Root shell on Samsung TV
twitter.comAs far as I can tell, there is no write up. This is a reddit post with screenshots of a Twitter/X post by @log4j. Even the Twitter post doesn’t seem informative, but Reddit isn’t a source for the information by any means.
The original link was on HN two days ago with no traction.
So a bit of background I spent the last 17 years going Smart TV app development.
Not a lot here is any sort of surprise to anyone in the industry.
Always on audio is common on a whole swath of devices - any that want to do Alexa style voice prompts need it. ACR is table stakes for modern TVs reselling viewer data has driven down TV prices to the point where there is no going back.
The policy file here is how you build Tizen apps. It’s a list of specific features your app needs access to and certain levels of access (controlling the TV For example) require different levels of contractual relationships with Samsung.
Everything Samsung collects should be covered in their, very dense, TOS.
I spent a couple years working at the platform (AndroidTV + Google Assistant) level on smart TVs.
> any that want to do Alexa style voice prompts need it
Not technically true. More modern smart TVs (~post-2021) put speech recognition models for the hotword only onto the DSP, and then have the DSP wake up the CPU for further auditory processing when it detects a hotword or remote button press. There are major power-efficiency benefits to this, because the DSP uses a fraction of a watt and can easily keep running while the TV is sleeping, while the CPU, Internet, and rest of the circuitry is a major power hog. Your smart TV or streaming dongle would basically never sleep if it were forever running the audio capture.
For that matter, it's pretty rare for the device to actually run the speech recognition models on-device. Smart TV manufacturers are very sensitive to BOM costs, which means that their hardware is only powerful enough to run machine-learning models at the extreme high end of their product range. Most of the time the full audio gets streamed back to the Internet and the speech recognition & voice assistant gets run in the cloud, with commands streamed back to the TV.
I won't say that it never has always-on audio listening; a significant number of devices I tested did. But I'll say this comes because they're poorly implemented. The DSP hotwording approach requires coordination between hardware and software developers, oftentimes across companies, and it's more complex to implement and a lot of executives don't care about the marginal consumer benefit it gives. Many smart TVs just guzzle power when they're off instead.
Yeah I don’t disagree, but should clarify that I’m just referring to the security policy that provides access to the ambient mic. How the app/platform handles that audio is a different path that I’m not entirely familiar with.
> ACR is table stakes for modern TVs reselling viewer data has driven down TV prices to the point where there is no going back
Really? There’s no “spyware free” marketing angle? Even Amazon has always sold a Kindle variant with no ads for $20 more.
Not really, that said it’s trivial to disable, you can opt out either via their settings or just by not connecting the TV to the internet.
I will say that how “completely” disabling ACR via settings works is not terribly clear and there are a few lawsuits questioning wether “opting out” actually works as presented.
Except it's not covered in their TOS
Oh no, how will Samsung survive this, they will have to send $5 to few thousands of people for the damages if they decide to start a class action lawsuit and if a judge/jury decides it was really causing damages... as a reminder Samsung Electronics made almost 250 B-B-B-B-illion dollars in revenues just last year.
Maybe not the same exploit but I found another guy who was also able to debloat his TV: https://github.com/jss8649/quiet-tizen
A little bit of traction https://news.ycombinator.com/item?id=50193579
These posts are engagement farming, they should be banned in HN, it’s like sending a marketing buzzword email to a technical engineer, useless.
What are you talking about? I have a samsung frame that I'm in the middle of rooting because of this excellent information.
I always advise people to never, ever use Reddit as a reference for anything.
Why and what reference do you rec instead? Even this site gets fake news that stays up for hours.
I think that's the point. If you have to go looking for a reference and don't easily find one, it's probably fake.
I would avoid and distrust most social forums, such as reddit, that have Pookas which are fond of rumpots, crackspots and how are you bix6? -- https://www.youtube.com/watch?v=NeEsx2QLJ40
Sure, ‘knee-jerk’ reddit is unreliable, but many upvotes over time with supporting comments can be a strong signal of truth
It’s unfortunately only a symbol of bot activity or popularity with the crowd in the subreddit.
What do you mean? I can post whatever I want on Reddit and I can make it have as many upvotes as I need. They basically make bot farms undetectable, by hiding histories and automatic username selection. Literally nothing should be considered truth with Reddit as its only source
Why? Their system is literally mined for its Q&A merit by search engines. If it was all very bad information, NYSE: RDDT would not exist.
Stock prices are based on earnings, not on factual accuracy of content.
To what extend do these "smart" products state how, when and what data they are collecting and selling?
Isn't it high time for legislation that at least makes it clear on the box what kind of sensors and integrations it has and how bad the product is privacy wise?
Also, shouldn't corporate spying disguised as automated content recognition and classification be limited or outright be banned?
And all this connected garbage should by law be able to function with the cord cut and without registration, and if an update negatively changes the advertised product characteristics the customer must be allowed to return it for full paid price even if it is 10 years old. Force the manufactures to make statements on the "box" about the features, the support duration and level, and the privacy characteristics.
Why don't politicians care about this?
Why don't politicians care about this?
Because there have not yet been any high profile cases involving the collected data used to embarrass or convict a politician. It just takes one embarrassed congress critter to get a giant red sticker on all these devices, their boxes and ads: "This Device Transcribes and Reports Your Conversations" and that is before including those that upload camera screenshots of what you are watching and video of the viewers doing something untoward.
This is literally the history of the Video Privacy Protection Act (1988, back when video rental was a thing). It was passed immediately after, in 1987, a journalist obtained and published the video rental history of a supreme court judge nominee.
Your outrage is well earned, but fairly rare. The reason nothing is done is because the public doesn't care.
> Why don't politicians care about this?
Because it is not the subject that will win them the next election.
And even if they care, some things are easier said than done. With politicians who care, you get things like GDPR, which goes in the right direction, but is a huge legal mess. And it is a mess because it can't be made simple.
Obviously, you have to take consumer interests into account, but you also have to take companies interest into account too, both big and small, otherwise they will just stop doing business, which is bad for the economy and will mean that consumers won't find what they want at a reasonable price. You have to think about the loopholes big companies will invariably find, and covering them will add complexity, but still make it so that small companies can comply without a full time team of lawyers assisting them.
For example, you say the consumer could return it for a full refund on a bad update, but to whom? Should the burden be put on the small retailer who sold it to you (and may be out of business)? There is already plenty of stuff the manufacturer must tell you, usually one a piece of paper no one reads, and the more you ask, the less people will read because there will be too much information. A "privacy score" may be a good idea, if well designed, it will share a place with the energy score, environment score, repairability score, safety signs, origin, etc... And how do you differentiate actual automated content recognition that consumers could want from spyware, formally, so that the first one is allowed but not the latter.
All that to say that it is not impossible, but even with politicians caring, it is complicated. European politicians care (because it can win them votes), the results are so complicated they are getting ridiculed by the rest of the world.
They don't care because of several reasons. 1. They get paid not to 2. They are told not to.
Alot of this spying ends up producing valuable data for different agencies to purchase which otherwise they cannot legally obtain without a warrant. Sure some of it has some commercial value, but a lot of it is probably more valuable for domestic spy operations.
They still can’t technically obtain it without a warrant. They do it anyway.
> Also, shouldn't corporate spying disguised as automated content recognition and classification be limited or outright be banned?
A lot of info collected by 3 letter agencies is based on it.
Politicians didn’t even do anything about Epstein. I’m pretty sure they’re all compromised.
Who bears responsibility when TV makers and data brokers undoubtedly had clear and verifiable evidence passing through their systems?
Follow the money
Remember guys, if they are reading it for wrongthink, they are reading it for industrial espionage.
Anthropic and openAI just announced they’re spinning out competitors to their biggest customers… right after ingesting all of their IP and contracts. Do the math.
(Or don’t, because they’ve been caught stealing math research)
> Anthropic and openAI just announced they’re spinning out competitors to their biggest customers… right after ingesting all of their IP and contracts. Do the math.
Do you have any links to further reading on this. I haven't read anything like this and also couldn't find anything in their newsrooms. The only thing I read is that they want to have more people work on implementations.
I think it's just a logical conclusion if everyone keeps feeding OpenAI and Anthropic data. I don't think it's imminent, but if AI keeps improving and regulation is very lax I could see this world developing in 5-10 years.
Say OpenAI builds GPT-11 and they realize it can do 95% of their clients work. Why not hire away key people from the client or the client's competitors and start eating into their market share? OpenAI would realize if GPT-11 has this ability then Anthropic is probably able to do the same thing. The company that has the most money and credit available is going to be able to buy more compute. OpenAI and Anthropic cannot slow down unless mandated to by the government.
>Say OpenAI builds GPT-11 and they realize it can do 95% of their clients work. Why not hire away key people from the client or the client's competitors and start eating into their market share?
Can you give me an example where this would make sense, because I just don't see e.g. OpenAI starting a grocery store chain or Anthropic building a wind turbine manufacturing company.
What I can see is that some company's products become useless, e.g. translation tools, as it is something that you can just do within ChatGPT or Claude.
You picked physical work, but how about consulting or sales?
Interesting. Could you share a link to this announce?
> because they’ve been caught stealing math research
No. They haven't been caught stealing. Tristan Buckmaster publicly expressed surprise that the OpenAI model has used the niche approach he pursued (with AI assistance). 10,000 agents exploring various approaches make it much less surprising.
Another anecdotal data from an arxiv preprint [0]: "After the paper was written, the author learned that the same result was also proved independently and by a different method by Ningyi Li [7]. The result is also contained in a paper [8] released by OpenAI on October 6, 2026, in which the proof strategy and specific choices of notation are identical to a preliminary version of the present paper that was uploaded to ChatGPT on September 8, 2026."
So not only did OpenAI publish a proof similar to one discussed on ChatGPT, but also there were other possible strategies to prove the result, that OpenAI didn't follow.
Also, up to my knowledge, on the free plan, OpenAI doesn't claim that they won't use the chat data for training. So this seems plausible.
The "coincidence" defense never works for random individuals so I won't buy it when it comes companies worth billions which will lie, cheat, and deceive to make even more billions. They've been caught lying and stealing again and again. After so many years of this I think any trust put in them is either blissful ignorance or willful misdirection.
Wait are you saying: 1) the simultaneous invention defense is never valid for individuals, and 2) you don’t base your beliefs on what’s right, but on some kind of tit-for-tat retaliation?
That seems like a fraught way to live.
It's not a coincidence... they literally just span up agents on every possible variant of the problem.
There are 10k agents and Tristan's approach wasn't even groundbreaking novelty, it was literally based on existing published literature from Diego Córdoba and Luis Martínez-Zoroa that the agents have seen.
It's not just probable that agents would persue his particular niche of trying to construct a blowup. It's guaranteed if you just consider the basic setup.
The coincidence is that they took the correct path out of “every possible one” exactly after one of their users took it.
Monkeys on typewriters excuse only works for huge companies who can afford to buy opinions.
> The coincidence is that they took the correct path out of “every possible one” exactly after one of their users took it.
From what I read online that researcher also used a lot of AI to find this approach, so it might just be the one that is the easiest to find for AI agents and the timing would then only be caused by them working on it around the same time.
They also took the correct path for hundreds of other problems, where they didn't have a user that already did.
> The coincidence is that they took the correct path out of “every possible one"
My point was explicitly that this didn't happen. I said that they took every possible path and published the one that worked. It's exactly throw shit at the wall and see what sticks.
> exactly after one of their users took it.
All parties agree that what happened is that someone had told someone working at OpenAI that Anthropic employees are working on Navier stokes. This is where the timing comes from.
---
What you are implying is this: OpenAI didn't do the throw shit at the wall approach as outlined in their release blog they actually secretely monitor everyones chat logs and found the idea in Tristan's Chat history and stole and published it.
Or some variant of this. You are only implying what you think is the truth, please state it outright. Don't beat around the bush.
Dog bark and baby crying detection are advertised features for SmartThings integrations. Door bell and door knocking are also included, this is surfaced directly if you hook the TV or sound bar into SmartThings.
I had to check this and it's true. Not saying Samsung or 3rd parties don't take advantage of any data they can collect, but the story the tweets tell is super biased against them. Here, the detection seems limited to specific pattens required for a particular feature.
In fairness, they have not earned the benefit of the doubt based on past behavior.
These folks recently did tests on major TV brands, not accepting the on-device consent forms, and then recording how much data they continued to send after consent was denied (which itself is not always possible). https://www.youtube.com/watch?v=t_GK06_3UY8
Samsung was the single brand reviewed which actually stopped sending data after the user told it to do so.
I'm pretty sure it was actually LG?
In February Samsung agreed to stop collecting the data when opted out in the US after a few lawsuits were lining up. LG was the target of a different lawsuit in Texas which they settled and then wrote up a forced arbitration clause and doubled-down on data collection, even when opted out.
Don't buy LG TVs.
Oh don't worry, shitty WebOS is enough to put me off that (I unfortunately own one). However in this case we were talking specifically about that recent study of the effects of opt-outs, and for LG TVs the opt-out did result in the no longer detecting any tracking.
Yep, lg. After all, they believe that own your living room and their actions reflect this.
Still, calling it "PRIVACY HORROR" as per original post is typical twitter-style sensation seeking. It's an opt-in dog barking detection that causes local chip to listen for... dog barking
But does this mean they don't turn this on by default resell that data? X house has dogs or whatever?
We clearly need a whole new privacy framework where everything is optional and opt-in, plus open software to monitor this compliance.
I will add: Use the TV as a HUB for all IOT, light sensor so you know when the light is on or someone passes in front of the TV and there are some others but I can't remember now. They are all advertised in the Smart Things app.
I hate 2 things about my samsung tv's (I have 4 only samsung): the simplistic remote and ads on home screen. Even with custom dns sometimes they get through.
This looks very bad on OpenAI.
What is next? Power and water company suspending your account if your home and car are not streaming enough telemetry.
I am glad that open models are closing on frontier.
Claude is absurdly sensitive to anything that touches cyber, and threatens bans if it gets a whiff of a pentest going on. it wouldnt surprise me if openai had a similar approach.
Yesterday I joked that claude gave me a cyber stop for asking how to make the anime girl video for an exploit, but it was a thing that happened. Meanwhile im trying to convince people around me that having grok nudify people isnt "fine" just because grok says it is. Its wild where the lines a drawn.
Claude is so absurdly bad about false triggers for security. It kept triggering when I asked it questions about how to plant some tree saplings I got because somehow that was a security risk.
Anthropic is a joke of a company.
Would you mind copy+pasting the text that you fed it that led to that refusal? That's so bizarre. It would be really interesting to try and guess at what exactly in the message set it off. Over-indexed on biology refusals, perhaps?
I'm in the new CVP, and even Mythos gave me a cyber stop just for creating an MCP server that can remotely connect to my Steam Frame to run commands. It must've thought I was trying to create a centralized malware or something.
Not even sure why Mythos even still exists when it's identical to Fable in every way, including refusing to do stuff that looks like it could ever be repurposed into unapproved use cases.
Wasn't the whole point of Fable that it's Mythos but with all the cybersecurity stuff lobotomized out of it?
The whole point of Fable was that it's Mythos but with extra guardrails, yeah. But now that Project Glasswing has been folded into the CVP, Mythos seems to have the exact same guardrails now
We're all naked under our clothes, I don't see the problem. One of the funniest shock-bits I've ever seen was when Jon Stewart's book on American government and history included full-frontal (Photoshopped) nudes of SCOTUS, but I guess he couldn't get away with that during the current era of Puritanical panic.
It's just an adult human body, Michael, what could it cost? One('s) dignity?
Tastelessness aside, this seems perfectly rational to me. Unlike reverse engineering for the purpose of developing security exploits, "nudification" doesn't meaningfully expand the capabilities of malicious actors with access to free online porn and Photoshop.
Actually violating one person is fine, but doing anything associated with a specific industry that has applications for harm and help is not?
That's like saying it's fine to beat someone up and take their lunch, but it's not fine to learn how locks work.
It's fine to imagine what your lunch looks like, even if that makes you feel that your dietary privacy has been violated.
You don't need Grok to imagine anything.
Well, maybe _you_ do.
People really out here directly comparing women to food, unbelievable.
Siri, what’s the name of this kind of bad faith argument?
Stating literally exactly what is being done is a bad faith argument now?
Yes. You’re 2 for 2 on bad faith now.
Today is Sausage Saturday, not Taco Tuesday. Some people like their links and eggs, and... okay, I don't know how to double entendre this next part, but suffice it to say that it's bad faith to assume that everything related to the depiction of the natural human body is about degrading women.
This thread is literally about Grok publicly nudifying women, and people are out here saying it's comparable to seeing somebody's lunch...I know HN is a boys' club and all but this is pretty bad even by normal HN standards.
No one else mentioned women. You're the one who assumed nudification was only of women and only of unwilling subjects. I've seen people use it on pictures of themselves for... enhancement and personal gratification. I've seen men and women use it to make spicy images of adult men. And if we're talking specifically about autonomy and control in depictions of a woman's body, I've seen some women complain about the overzealous guardrails of other image generators, where something as innocent as a touch-up of a photo of their own fully-clothed body was flagged as unsafe.
This issue isn't black and white, and you're casting unearned aspersions. We should have this conversation, in actual good faith.
That's one of the weirdest takes on here in a while.
I can develop security exploits myself just fine without the use of AI. Many people don't have the capability to use Photoshop and free online porn to deepfake someone else they might or might not personally know into a realistic video...
> I can develop security exploits myself just fine without the use of AI.
Are you seriously claiming that the untrained masses can start developing exploits on their devices?
Honestly as someone with faily little experience in pentesting or photoshop but who has dabbled in both back in highschool. I think they would both take a similar amount of practice and skill to develop those abilities. They are complete opposites so i suspect the kind of person who is naturally good at photoshop to be bad at pentesting and vis versa. But if your going to dedicate a few hours of every day to learning a skill i dont think it matters which one you choose, a bit of time, practice, and natural interest and you will be able to do the kind of work that people ask these llms to do.
The major difference being that it usually takes longer to study code because projects can be massive but at the same time the llms are looking for everything and anything while a real attacker is usually looking for specific vulnerabilities in common libraries and such so i wouldnt exactly call the amateur hacker on par with an llm but i would say they could do a similar amount of damage without llm assistance. And photoshop is basically the same. Spend a few hours every day dedicated to one thing and you will learn how to do it.
Seriously this. If my high school self could figure it out an adult with will can too. It really doesn't take that long.
Bad on OpenAI? The account got suspended.
This reflects horribly bad on Samsung!
The headline is misleading, but the "article" is only five screenshots of Twitter posts. Someone used AI to decrypt the Samsung TV's firmware and found always on listening, classification, lots of privacy nightmares. Then their (the hacker's, not Samsung's) OpenAI got suspended for "Cyber Abuse".
Right, but still: OpenAI provided a service and it's up to them to determine acceptable use. Samsung sells a $3000 (or actually $2999,95) TV and you get big brother watching you.
Surely acceptable use is not exposing crooked corporations that are using spyware to listen to you and your children.
They don't know that. It's the hypocracy of not using their service to infringe on someone elses copyright. The only way Compaq could legally have a PC BIOS is by having one team reverse engineer the original and describe what it does and another team reimplement it.
Yes, it's completely hypocritical because of how LLMs are built. That doesn't change the fact that if you're offering a service online, you'd try to prevent abuse in order not to be held liable.
You can do the same with LLM. Have one session document something, and other session write up an implementation, with nothing but docs shared in between.
I agree with you, but if OpenAI wasn't suspending users for exposing cyber-spying that these consumer companies like Samsung are doing, more and more people would expose such issues.
With Kimi K3.1 Pro, today WAS the day they actually caught up.
I’ve been saying this for the past few months - once the inflection point happens (such as today), it’s only a matter of 2 years when American closed models won’t be able to catch up with Chinese open models
Did you mean Mimo v2.6 Pro? Because Kimi K3.1 Pro is proprietary, and not open.
Also leading the frontier is much much harder than catching up to frontier.
Iirc They specifically trained them to not be useful in cyber stuff. But the point still holds
The Chinese models?
I have been wondering if that would be a benefit of using say GLM5.3 on HF if only to get fewer cyber stop flags.
There is more written here: https://www.kimi.ai/blog/kimi-k3-1-pro
They claim that the default datasets contain so much cyber stuff, so that their models became quite good at it.
They filtered the data before training to not contain this. So it's not that they train the model to not do it. They rather not have it in the Corpus at all. In theory
Even if there were stop flags, you simply get one of the ABLITERATED models that have all the guards NOP’d out
… IF chinese models stay open.
Chinese models don't need to stay open. The open weight ones just need to have nearly the same capabilities as the models sold by US companies. When that's the case, no one will have a reason to continue paying anything to any US company, thus destroying their business model and business case.
Idk about the destroying their business model part. Running your models on your own is akin to running a media server, which most people seem to be fine paying $20/m for someone else to do.
Well, maybe if your own media server only showed videos in 1fps compared to hosted ones. You’re not comparing apples to apples here. You might argue that local inference speeds are acceptable, but that doesn’t make them comparably equivalent.
AI companies can’t sustain this bleeding of runway. Eventually they’re all going to have to significantly bump up pricing. I guarantee personal accounts will be between double and triple your current internet and phone plan combined
People have been saying this for three years. When will it happen?
In fact they can be clearly worse, so long as they are same ballpark for many use cases.
I.e. I would pay 15-20% more for comparable work from an open model I control,rather than use anthro et al.
And it seems we've gotten to the point that frontier models benefits are marginal/nominal. Whereas, not long ago Claude was materially ahead and could do things the other models simply couldn't.
... IF the world doesn't end tomorrow.
Yes, of course, IF a lot of things. But if we go by what the official positions of the government is, and where the ecosystem is today, then the models stay open, as they are. And today, the US models stay closed, so lets not assume "... IF they stay closed" as nothing is pointing to this changing, just like China seems to point towards staying with open models.
This is again will be very funny race to watch. Right now you can fairly do lots of reverse enginnering or "hacking" with off the shelf Chinese models, and they are very capable (GLM3 for example) which can be run locally if you stomach the hardware. Tools like rea are also becoming more popular. If openai continues to be pedantic, they will lose the technical base given how close is the gap getting.
The company I work for has a plan B to switch to alternative, self-hosted models if providers pull off strange moves, such as drastic pricing changes. We have already switched from one provider to another once, and we do not want to get burned again. We do not reverse engineer, just regular development.
Until Kevin quits and you need to modify his work
Right, because admin is some obscure work in 2026 and no one knows how to do it.
How do you correlate owning an LLM implementation and "admin"?
Owning implementation = self-hosting one?
Have you tried Qwen3.8-Flash-Next? I found it significantly better in anything I tested than GLM5.3-Flash, but the thing is you either use Alibaba cloud or you host it yourself (no lesser quant than unsloth 4.5 or mx4/nvfp4 and kv no more quantized than int8).
It is great if you have enough RAM. I have enough to run it, but not enough for prefix cache. 27b is so much faster with long term tasks.
The Qwen3.8-Flash-Next model is ok for reverse engineering. It can definitely solve simple tasks, but it gets easily confused, and burns tokens like crazy.
I've been sick with stomach flu the past few days, so I put together a small benchmark for all the top AI models where I ask them to reverse engineer a very simple small crackme (https://www.crackmes.one/crackme/6ab4ce5195b976f8f1300b54) using Ghidrasql (https://github.com/0xeb/ghidrasql)
(As an aside: I think representing tools for AI as virtual SQL tables is much better than traditional MCP servers, and that this approach deserves a lot more attention. Props to Elias Bachaalany for thinking out of the box and putting it together).
The results of the benchmark are below. It's sorted by cost, but I added notes about each AI model for myself. The notes try to talk about intelligence, speed, and cost. Intelligence is a subjective measure which I assign to each model by reading its thinking tokens and watching its tool calls. You can definitely tell which models are more intelligent, even on this small benchmarks, by their tool use, false starts, and how easily they fall for the simple misdirection in the crackme.
(gist is more readable on mobile) https://gist.github.com/markoglasgow/8812c19b6fbffed30b2661b...
```
```+----------------------------------+--------------------------------------------------------------------------------+-----------------+-----------------+-----------------+ | Model / route | Experience / outcome | Total tokens | Output tokens | Cost | +----------------------------------+--------------------------------------------------------------------------------+-----------------+-----------------+-----------------+ | Xiaomi MiMo v2.6 Flash | Incredibly fast and intelligent; solved. | 225,291 | 3,281 | $0.005 (0.5c) | | GLM 5.3 Flash / OpenRouter | Least intelligent group but cheapest; solved. | 340,006 | 11,205 | $0.01 | | ChatGPT Luna Latest | Very fast, concise, token-efficient, intelligent, and cheap; solved. | 401,837 | 5,119 | $0.01 | | Qwen 3.8 Flash | Initially confused and failed, then eventually solved. | ~418,000 | 16,676 | $0.02 | | Qwen 3.8 27B | Fast; one wrong answer, then quickly self-corrected. | ~774,784 | 12,658 | $0.02 | | DeepSeek Flash / direct API | Fast, cheap, and accurate; solved. | 369,334 | 9,089 | $0.03 | | DeepSeek Flash / OpenRouter | Fast and intelligent; solved in 76 seconds. | 371,689 | 13,888 | $0.03 | | Claude Haiku 5.5 | Fast and surprisingly intelligent, but overly thorough and token-heavy. | 662,346 | 26,244 | $0.03 | | Thinking Machines Inkling | Low intelligence; good speed and relatively token-efficient; may loop. | N/A | N/A | $0.03 | | Xiaomi MiMo v2.6 Pro | Overthought the task and wandered into tangents. | ~420,644 | 13,638 | $0.05 | | GLM 5.3 / OpenRouter | Cheap and intelligent but behind DeepSeek; solved in about 2m 20s. | ~702,473* | 11,879 | $0.08 | | GPT Sol Latest (Sol 6.1) | Slow, but highly intelligent and extremely token-efficient; solved in ~89s. | 294,957 | 1,406 | $0.15 | | MiniMax M3 | Least intelligent comparison; many wrong turns and repeated small mistakes. | 4,546,552* | 116,804 | $0.20 | | GPT Terra Latest | Very fast, very smart, and token-efficient; solved. | 229,888 | 2,342 | $0.20 | | Mistral Large 4.0 | Medium speed; small mistakes but reached the solution. | 303,687 | 4,041 | $0.21 | | Kimi K2.7 Code | Less intelligent, slow, expensive; not worth using. | 100,740 | 8,938 | $0.22 | | Claude Opus 5.5 | Highest intelligence and token efficiency; pretty fast; cheaper than Sonnet. | 312,837 | 1,862 | $0.24 | | Qwen 3.8 Max 0902 | Fast and intelligent, better than Kimi K3, but token-hungry. | 1,305,241 | 26,390 | $0.25 | | Kimi K3 / OpenRouter | Faster and cheaper than direct K3, but still very expensive. | 750,064 | 5,922 | ~$0.32 | | Meta Muse Spark Open | Intelligent and medium speed, but among the more expensive runs. | 872,850 | 10,284 | $0.35 | | Gemini 4 Argon | Extremely fast and intelligent; strong value for capability. | 384,046 | 2,870 | $0.35 | | Grok 4.7 | Very fast and intelligent, but expensive. | 655,648 | 4,217 | $0.46 | | Meta Muse Glimmer 30B | Fast but low intelligence and token-inefficient; cancelled unsolved. | N/A | N/A | ~$0.50 wasted | | Claude Sonnet 5.5 | Smart and fast but very token-inefficient and expensive. | 1,243,285* | 20,381 | $0.57 | | Kimi K3 / direct API | Intelligent but not notable; extremely slow and far too expensive. | 205,710 | 3,215 | $0.72 | | ChatGPT Astra Latest | Medium speed, token-efficient, highly intelligent; cost similar to Fable. | 278,603 | 1,536 | $0.83 | | Claude Fable 5.1 | OK speed, very expensive, medium token efficiency; solved in ~53s. | 397,806* | 2,857 | $0.85 | | Gemini Flash | Refused the reverse-engineering task. | N/A | N/A | Not reported | | Thinking Machines Inkling Small | Got stuck immediately and was terminated. | N/A | N/A | Not reported | +----------------------------------+--------------------------------------------------------------------------------+-----------------+-----------------+-----------------+The latest Deepseek-Flash model was the most impressive of the open models. Intelligent, fast, cheap. What more do you want? I think the Xiaomi MiMo - Flash model did well, but it might be a fluke, given how poorly the Xiaomi MiMo Pro model performed (it wandered off on a million side quests).
What's also interesting is how competitive Luna is to the lower end Chinese models. It's quite a bit more intelligent, burns a lot less tokens, is plenty fast and very cheap. After doing this experiment, I don't think it's a foregone conclusion that the American AI labs will be having the bottom of the market eaten away by the Chinese models. It seems they (at least OpenAI) can compete on the bottom end as well.
Finally, the benchmark's results reinforce a few things people have been saying about Sonnet and Terra, namely, they're not competitive for a range of tasks. For easier tasks, Haiku/Luna are plenty capable. For harder tasks, Opus and Sol are so much more intelligent, that they can get them done quicker and cheaper. Maybe there is a place for them in the "Opus designs and Sonnet implements" workflows, but in that case, why not "Opus designs and Deepseek implements?"
PS Code harness used was Claude Code. $20 in openrouter credits was used to access the models. Kimi K3 and Deepseek were also accessed directly via the API and then compared to Openrouter. Kimi K3 specifically was much faster on Openrouter, and the difference in output tokens to get the same results for Deepseek/Kimi makes me think the Openrouter providers are hosting significantly quantized versions of the model. Goes to show that Moonshot.ai's acceptance testing of providers is a useful thing, otherwise your model might look artificially bad when it's hosted as heavily quantized by providers who dont care about quality, just top-line cost and speed.
Thanks for the data-centric perspective. It reinforces my vague impression that DS 4.1 Flash eclipses almost everything out there, or at least ties it. We're really lucky to have that one.
For the past few days I've been trying to use rea to reverse engineer my Samsung Galaxy Buds 4 Pro to get seamless functionality working outside of the Samsung ecosystem with no luck (the buds 4 pro require Samsung devices + a Samsung account for seamless to work, I only have a Galaxy phone + my macbook). I think we're still far away from having something that works that well.
If OpenAI really banned him for that reason, that would be pretty stupid. If you can no longer use an LLM for reverse engineering without worrying about your account, that drastically reduces its value.
In my efforts to reverse engineer the LG firmware to validate Gamers Nexus claims, I got a warning very quickly. I appealed that quickly and successfully.
Then I applied for daybreak blue and haven’t had any issues since.
It also a matter of phrasing. Say what you want to achieve not how if it is approaching defensive or cyber stuff.
Also state you are working on your own devices and software. (If that is the case of course.)
Was this through a personal subscription? Last I looked into daybreak access, if required an organization based account.
Yes a personal subscription. Individuals can also request it. only blue for defensive questions. Red for offensive is gated to organisations that are proven to be using it for protecting.
> If you can no longer use an LLM for reverse engineering without worrying about your account, that drastically reduces its value
...for a niche of customers, from whom the revenue is probably entirely negligible to OpenAI.
Not saying it's right, ethical or anything like that, but it's not "stupid"
> ...for a niche of customers, from whom the revenue is probably entirely negligible to OpenAI.
> Not saying it's right, ethical or anything like that, but it's not "stupid"
Quite many well-paying customers become very suspicious of whether integrating OpenAI into their workflows if they read about cases where OpenAI banned accounts willfully.
If OpenAI wants to become "infrastructure provider other companies should depend on (and thus pay)", OpenAI must behave like an infrastructure provider as for electricity or internet, i.e. must only ban accounts in exceptionally rare cases that 99.9 % of society will consider to be acceptable.
I think the biggest issues here will be for individuals.
For corporate accounts they are apt to get a message to the admin account saying suspicious shit is occurring provide more verification.
>OpenAI must behave like an infrastructure provider as for electricity or internet
Oh, you mean like Google? Yea, how's that working out.
> Oh, you mean like Google? Yea, how's that working out.
Because of such reasons many companies are indeed quite suspicious of using GCP, and for the same reasons prefer to use AWS or Microsoft Azure.
It creates a lot of uncertainty. Especially if you depend more and more on them in your day to day life.
If your always-on agent hears to wrong thing, you could lose access to things you depend on. The same reason makes GCP very unattractive.
There is a Rick and Morty episode where Rick opens up a store that removes the curses from free items the Devil is handing out.
(Surprisingly, the Devil does not consider lawyering up, which is odd because I would expect the Devil to be pretty familiar with lots of lawyers. It’s fiction of course.)
It would be great to use AI to do something analogous with these spy TVs. I would be very interested in stripping the “smart” crap out of my TV to turn it into a kiosk.
If only. The EFF has been harping about this for literally decades. The DMCA in the US and similar anti-circumvention laws around the world (primarily EU) make doing that a serious crime, with punishments well beyond any conceivable impact to society at large.
> The DMCA in the US and similar anti-circumvention laws around the world (primarily EU) make doing that a serious crime
Yes, this is the main reason why I mentioned the cartoon episode was unrealistic fiction.
I’d imagine a paid service that did this could be illegal. But prosecuting people who did this to their own hardware would be essentially impossible.
What’s a better way to make money than to sell the solution to a problem you created! Don’t let Samsung read this…
"Why you bother about local AI? LOL with the $ you spend for <insert hardware here> you can buy 5 years of 20x account". Yeah. 5 years where you are told what you can do.
My guess is we're within a year of the first governments banning local LLM AI at any scale. Declare GPU's as munitions that must be registered and ban any card with more than X GB of RAM and more power than a 3060 or something like that.
That could be done with an executive order in the US, and it probably wouldn't take much to get it to pass congress.
Now, don't worry, businesses can still buy a lot more GPU compute than that, but they have to register and track their usage just like any other company making/buying weapons.
It’s quite something when western companies act more like authoritarian governments than actual authoritarian governments. I just wish RAM was less expensive ;(
5 years where you can get arrested if the AI tattles to the feds.
Just wait till your GPU legally requires an online connection to run an LLM.
5 years, assuming the terms don't change.
from his twitter account:
> if your samsung TV model doesnt have a microphone, it will look for connected devices with one, such as a soundbar or a connected bluetooth device. this gets shipped off to bixby.
Is this not the expected function when you use bixby?
(Not saying people should use bixby mind you)
an aside. I'm visiting my wife's parents and they have a new samsung TV. the remote doesn't even have an input selection button on it anymore. the default state of the tv pushes you to a big complicated home screen covered in ads and always pushing you towards their streaming live TV.
they like most poeople have a cable box plugged in, but every time the tv is switched off, the default behaviour of the tv is to forget what input it was on and goes back to the home screen where you have to use a convoluted set of button presses to get back to a non-samsung, non-app device.
You can switch this off somewhere in the settings, something they would never figure out on their own, so that it re-opens on the last selected input. I did this in the past, but then it seems an OS update wiped out all the settings, including the one that makes their cable remote work with the TV.
there is now no doubt that the TV is made to function primarily as an ad-delivery machine with explicit dark patterns built in for the task. I did them one more favour and reset their ad id and unchecked all the consent boxes on the privacy terms pages
going forward anyone who cares about their choice and privacy should not allow their TV to connect to a network at all and simply use it as a dumb link to their devices of choice, or ideally, find a better company to buy from, if that exists.
My CPAP machine has a modem in it that transmits data without ever being connected to my network. I can see a future where Samsung decides to do something similar. Perhaps even contracts with Starlink so they can transmit even larger chunks of real-time data. It makes me sick to think about that possibility, but I don't think it's far-fetched.
"SAMSUNG ERROR 306, TV will not unlock until you give it a clear connection to starlink"
Just another line in our "Please drink verification can world".
Also reminds me of the TVs that eventually crashed by collecting to much data but having no internet connection to upload it so it fills up the disk.
> the default state of the tv pushes you to a big complicated home screen covered in ads and always pushing you towards their streaming live TV.
Audible does this now too. Instead of opening up to your library and your books on default it opens to an obnoxious home screen that promotes Prince Harry and Kara Swisher's books. It's so annoying. I'm not in the market for Prince Harry's book and I will never be in the market for Prince Harry's book, stop shoving his face at me all the time. I wish we could block certain authors and books from showing up in our ads constantly but there is no way to do it!
Samsungs interface is godawful.
TVs once had a simple input button on the controller. Now there’s just the Home Screen that you described. It’s very confusing to use
I'm curious if someone has started cataloging these cases. These spying TVs seem like a common industry practice.
I believe GamersNexus were talking about doing more exhaustive testing across different brands after their video on LG models specifically.
A video on vulnerabilities in Asus routers was released yesterday: https://youtube.com/watch?v=kYnwCx2SNYY
But if you are talking TVs specifically, I expect they will get back to that too.
https://mrbruh.com/asusrouter/
Nice.If you avoided downloading the app for whatever reason, you can be proud because it just saved you from potentially getting hacked.It's been many years, but they last time I tried that app it was completely useless.
He bought a whole bunch of TVs and he’s testing them right now. Check out his YT channel.
They could have started with actually decrypting what was sent rather than a breathless 2 hours of monetised nonsense. Show me the http posts with the data that's actually sent.
This is why you buy a display panel and not a smart TV.
Or just buy the subsidized TV and never connect it to the internet. I've yet to see them adding cellular modems.
>and never connect it to the internet
That saves you from some potential issues but not others. My google search magic isn't working, but there was some manufacture a number of years ago who's TVs were crashing because they'd save up massive amounts of data to upload and either exhaust all their memory or fill up the disk because it never connected online.
Also, if you live in a household you have to make sure every other person behind that TV doesn't connect it to the net at some point for some reason.
I'll probably get a kiosk panel next time. But this time, yeah, I've just not configured the network. All content comes from the console. Which is a different bundle of problems and no better in a lot of ways, but it's more of a known quantity.
Right......
Like none of these TV's hooks to an Open Wifi Hotspot and abuse you.
Trivial to prove that, just set up an open wifi and show it connecting, getting an IP, and then transmitting data.
I haven't managed to reproduce it.
I've seen this "yeah well they'll just connect to open WiFI" for 10+ years at this point and have never seen any vaguely credible reproduction.
Best you get is "my friend's LG TV connected to the internet even after changing his WiFI password" vaguely recalled hearsay.
Or the second order retreat-to-motte argument of "well they'll add that to the firmware overnight when ad revenues drop if enough people stop connecting to the internet".
Underrated approach. Even if the screens are a tad more expensive by today's subsidized standards. There are also many FHD and 4k non-smart TVs on the second hand market, with many given away for free on your preferred marketplace website.
Claude Code for coding and DeepSeek + opencode for reverse engineering at this point.
DeepSeek will break the binary into code and Claude will search for problems. Now you have most of exploitation chain.
An interesting question is whether LLMs are going to make HDTVs more or less expensive.
Just a reminder. You can buy a kiosk TV from major TV manufacturers. You just have to look a little. Kiosk TVs are intended for lobby signage or conference rooms. You pay a little more but the garbage UI and always on piece is gone. You get 4 HDMI with them too (from what I've used).
Just don't buy any tech home appliance that was made after 2022.
I mean it may work for you, but the bathtub curve means that someone isn't going to have a chair when the music stops.
Without a reddit account
https://web.archive.org/web/20261024082057if_/https://www.re...
so is the hardware mic-off button on Samsung TV any good?
If we read the tweet - the TV then connects to whatever device it can like a home theater, other speakers and uses its microphone if it has one
Man, it would really be interesting if they were caught using the speakers themselves as microphones.
The AI in the TV was listening for dogs barking and babys crying, a gross invasion of privacy. Possibly this could have been for ad targeting purposes.
With projects like REA, it's going to get harder and harder for companies to get away with things like this.
I'd love to see the major social media apps reverse engineered and analyzed in this way. I can only imagine the dodgy practices that'll be found if this happens.
Jaded part of me thinks companies are going frontier-model-vibe-code their security into security chips everywhere, encrypted firmware everywhere, locked down APIs everywhere and it’s going to take away the fun out of it.
> With projects like REA, it's going to get harder and harder for companies to get away with things like this.
Nope, don't worry, if you clicked on the linked you'd see OpenAI has already banned this person's account. Chinese models you say? See today's new Kimi release which intentionally had all cyber-related data removed from its training data. The window is already rapidly closing and will soon be completely closed.
Won't be too much longer after that where modifying LLM weights to ignore safety rules is considered a felony.
No libre agents and LLM's?
The tile references on-board ai. “On-board ai” here is automatic content recognition. Imagine my disappointment
AI is simply an if/else statement in disguise, you didn’t know that?
Computer vision and categorization is ai. It isn't just generative AI.
There’s a meaningful difference between ML algorithms and the Grail that is (general) AI. Many of us spent years arguing that words matter and now with the VC backed hype train just about everything is AI, under penalty of stock market collapse
Because just about everything is AI. AI in itself is a rather useless term, and that was years ago.
The differences are not really that great most of the time. Quite often one type of logic in one type of ML network can be easily simulated in something like a generic transformer network and vice versa. Efficiency is typically the biggest constraint.
way to miss the point.
so glad i bought a dumb tv.
i remember going store to store in search of one.
gg me, that was a smart move.
I haven’t been able to sleep since I found out that Samsung knows my dog barks.
What a load of shit. Samsung TV always listening if you enable listening, why would they do this? /s
How dare they run a classifier locally on the TV, they should upload it uncut to Czechia like Amazon did with their ring cameras!
the point is the setting does nothing, now go back to your claude session
When I was in college this sharp phd student told me that your phone, tv, and other devices talk to each other via radio frequency to build a profile on you. I've seen zero proof of this anywhere and I still think about why he told me that.