instances as this thing that ATProto doesn't have. But if I run it by the various computer science definitions I can find (e.g. https://en.wikipedia.org/wiki/Instance_(computer_science)) then your argument seems like it depends on a niche interpretation. So it can be technically correct, but just irrelevant or inconvenient when used in a comp sci discussion involving the word "instance".

But your blog post on the subject is interesting, and informative, and has some great diagrams, and the original parent poster agreed and learned stuff from it. Despite that, I think the implication in the parent post was, IMO, "so the main, largest implementer (instance) of the protocol is also a for-profit which owns the trademarks? I never realised that, I'd assume the protocol was governed and run by a not-for-profit independent governance org, charity or consortium" (like some other open protocols).

I think that's a fair and slightly critical implication that has nothing to do with network topology. I completely get how you would want to take issue with that implication alone, because without context like the trademark was acquired with good intent to protect protect openness, and that the Bluesky org is a specific type of for-profit (PBC) that expresses intent to favour public benefit and openness, it makes things sound more sinister than they are.

Ultimately, from an openness and independence position, I think it would be better looking from the outside if the ATProto and trademarks were governed and managed by a separate not-for-profit entity that wasn't tied to the biggest instance/implementation of the protocol. I think that was the point being made, and I don't think it's absurd. But I think refuting that argument by focusing on network topology and instance definitions seems like a misunderstanding somewhere.

Dan, you really should refine this line of reasoning a bit as it feels not very nice in some way: it feels like you're picking an unintuitive/uncommon definition for "instance" and then using that to tell everyone they're misunderstanding stuff then they're using different, arguably more widely accepted, definitions.

We've had a very similar discussion in response to this same argument recently elsewhere on HN. If you enjoy having this debate, then carry on, but I don't think it endears non ATProto users to the ATProto cause. I think it seems needlessly controversial/pedantic. Sorry for the blunt feedback, it's because I expect better from you:)

Gwern has been around for ages. I remember citing a few of their pages during my postgraduate work over 15 years ago. At the time, it wasn't easy (or even possible?) to find their name at all.

I've always assumed Gwern was anonymous because the norm on the internet generally was to try to be anonymous, especially those who've done research around crypto, darknet markets and other spicy areas.

Maybe I'm just getting old and losing touch, but is the expectation now that you can only write long articles and ask for donations or contributions if you publish your real name somewhere?

Could I ask what sort of age group you are?