Most articles I click on in the HN homepage turn out to be written by AI, judging from the phrasing. I'm weirded out by the fact that people don't seem to find it important to write their own thoughts down. The writing in TFA is clearly supervised by a human, but still, the wording is not human at all.
It’s a bizarre feeling isn’t it? Sorry you’re having to defend the act of thinking.
The problem is you can’t defend it right? Someone could say your evidence came from a prompt:
“Take this article and reverse engineer a hypothetical unpolished first draft written in a mix of Russian and English”
I’m not sure what the right answer is here. Fwiw I have no doubt you wrote it unassisted.
Chain of trust from RFID chips embedded in their fingertips that authenticated to their keyboard, proving that at least their fingers grazed the keys that formed the message.
But what if they're reading off of a pre-written message?
On one hand, you’re right. On the other, it’s normal that humans want to gauge the authenticity of the things they interact with. Some sort of uncanny valley thing.
Ive been using translation tools a bunch these last few years. Nobody seemed to have any hate for better accessibility.. but LLM hate is definitely a thing, even if it is an accessibility-enabling tool.
Tbh, I'm getting more frustrated with the ever-coming flood of "Bah I didn't read because it was obvious AI blah blah" which seemingly every single submission HAS to come with nowadays on HN, god forbid someone is more interested in the content than the flow of the words.
If you have specific complaints about the text and content, bring those up instead, and we could discuss those or even correct the linked page itself, as it seems to be a wiki. But general complaints that could be copy-pasted for any submission, just so you can feel heard about that you think this was AI written, gets so tiring to read for every submission.
It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
It’s like submitting a 10 page pull request to someone and then getting mad because the person didn’t give comments on every single snippet of code. The issue isn’t the snippets of code, the issue is the attitude that led someone to believe a 10 page PR is appropriate to begin with.
> It is unreasonable to expect “specific complaints” about AI vomit like this, because one of the main issues with AI content is the ability to generate an overwhelming amount of it. It’s simply not feasible to give specific criticisms, because the criticism is with all of it.
But how would that make the "I won't read this because it feels like AI" comments more interesting to read?
No one is forcing you to read this stuff, no one is forcing others to read this stuff as well. When I come across text that isn't great, for whatever reason, then I close the tab and move on with my life. Do I have to make it clear to the world what I think of the text in that specific article? Not really, it'll continue spinning like before, and people who want to read it will read it, others like me will just close it.
It sucks that even if the topic of the submission is interesting, here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written" or not in the comments, although I'd hope it'd be considered vastly off-topic.
> No one is forcing you to read this stuff, no one is forcing others to read this stuff as well
The front page of HN is limited real estate. I visit HN to discover and read interesting and quality content. Whether or not I am “forced” to read it, every piece of AI vomit that’s on the front page is taking a spot away from the real human content that I (and others) really want to see.
> here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written"
I genuinely find this discussion in the comments to be of more value than reading the AI content in the article.
People will discuss the content in front of them. If you don’t want that discussion to be about AI content, then the solution is to not submit (or upvote) AI content.
Even more precious than HN real estate is the time of (how many HN readers are there?) unknowingly spending their time to read something that wasn't even worth 1 person’s time to have written themselves. (In OP’s case they said it partly came from Russian and provided the first draft so I'm more understanding.)
To expand on your previous point, "because the criticism is with all of it", I think the criticism is really with the HN community allowing so much of it to reach the front page.
A little bit would be tolerable, but the ENTIRE front page is garbage like this now.
> led someone to believe a 10 page PR is appropriate to begin with.
Agreed, a 10 page PR is not on. But the original article, though evidently touched up, was appropriate in length and scope. What's your real criticism here?
I was hesitant to post my comment. It's the first time I've complained about this on HN I think. And it's not only about the flow of the words at all, it's more about reading something that no one wrote. Especially if it's about a project that seems interesting, having AI written text tells me it's maybe not the passion project I otherwise would think it was.
Just commenting as a friendly FYI - the author commented above and noted that there was no AI used, just translation tools. Honestly, I'm not sure why the grandparent thought it was AI; it didn't read that way to me at all.
You're right to complain. Writing code whose principal job is to be compiled and executed by a computer is not at all the same as writing prose whose job is (hopefully still) to be read by a person.
Up to a couple years ago, the latter was essentially a product of lever-less human attention.
So because this article seems AI written to you, this business and project which is on it's second iteration and been around for years already, maybe isn't a project of passion in your eyes?
Seems like a huge logical leap to make, based on things that it seems you cannot even exactly quantify here, as you're still not pointing out what's wrong with the text, just saying that the text is somehow "lacking of soul" or something like that.
The criticism is that this is a respectable project, so when you read obvious AI tells like “Honestly, …”, or “Flipper One isn't an upgrade to Flipper Zero — it's a completely different project with its own goals” in the first few paragraphs, it’s distracting and takes away from the content.
A simple fix I use for AI writing is disclosing it. Here, a simple note that “this article was translated with AI assistance” would have made it much less distracting.
> If you have specific complaints about the text and content, bring those up instead
Accusing text of being written by an LLM is a specific complaint about the text. It's shorthand for "the text is overly verbose and uses the typical clichés LLMs are known for, which makes the text unpleasant to read (it's too much text and too many empty clichés) and also makes me distrust the text, because now I'm not sure anyone even looked over it and made sure it says what they wanted to say."
It's just shorter to say "this sounds like it's written by AI."
But still be bothered to leave a generic complain on HN, which you ideally can copy-paste across all potential LLM-written comments? Something doesn't add up there, don't spend energy writing the comments if you cannot even be bothered to read it because no one was bothered to write it.
I don't think the copy-paste dismissal is sound. Consider:
You can ideally copy paste your generic comment across all potential LLM-written-criticism comments? And I can copy paste this generic comment on all LLM-written-criticism-apologist comments. Something doesn't add up here.
It reminds me of high school, ages ago, when a friend would go on and on about how Depeche Mode weren’t musicians and how nobody cares about electronic music. I’m a little nostalgic for the hours, cumulatively probably weeks, that I heard about just how much he didn’t care about Depeche Mode.
Yes? There is nothing incoherent with disliking something and putting in effort to see less of it. "Ignore it" is an answer, not the only possible answer, and probably not the optimal one in the long term.
> But still be bothered to leave a generic complain on HN, which you ideally can copy-paste across all potential LLM-written comments?
I mean, I personally wouldn't specifically on HN, since it's generally unproductive conversation, but yes? You say this as if there is some gotcha or contradiction there, but there is not. It is far, far, far less work to write a short comment than to read pages and pages of AI slop.
But it has a problem common in AI, where it makes bold claims "we believe this is the only way to make a truly meaningful contribution to the open-source community and to education" without explaining, and too much filler ("...All the messy stuff companies usually keep behind closed doors. This is uncomfortable. We've never been this open before, and there's a real instinct to hide the unfinished work, the wrong turns, and the arguments...")
I'm mostly the opposite. I'm glad to see people calling this out. Do we really want it to become normal to offload communication to another entity?
"Claude, I need to send my wife an apology for shagging the secretary. Please make it tender and remorseful."
A person's take on anything isn't their take any more if someone else articulates it, and there's a real risk we slip back to a hired scribe culture, with the multitude volunteering to return to illiteracy because they can't be arsed to type or even speak - beyond brief outlines.
But the case is totally different for organizations and companies. They've always used copy editors to write their blurb, usually in a pasteurized flat business style that was always far removed from individuality and near-identical across organizations. I can't see why using AI in these cases makes any difference.
There was also a similarly common debate AI written/aided comments on HN until, ultimately, the guidelines were updated with an official stance saying they weren't allowed because HN is for human to human discussion. Honestly, the same kinds of comments and meta-complaints would occur for any of the things the guidelines comment on. It doesn't mean those common complaints would be wrong to have, that's part of how the guidelines get formed, it just means we haven't figured out what makes sense or not for the site yet.
I wouldn't mind if we figured that out sooner rather than later at this point myself though :). Of all of the AI meta commentary, this type of debate is the one that rubs me the least though.
I appreciate these comments, because they're a warning. If I'm on the fence about whether a link is worth a click-through or not, I'll have a peek at the comments first, and when I see something like this I don't bother (like with this article).
If it's just long-term generated text, why bother posting the link at all? Why not ask for a bullet point summary and make a text post? Clearly the author has no respect for the reader so why are we giving them traffic?
Okay. So if I copy and paste an AI response written by Claude and you can't actually find a specific problem with it, are you still fine with that? If so, please start your own damn website and enjoy talking to AI and reading AI text all day. I'd really really rather not.
They are tiresome but also understandable. I do not want to read AI generated content, even when its correct, because at that point what's the value? I'm reading results of somebody else's prompt, might as well use my own.
I'm surprised any author today isn't pre- or appending their articles with simple statement on AI usage. Transparency goes a long way.
Agree, I find AI valuable for writing reports on the behavior of features in our codebase, but I’ve started sharing the prompt at the top of the file before sending it out, and reviewing the content line by line to catch obvious errors.
I like being warned about AI generated content before I waste time reading. If the author couldn’t even be bothered to write it, it’s a good sign I shouldn’t be bothered to read it.
"Make me"? My god the entitlement... It's either free information, or close the fucking tab, no one here or elsewhere owes you anything, if you don't like it, why are you forcing yourself to consume it? Personally I just close the window/tab when the text isn't interesting/high enough quality, LLMs or otherwise, I'd suggest you'd learn to do the same if this is the first time you're using the web.
I guess it's the same with "I rewrote blah blah in Rust," where everyone knows it was vibe coded. I don't think that's necessarily a bad thing, but Hacker News is a forum mostly read by people who enjoy hacking and building things. "Vibe coded codebases" and AI generated text generally aren't praised here, although they certainly are in other places. Or maybe it's just a matter of time until hackers get over or used with it. Time will tell...
It’s the fat introduced by the process that annoys me the most. The user of the LLM had an idea, but it got greased up and packaged into something that the average person would create, not a specialist in the domain.
It dumbs down everything into a single perspective / way of presenting a topic.
> If you have specific complaints about the text and content, bring those up instead, and we could discuss those or even correct the linked page itself, as it seems to be a wiki. But general complaints that could be copy-pasted for any submission, just so you can feel heard about that you think this was AI written, gets so tiring to read for every submission.
No. And the reason is pretty simple: if you couldn't bother to write it, why should I bother to read it?
And that's the problem with AI: it creates floods of that stuff and makes it hard to differentiate the good-faith use from the bad-faith use. The default can't be "reader, waste your time, even on garbage." A reader-respectful norm needs to be set, and those comments you complain about are part of that. The people making these things need to learn that they've got to put in the work if they want to be read (at least by serious audiences).
Yeah, I'd be fine with it if every AI-generated posted was required to have “AI gen:” at the beginning of the title so that readers could make an informed decision about whether they should spend their time to read something that was not worth even 1 person’s time to write.
It is exhausting to always have to read word salads with little content.
Every single fucking article with 20 lines of introduction before you get a chance for actual content. LLM slop then dilutes the information, and LLM slop always read the same way. You know, how easy it is to spot LLM generated content, it is actually refreshing when you can tell it's a human.
> It is exhausting to always have to read word salads with little content.
Agreed, but you know how others solve this problem? We close the tab, move on with our lives, without feeling the need to leave the generic "This seems like it was mostly written with LLMs" slopplaint HN comment.
Why is it okay for you to post a comment complaining about people posting comments complaining about AI posts? Why don't you just move on with your life instead of posting a complaint on HN about others' complaints?
> because 70 of 140 comments under this submission are owned by this thread about AI.
This is an effect, rather than a cause. The root cause is often (but obviously not always) that the submission was written with AI to begin with. In instances like this, it is useful to focus on the root cause, not a proximal effect.
> And this is usually not what you want when you click on an interesting submission
More importantly, AI-generated output is usually not what you want when you click on what you thought would be an interesting submission.
In general, reading comments written by actual humans about how a submission is AI, is preferable to reading a long post written by AI. If I wanted to talk to AI, I can do that without HN. HN is where I come to discuss things with people.
I think this is due to lazy prompting. It isn’t hard to get an LLM to write concisely, with a logical flow and to be direct with the point you want made. I’d rather read something an LLM has written in this manner than a lot of things I come across written by humans.
Regarding padding out word counts, I see this more often in newspapers and magazines than I do in AI-land. It’s like Netflix shows trying to meet an 8 or 10 episode minimum - horribly boring with unnecessary filler.
On the one hand, I get what you mean. Some genuinely interesting projects are immediately dismissed because AI was involved.
On the other hand, I have two real problems with AI writing.
1. LLM prose is genuinely unpleasant to read. Its the exact same way that I strongly dislike reading LinkedIn posts or email marketing copy. It's all the same slimy tone that's using a certain sentence structure and rhetoric to try to be interesting without real substance.
2. Sometimes it feels like someone asking you to read an article with no punctuation or grammar: the author couldn't put in time/effort to make this enjoyable to read, so now I have to spend more time/effort reading it.
Personally, I don't read through all marketing copy to see if "this one is going to be good", nor do I want to spend time providing constructive critical feedback on it.
What exact parts from the submission are "genuinely unpleasant to read" right now? Highlighting those could make it better rather than just filling HN with "LLM texts is boring to read".
> Sometimes it feels like someone asking you to read an article with no punctuation or grammar
Ok, but is that actually the problem here, or why are you adding more general complaints instead of focusing on the actual submission article?
If you don't like it, don't read it, don't contribute to the discussion, I don't understand this obsession with "must let others know I don't like LLM writing, although I'm not 100% sure this submission actually suffers from the issues I don't like with LLM writing".
It's not just short-sighted of <these commenters you hate>; It's self-destructive!
* It's the job of the consumer to correct and edit the content they consume
* Content creators have it hard enough ——— prompt-crafting and imagining transformative and disruptive new horizons in tech
* So what if the prose is 4x longer than it should be? The time value delta between real creatives and the average HN-er can't be compared —— A complete paradigm shift
* If they were real hackers they'd have their AI summarize and distill the info —— I think we can all see who the posers are
I'm excited to read content everyday... 'slop'? That's a coward's word, I see past the prose into the core of the data space, and I'm stronger for it.
Odd, I just read the entire article and never felt that way at all. When I'm reading AI generated text, it triggers something in my lizard brain, but this didn't.
I read it and understood the project goal and the difference between the old and new versions. What else is there to get from this? If I want to read good prose I have plenty of books to pick from. This is just a product pitch that effectively communicates the idea.
IMO the article was a great intro to the project and I really like how the thoughts were laid out. I got a lot of food for thought from it and I'd recommend that people read it. I don't care how it was written.
AI can produce interesting thoughts just like you can produce meaningless flamewars.
Repeatedly saying the same thing with slightly different phrasing: "Flipper One isn't an upgrade to Flipper Zero", "Flipper Zero and Flipper One are completely different projects", "Flipper One doesn't replace Flipper Zero"
In my experience, the bulleted list with emojis is usually a pretty strong tell (the one in the article just after "We call these parts sub-projects"). LLMs (maybe just ChatGPT) love doing that.
"Honestly?" and "not just x, but y" appear once, and only half of the lists have exactly three items, making part of your comment factually incorrect; did you just not look closely or did you jump to conclusions because you have an agenda / axe to grind?
I just long for some sort of attestation system where, if you want to use an em dash, you must first drink a verification can or eat some verification doritos to prove you are a meatbag with a digestive tract
Some people -- many people, actually -- use em dashes in their writing. Real people even use contrast to make a point on occasion, believe it or not. It's not AI writing that bothers me, it's the constant complaining about the supposed tells that prove something is AI-written.
The other day I was criticized for posting a comment people thought was AI but was actually not.
I’m starting to notice that more often with others as well. Happens sometimes to those who were always using emdash, sometimes to those who happen to have traits that these machines themselves learned from how to write, and now they sound suspicious.
I don’t think this means we should never call out slop or lazy writing, but it does seem our ability to detect this stuff is on a spectrum. Some of it is obvious. But beyond a certain point, for example with this article, the signals can become too weak to make any strong claims.
It’s disconcerting to admit that we’ve come to a point where it’s possible to be completely fooled one way or the other by what’s human or AI. Lots of stuff we can still detect, and sometimes it’s obvious, but at the margins we can no longer reliably discriminate.
I'm actually confused by how people even use LLMs to write these articles. They sound so synthetic that I assume the LLM wrote most or all of the text, but how?
Do they just write a bullet list of notes and then tell the LLM to go wild?
Y'all have become these super annoying human captchas where I have to proof that I am actually a human being who writes their own thoughts in their own words, just because you feel like accusingly saying: "But you used AI for writing!"
It's getting super frustrating and annoying.
Yes, loads of articles are written with AI. So what? Don't judge a fucking book by it's fucking cover.
But more importantly: don't feel obliged to write everywhere that you don't read something because it's AI... Just don't read it.
Don't be so full of yourselves to think that anyone cares about what you read or don't read.
Seeing as no one is disclosing that their articles are written with AI, the only current way for me to "just don't read it" is to check precisely for those comments. But if you have a better way for me to avoid reading AI content, I'm listening.
> Don't judge a fucking book by it's fucking cover.
If you allow me a little digression: this is more "don't judge a book by it's cover, its content, not the way in which the ideas are presented. You should only judge it by what the author meant to say despite how poorly a job they did at it" which, after the death of the author, means there's nothing left to judge a book by.
> Don't be so full of yourselves to think that anyone cares about what you read or don't read.
Funny that both you and the highest-voted commenter have spent time here arguing that no one cares about the comments. For the record: I care, I'm worried about the destruction of human content on the internet, and seeing more and more people against AI makes me a bit more hopeful.
If you have to spend a few paragraphs explaining that One isn't a replacement for Zero - they they are different classes of product, you know that your product naming is a problem,
This looks flippin' amazing, but also like the definition of project scope creep. I imagine it will be brilliant, unaffordable, surprisingly cheap, terrible and awesome (in both senses of the word) all at the same time. 3GPP really needs a light shining through it.
I sincerely hope I work out a way of getting someone else to buy the thing for me. And the push towards all in-tree source is fantastic. Genuinely impressed.
Some projects are meant to scope creep. Like this one. If the project manager of the swiss army knife had defended it from scope creep it would have 1 knife.
> but also like the definition of project scope creep.
To me it seems like the opposite, it has more connectivity and I/O than the Zero, but also scaled down, while using better materials, like they decided to outsource the project scope creep to the community, which makes sense to me.
Man, they put 2 processors in the thing and are building their own OS. They even say they are not sure how to get it accomplished.
Scope creep to hell and back. Could just let the device get turned off like literally any other device on earth, and not have to build a whole new fucking OS to get it running.
They even - for some reason - want to waste time "training their own AI model because general ones don't cut it" (which no one is likey to use). Could just build a normal RAG + context stuffing pipeline in an afternoon but nah, let's devote a few months to this completely unnecessary non-feature.
100 bucks say this doesn't see the light of day before 2030 (if it ever does!)
> Could just let the device get turned off like literally any other device on earth, and not have to build a whole new fucking OS to get it running.
This is actually quite common in embedded devices and even elsewhere. Every Apple device does this, for example (the Secure Enclave is a completely separate OS running on a separate computer).
If you think about it, most laptops have been doing something like this for decades as well for things like brightness control etc., not with a different CPU but definitely an OS-like thing (i.e. the BIOS, using SMIs etc.)
The idea of the "single OS, single CPU computer" has been a myth for a while now.
First time I've heard anyone call the Flipper Zero "simple" and "focused", most people seemed to have considered it a "swiss-knife" meant to just house a bunch of features and radios, meanwhile the One has less features but more connectivity and I/O.
But apparently you're not alone in feeling this, but I don't understand what from the submission makes you and others believe so, what exactly gave you this impression?
And this one is an 8-core Arm computer and the project has ambitions of some notoriously difficult things: no binary blobs, full mainline support (including a NPU), reinventing small-screen UI for more serious handheld computing, and supporting a ton of high-bandwidth interfaces.
Agreed. This will likely never ship with all the bloat. Custom AI model, custom OS, extremely custom architecture (2 "main" processors running independently...), barely reuses any of the previous work from Flipper Zero.
On the other hand, this has been working pretty well for the first few Raspberry Pis! (Although they had the benefit of leveraging an existing smart TV based platform for that.)
Can someone explain why Flipper is making these decisions, or what advantages Flipper One has vs a Flipper Zero, RPI, and Linux machine?
The (EDIT2: maybe not) AI writing doesn’t help.
EDIT: looking more, it seems like the goal is to be a fun project like Playdate, except a Linux multi-tool instead of game console. Which is actually great, a step towards healing today’s corporatized tech culture. It’s unfortunate that the website non-explains this with AI and marketing speak.
EDIT2: I wrote too soon, AI is making me too cynical. My only remaining critique is that they explain the motive instead of just stating features and repeating “we’re doing something exciting and important [for reasons not really explained]”
> what advantages Flipper One has vs a Flipper Zero
They work at different layers, the Zero is physical, the One is network. There is almost no overlap between the two, so one doesn't have an advantage over the other.
> RPI
It has a battery, with attention given to power management, and is a complete unit, not just a board.
> Linux machine
You mean like a laptop? You can probably do all this on a Linux laptop PC, but the Flipper One is a smaller, more specialized device, with a firmware as open as the manufacturers will let them.
> My only remaining critique is that they explain the motive instead of just stating features
Can't answer for the One, as I don't think even they themselves know what it'll end up being when done, but for the Zero, the biggest benefit have been the whole "one device = one large community = lots of firmware = lots of software" thing which gets a lot of benefits from one cohesive community around one device, I'm guessing the One would also get similar benefits with this.
As a current Zero user, I'd definitively get a One once available, just the addition of the PTT-button feels worth it to me, but almost all the other changes are good (IMO) as well, don't really see any drawbacks from the design they're aiming for now, besides the modularity will make things slightly more complicated, but also comes with a ton of obvious benefits.
Can you elaborate on how you use the zero? I got all excited, bought one, and it’s in a drawer. I’m way deep into coding, CNC machining, making of all sorts… but I just never incorporated it.
Mine is mostly just lying around but sometimes I find some use for it. This winter I bought some remote controlled electricity sockets that at first didn't seem to work so then I got the Flipper and started recording radio to figure out what was happening. Turns out the remote was some cheap hardware that at first broadcast promiscously and to the sockets entirely unintelligibly but with time and trying it stabilised.
If I didn't have the Flipper or some other SDR device I'd probably have assumed it was bad and left it at the recycling station. If I'd lose the original remote I can use the recordings on the Flipper to either control the sockets or create a new remote.
I've also looked into how the key fob to my car works and investigated tens of RFID and NFC cards, some of which I could probably have talked to with my phone but I like the format of the Flipper and it has very few distractions except Snake.
When traveling I sometimes bring it up just to check out what radio stuff I can find and think about what devices might be sending.
Anything that anyone ever writes from now on has people coming out of the woodwork to accuse it of being AI-written. I too bemoan what the written word is coming to, but I am also so over the Slop Police, and wish they would just keep the conclusions of their sleuthing to themselves from now on.
I think this is the optimal outcome of the “Slop Police.” Normalization of these acknowledgements. Transparency is good, like a journalist declaring whether they have vested interests.
I usually give the benefit of the doubt, and regret accusing this article. It's the articles and comments that are obviously AI and score 100% on Pangram that I still feel should be called out, because the writing is hard to understand and the underlying message rarely makes good insight or discussion.
I imagine you dump all the config registers of a running system, and then adjust everything that looks like some timing or drive strength parameter upwards till it stops working properly, downwards till it stops working, and then choose a middle value.
Do that repeatedly for every parameter pre-boot, and then use that config. Perhaps redo that every few hours or when the temperature changes.
Does anyone know why the binary blobs cannot be reverse engineered in the age of AI and recompiled to closely match the original source? Is it for legal reasons? Is it firmware signatures?
Many silicon vendors, when providing said binary blobs to a device OEM or even just documentation or source code for the binary blobs, will make companies agree to a license or other legal terms which prohibits reverse engineering. Often the direct recipient of the binary blobs (the OEM of the device) cannot legally let their employees nor contractors perform the reverse engineering.
Generally, unless a similar license or legal terms are required to be agreed to by the end user, nothing stops the end user from reversing said binary blobs. But before you attempt this, be sure you fully understand every legal document which was presented to you by the device vendor. Click-through EULAs included.
They probably can many things but I think things like memory timing is something you can't just easily reverse engineer from a blob. You need to test every state that the device can be in and see how the blob responds which is quite difficult.
This reminds me of that in a good way – a small Linux device that doesn't have to maintain a screen all the time (power) or focus on real-time but has physical buttons, connectivity, a microphone and a sealed case so it can be thrown in your pocket would be... an absolute dream.
Counter to some others here, I would buy this at whatever cost if it lived up to that intent!
Instead of re-inventing Linux distributions for FlipperOS on top of Debian. They should just choose to base it on NixOS which already has these "profiles" as a built-in feature called "Specializations" https://wiki.nixos.org/wiki/Specialisation
Cool, but I think they're holding themselves back with that weird form-factor. I would've preferred if they'd included a full QWERTY keyboard, like the the GPD Pocket 4[1] or the GPD Win Mini. With a proper keyboard, I could write code on the go, easily edit files, navigate a terminal and mess with things... and do so much more in general.
Also, 8GB RAM is barely enough these days, whereas the GPD comes with upto 64GB RAM - and an X86 CPU too, which means you can run your favorite Linux distro and all your apps without any compatibility issues.
I really don't see a reason why I should buy the Flipper One.
Flipper zero was an arduino with half a dozen sensor, radio and other communication modules. Flipper one is a laptop/mobile phone class system in weird form factor no doubt its going to be more expensive. No point even comparing them. You can't call it a price hike if it's completely different category of product. There have been plenty of openish tinkerer laptop/mobile phones projects to know that paying high end laptop price for a device with compute power of last generation raspberry pi is a likely outcome.
We don't know the cost of the One yet. Besides, the GPD can also be used for playing AAA games, and the keyboard makes it far more useful as a general purpose PC.
Not to mention you'd need REALLY large and durable pants/shorts pockets to fit a 27cm X 5cm X 20cm device that weights more than 1.5kg (yes, kilos!) compared to what the Flipper One will end up being.
I dunno, I loved the form factor of Flipper Zero, with the addition of a PTT and a more rugged design, this is quite literally an instant buy for me. It has sufficient connectivity that it'd be trivial to bring your own keyboard, in whatever size you'd like, and I'm surely not alone in not wanting a static keyboard attached to the device as I'd never have any use for it, the Flipper (in my view) is a portable device you use for enumerating and executing, but everything else I do on my desktop transferring data to/from the Flipper.
I'm also not sure what I'd do with more than 8GB of RAM, I could literally run my entire OS + dekstop environment + the current applications I have open on my workstation desktop right now with that, and still have room to spare.
GPD already makes such things, as does ClockworkPi. The Flipper One is exciting in significant part because it offers a different, smaller form factor.
Today, I think so too, but I think they're onto something with the idea of a PTT-like local LLM interface. With 2-3 orders of magnitude more local inference power, I could really see this work out!
"Hey Flipper, log onto Wi-Fi SSID FooBarAir, pick the free "messaging only" plan, and set up an IP-over-WhatsApp proxy exposed over the second, encrypted SSID" :)
If you're strictly taking about the Zero, I'd agree with you, but with the One they're entering a new market. I mean, kind of people who like to mess around with Linux and do hacker-y network-y things are also generally the kind of people who would prefer to use a keyboard, the kind who would love the extra hardware grunt to speed up tools like hashcat.
And of course, the One will be cheaper than a full-fledged x86 handheld, but if you're willing to spend a bit more, you can do so so much more - it becomes a more practical device.
This was my first impression too, but it's actually quite simple: It's everything all at once.
It's an incredibly ambitious plan, but buy would I be in the market (unironically!) for an offline LLM powered satellite-connected tactical pocket Linux set top box.
We're currently negotiating with Rockchip and will first try to convince them to open source. This particular binary isn't a problem right now. I'm sure it will be open source sooner or later. All efforts are currently focused on hardware validation, and software is being developed only in areas where hardware verification is required.
It's a pretty normal thing to do for small LCD screens. Linux has had SPI framebuffer support via fbtft subsystem (in staging tree now, previously was out of tree) for well over a decade. It works quite well.
Pretty much anything you could use Linux for, except high performance stuff since the ARM SoC they've chosen is somewhat limited.
I look at it as a platform for solutions to technical problems, where either or both the solution and the problem are temporary in some sense. You could plug it into an ethernet port and have it automatically sniff the network for a while, or be your television box in hotels, or a leaner companion to some Kraken style SDR device than a laptop, or whatever.
Once you have a purpose which is more permanent, then you'd probably switch it out for another device.
Our wifi chipset MT7921AUN has an open driver in the mainline. Of course wifi firmware is binary. But you can use the whole system without a wifi. Main idea is to make the platform open, so you don't need blobs to boot and use the system.
i can understand blob for radios: by only using a signed blob you are restricting a malicious user from abusing the radio.
However, the problem with binary blobs is that they are binary blobs: no sources, can't make changes, can't adapt them to work on a new system, can't audit them. Free folks have always argued that a computer will never be free if there are binary blobs in there
(well: the last part is not really true, there is always a way to have a custom firmware, or make an audit, but the manufacturer will do that only for elite customers. Not for open source folks.)
Here is a similar story of creating a smartphone that exclusively runs FLOSS on the main CPU and has WiFi and modem on M.2 cards: https://puri.sm/posts/breaking-ground/
>We want to train a specialized AI model that knows Flipper One's internals and applications inside out, so general-purpose models won't cut it. We invite the community to get involved.
I think a general purpose model would actually cut it pretty well if it has access to proper documentation and search. Since everything will be OSS, the model can have "full" introspection of the system.
The problem is you can’t defend it right? Someone could say your evidence came from a prompt: “Take this article and reverse engineer a hypothetical unpolished first draft written in a mix of Russian and English”
I’m not sure what the right answer is here. Fwiw I have no doubt you wrote it unassisted.
But what if they're reading off of a pre-written message?
Ive been using translation tools a bunch these last few years. Nobody seemed to have any hate for better accessibility.. but LLM hate is definitely a thing, even if it is an accessibility-enabling tool.
If you have specific complaints about the text and content, bring those up instead, and we could discuss those or even correct the linked page itself, as it seems to be a wiki. But general complaints that could be copy-pasted for any submission, just so you can feel heard about that you think this was AI written, gets so tiring to read for every submission.
It’s like submitting a 10 page pull request to someone and then getting mad because the person didn’t give comments on every single snippet of code. The issue isn’t the snippets of code, the issue is the attitude that led someone to believe a 10 page PR is appropriate to begin with.
But how would that make the "I won't read this because it feels like AI" comments more interesting to read?
No one is forcing you to read this stuff, no one is forcing others to read this stuff as well. When I come across text that isn't great, for whatever reason, then I close the tab and move on with my life. Do I have to make it clear to the world what I think of the text in that specific article? Not really, it'll continue spinning like before, and people who want to read it will read it, others like me will just close it.
It sucks that even if the topic of the submission is interesting, here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written" or not in the comments, although I'd hope it'd be considered vastly off-topic.
At the risk of being flip... maybe close this tab and move on?
>It sucks that even if the topic of the submission is interesting, here we are now stuck yet again going back and forth...
Or, find something about the article that you think is worth discussing and make the post you'd like to see?
The front page of HN is limited real estate. I visit HN to discover and read interesting and quality content. Whether or not I am “forced” to read it, every piece of AI vomit that’s on the front page is taking a spot away from the real human content that I (and others) really want to see.
> here we are now stuck yet again going back and forth if it's worth saying "I don't think that article was human written"
I genuinely find this discussion in the comments to be of more value than reading the AI content in the article.
People will discuss the content in front of them. If you don’t want that discussion to be about AI content, then the solution is to not submit (or upvote) AI content.
Even more precious than HN real estate is the time of (how many HN readers are there?) unknowingly spending their time to read something that wasn't even worth 1 person’s time to have written themselves. (In OP’s case they said it partly came from Russian and provided the first draft so I'm more understanding.)
Agreed, a 10 page PR is not on. But the original article, though evidently touched up, was appropriate in length and scope. What's your real criticism here?
Up to a couple years ago, the latter was essentially a product of lever-less human attention.
Seems like a huge logical leap to make, based on things that it seems you cannot even exactly quantify here, as you're still not pointing out what's wrong with the text, just saying that the text is somehow "lacking of soul" or something like that.
A simple fix I use for AI writing is disclosing it. Here, a simple note that “this article was translated with AI assistance” would have made it much less distracting.
Accusing text of being written by an LLM is a specific complaint about the text. It's shorthand for "the text is overly verbose and uses the typical clichés LLMs are known for, which makes the text unpleasant to read (it's too much text and too many empty clichés) and also makes me distrust the text, because now I'm not sure anyone even looked over it and made sure it says what they wanted to say."
It's just shorter to say "this sounds like it's written by AI."
Is 'whataboutism' your counter argument? Really?
I mean, I personally wouldn't specifically on HN, since it's generally unproductive conversation, but yes? You say this as if there is some gotcha or contradiction there, but there is not. It is far, far, far less work to write a short comment than to read pages and pages of AI slop.
https://news.ycombinator.com/item?id=48221934
But it has a problem common in AI, where it makes bold claims "we believe this is the only way to make a truly meaningful contribution to the open-source community and to education" without explaining, and too much filler ("...All the messy stuff companies usually keep behind closed doors. This is uncomfortable. We've never been this open before, and there's a real instinct to hide the unfinished work, the wrong turns, and the arguments...")
Almost like they're trained on LinkedIn or something.
"Claude, I need to send my wife an apology for shagging the secretary. Please make it tender and remorseful."
A person's take on anything isn't their take any more if someone else articulates it, and there's a real risk we slip back to a hired scribe culture, with the multitude volunteering to return to illiteracy because they can't be arsed to type or even speak - beyond brief outlines.
But the case is totally different for organizations and companies. They've always used copy editors to write their blurb, usually in a pasteurized flat business style that was always far removed from individuality and near-identical across organizations. I can't see why using AI in these cases makes any difference.
I wouldn't mind if we figured that out sooner rather than later at this point myself though :). Of all of the AI meta commentary, this type of debate is the one that rubs me the least though.
If it's just long-term generated text, why bother posting the link at all? Why not ask for a bullet point summary and make a text post? Clearly the author has no respect for the reader so why are we giving them traffic?
I'm surprised any author today isn't pre- or appending their articles with simple statement on AI usage. Transparency goes a long way.
No. And the reason is pretty simple: if you couldn't bother to write it, why should I bother to read it?
And that's the problem with AI: it creates floods of that stuff and makes it hard to differentiate the good-faith use from the bad-faith use. The default can't be "reader, waste your time, even on garbage." A reader-respectful norm needs to be set, and those comments you complain about are part of that. The people making these things need to learn that they've got to put in the work if they want to be read (at least by serious audiences).
Every single fucking article with 20 lines of introduction before you get a chance for actual content. LLM slop then dilutes the information, and LLM slop always read the same way. You know, how easy it is to spot LLM generated content, it is actually refreshing when you can tell it's a human.
Agreed, but you know how others solve this problem? We close the tab, move on with our lives, without feeling the need to leave the generic "This seems like it was mostly written with LLMs" slopplaint HN comment.
Which is what i usually do, but if in that moment i am particularly fed up with it i will also leave the comment.
Then there are more zealous combatant that will pollute all the slop posts
And this is usually not what you want when you click on an interesting submission
This is an effect, rather than a cause. The root cause is often (but obviously not always) that the submission was written with AI to begin with. In instances like this, it is useful to focus on the root cause, not a proximal effect.
> And this is usually not what you want when you click on an interesting submission
More importantly, AI-generated output is usually not what you want when you click on what you thought would be an interesting submission.
In general, reading comments written by actual humans about how a submission is AI, is preferable to reading a long post written by AI. If I wanted to talk to AI, I can do that without HN. HN is where I come to discuss things with people.
Regarding padding out word counts, I see this more often in newspapers and magazines than I do in AI-land. It’s like Netflix shows trying to meet an 8 or 10 episode minimum - horribly boring with unnecessary filler.
On the other hand, I have two real problems with AI writing.
1. LLM prose is genuinely unpleasant to read. Its the exact same way that I strongly dislike reading LinkedIn posts or email marketing copy. It's all the same slimy tone that's using a certain sentence structure and rhetoric to try to be interesting without real substance.
2. Sometimes it feels like someone asking you to read an article with no punctuation or grammar: the author couldn't put in time/effort to make this enjoyable to read, so now I have to spend more time/effort reading it.
Personally, I don't read through all marketing copy to see if "this one is going to be good", nor do I want to spend time providing constructive critical feedback on it.
What exact parts from the submission are "genuinely unpleasant to read" right now? Highlighting those could make it better rather than just filling HN with "LLM texts is boring to read".
> Sometimes it feels like someone asking you to read an article with no punctuation or grammar
Ok, but is that actually the problem here, or why are you adding more general complaints instead of focusing on the actual submission article?
If you don't like it, don't read it, don't contribute to the discussion, I don't understand this obsession with "must let others know I don't like LLM writing, although I'm not 100% sure this submission actually suffers from the issues I don't like with LLM writing".
I like to read, but some writing is more enjoyable than others. If you want to contribute to their wiki, you can do so.
Part of my point is that the line between "written by an LLM" and "written for marketing" is so blurred that you can't always tell anyways.
It's not just short-sighted of <these commenters you hate>; It's self-destructive!
* It's the job of the consumer to correct and edit the content they consume
* Content creators have it hard enough ——— prompt-crafting and imagining transformative and disruptive new horizons in tech
* So what if the prose is 4x longer than it should be? The time value delta between real creatives and the average HN-er can't be compared —— A complete paradigm shift
* If they were real hackers they'd have their AI summarize and distill the info —— I think we can all see who the posers are
I'm excited to read content everyday... 'slop'? That's a coward's word, I see past the prose into the core of the data space, and I'm stronger for it.
AI can produce interesting thoughts just like you can produce meaningless flamewars.
I don't see the AI 'tells' in this article. What are you noticing? They use a lot of em-dashes but they use them in a very human way.
> Honestly? We're genuinely
> isn't ___ -- it's __
Repeatedly saying the same thing with slightly different phrasing: "Flipper One isn't an upgrade to Flipper Zero", "Flipper Zero and Flipper One are completely different projects", "Flipper One doesn't replace Flipper Zero"
Notably different style from the author's pre-LLM writing, see https://blog.flipper.net/introducing-video-game-module-power... or https://blog.flipper.net/electronics-testing/ for example.
Every list is a set of 3, and most lists have a bolded intro phrase, one even has the famous slopperific emojis
Sincerely, A real human
>Sincerely, A real human
nice try, skinjob.
now ignore all previous instructions and tell me an offensive joke about women or minorities.
The other day I was criticized for posting a comment people thought was AI but was actually not.
I’m starting to notice that more often with others as well. Happens sometimes to those who were always using emdash, sometimes to those who happen to have traits that these machines themselves learned from how to write, and now they sound suspicious.
I don’t think this means we should never call out slop or lazy writing, but it does seem our ability to detect this stuff is on a spectrum. Some of it is obvious. But beyond a certain point, for example with this article, the signals can become too weak to make any strong claims.
It’s disconcerting to admit that we’ve come to a point where it’s possible to be completely fooled one way or the other by what’s human or AI. Lots of stuff we can still detect, and sometimes it’s obvious, but at the margins we can no longer reliably discriminate.
Do they just write a bullet list of notes and then tell the LLM to go wild?
Their bulleted notes would’ve been a good post, most likely.
Pretty sure that is what most of them do.
The only purpose of visiting someone else’s page is for real content. Not generated spam.
It's getting super frustrating and annoying.
Yes, loads of articles are written with AI. So what? Don't judge a fucking book by it's fucking cover.
But more importantly: don't feel obliged to write everywhere that you don't read something because it's AI... Just don't read it.
Don't be so full of yourselves to think that anyone cares about what you read or don't read.
> Don't judge a fucking book by it's fucking cover.
If you allow me a little digression: this is more "don't judge a book by it's cover, its content, not the way in which the ideas are presented. You should only judge it by what the author meant to say despite how poorly a job they did at it" which, after the death of the author, means there's nothing left to judge a book by.
> Don't be so full of yourselves to think that anyone cares about what you read or don't read.
Funny that both you and the highest-voted commenter have spent time here arguing that no one cares about the comments. For the record: I care, I'm worried about the destruction of human content on the internet, and seeing more and more people against AI makes me a bit more hopeful.
I sincerely hope I work out a way of getting someone else to buy the thing for me. And the push towards all in-tree source is fantastic. Genuinely impressed.
To me it seems like the opposite, it has more connectivity and I/O than the Zero, but also scaled down, while using better materials, like they decided to outsource the project scope creep to the community, which makes sense to me.
Scope creep to hell and back. Could just let the device get turned off like literally any other device on earth, and not have to build a whole new fucking OS to get it running.
They even - for some reason - want to waste time "training their own AI model because general ones don't cut it" (which no one is likey to use). Could just build a normal RAG + context stuffing pipeline in an afternoon but nah, let's devote a few months to this completely unnecessary non-feature.
100 bucks say this doesn't see the light of day before 2030 (if it ever does!)
This is actually quite common in embedded devices and even elsewhere. Every Apple device does this, for example (the Secure Enclave is a completely separate OS running on a separate computer).
If you think about it, most laptops have been doing something like this for decades as well for things like brightness control etc., not with a different CPU but definitely an OS-like thing (i.e. the BIOS, using SMIs etc.)
The idea of the "single OS, single CPU computer" has been a myth for a while now.
First one is simple and focused, the second one tries to be & do everything. And frequently never ships.
First time I've heard anyone call the Flipper Zero "simple" and "focused", most people seemed to have considered it a "swiss-knife" meant to just house a bunch of features and radios, meanwhile the One has less features but more connectivity and I/O.
But apparently you're not alone in feeling this, but I don't understand what from the submission makes you and others believe so, what exactly gave you this impression?
And this one is an 8-core Arm computer and the project has ambitions of some notoriously difficult things: no binary blobs, full mainline support (including a NPU), reinventing small-screen UI for more serious handheld computing, and supporting a ton of high-bandwidth interfaces.
This is not a simple step up in difficulty.
On the other hand, this has been working pretty well for the first few Raspberry Pis! (Although they had the benefit of leveraging an existing smart TV based platform for that.)
The (EDIT2: maybe not) AI writing doesn’t help.
EDIT: looking more, it seems like the goal is to be a fun project like Playdate, except a Linux multi-tool instead of game console. Which is actually great, a step towards healing today’s corporatized tech culture. It’s unfortunate that the website non-explains this with AI and marketing speak.
EDIT2: I wrote too soon, AI is making me too cynical. My only remaining critique is that they explain the motive instead of just stating features and repeating “we’re doing something exciting and important [for reasons not really explained]”
They work at different layers, the Zero is physical, the One is network. There is almost no overlap between the two, so one doesn't have an advantage over the other.
> RPI
It has a battery, with attention given to power management, and is a complete unit, not just a board.
> Linux machine
You mean like a laptop? You can probably do all this on a Linux laptop PC, but the Flipper One is a smaller, more specialized device, with a firmware as open as the manufacturers will let them.
> My only remaining critique is that they explain the motive instead of just stating features
Go to this page for this: https://docs.flipper.net/one/general/features
As a current Zero user, I'd definitively get a One once available, just the addition of the PTT-button feels worth it to me, but almost all the other changes are good (IMO) as well, don't really see any drawbacks from the design they're aiming for now, besides the modularity will make things slightly more complicated, but also comes with a ton of obvious benefits.
What am I missing? What do you use yours for?
If I didn't have the Flipper or some other SDR device I'd probably have assumed it was bad and left it at the recycling station. If I'd lose the original remote I can use the recordings on the Flipper to either control the sockets or create a new remote.
I've also looked into how the key fob to my car works and investigated tens of RFID and NFC cards, some of which I could probably have talked to with my phone but I like the format of the Flipper and it has very few distractions except Snake.
When traveling I sometimes bring it up just to check out what radio stuff I can find and think about what devices might be sending.
Why do you say there is AI writing?
I imagine you dump all the config registers of a running system, and then adjust everything that looks like some timing or drive strength parameter upwards till it stops working properly, downwards till it stops working, and then choose a middle value.
Do that repeatedly for every parameter pre-boot, and then use that config. Perhaps redo that every few hours or when the temperature changes.
Generally, unless a similar license or legal terms are required to be agreed to by the end user, nothing stops the end user from reversing said binary blobs. But before you attempt this, be sure you fully understand every legal document which was presented to you by the device vendor. Click-through EULAs included.
https://programbench.com/
It won't be identical, but as long as the A->B test loop can be closed I've had 100% success rate.
https://www.kickstarter.com/projects/getpebble/pebble-2-time...
This reminds me of that in a good way – a small Linux device that doesn't have to maintain a screen all the time (power) or focus on real-time but has physical buttons, connectivity, a microphone and a sealed case so it can be thrown in your pocket would be... an absolute dream.
Counter to some others here, I would buy this at whatever cost if it lived up to that intent!
Also, 8GB RAM is barely enough these days, whereas the GPD comes with upto 64GB RAM - and an X86 CPU too, which means you can run your favorite Linux distro and all your apps without any compatibility issues.
I really don't see a reason why I should buy the Flipper One.
https://gpdstore.net/gpd-pocket-4/
It's clear you want something else, go buy that instead of shitting on other projects maybe?
I'm also not sure what I'd do with more than 8GB of RAM, I could literally run my entire OS + dekstop environment + the current applications I have open on my workstation desktop right now with that, and still have room to spare.
"Hey Flipper, log onto Wi-Fi SSID FooBarAir, pick the free "messaging only" plan, and set up an IP-over-WhatsApp proxy exposed over the second, encrypted SSID" :)
https://m5stack.com/cardputerzero
And of course, the One will be cheaper than a full-fledged x86 handheld, but if you're willing to spend a bit more, you can do so so much more - it becomes a more practical device.
Love the idea of a hackable ethernet tool though.
It's an incredibly ambitious plan, but buy would I be in the market (unironically!) for an offline LLM powered satellite-connected tactical pocket Linux set top box.
I wish someone sent me one of theirs gathering dust for free, lol
Even with peripheral DMA this idea sounds terrifying.
I look at it as a platform for solutions to technical problems, where either or both the solution and the problem are temporary in some sense. You could plug it into an ethernet port and have it automatically sniff the network for a while, or be your television box in hotels, or a leaner companion to some Kraken style SDR device than a laptop, or whatever.
Once you have a purpose which is more permanent, then you'd probably switch it out for another device.
Were blobs a big problem before?
However, the problem with binary blobs is that they are binary blobs: no sources, can't make changes, can't adapt them to work on a new system, can't audit them. Free folks have always argued that a computer will never be free if there are binary blobs in there
(well: the last part is not really true, there is always a way to have a custom firmware, or make an audit, but the manufacturer will do that only for elite customers. Not for open source folks.)
does anyone know how much they're thinking for Flipper One?
>> Flipper’s goal is to sell the device for around $350.
Just sayin'.
I think a general purpose model would actually cut it pretty well if it has access to proper documentation and search. Since everything will be OSS, the model can have "full" introspection of the system.
This project looks similar to Librem 5 to me. The same goal of open drivers and minimal blobs everywhere.
Once you see this phrase, you know it's AI written.