> 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.

Really? There’s no “spyware free” marketing angle? Even Amazon has always sold a Kindle variant with no ads for $20 more.

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.

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.

[0]: https://arxiv.org/abs/2610.10072

Anthropic is a joke of a company.

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.

It's just an adult human body, Michael, what could it cost? One('s) dignity?

Also leading the frontier is much much harder than catching up to frontier.

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.

Nothing particular special about 5.3 Flash -- you might as well run DS 4.1 Flash at this point, as it's better in every way. But there is as yet no all-around replacement for the full 5.3 model.

> 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.

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.

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.

But if you are talking TVs specifically, I expect they will get back to that too.

Whereas your CPAP machine is likely a high margin device and the reason it has that modem is because insurance companies require compliance data in order to continue providing coverage for replacement supplies. Resmed or whoever is happy to bundle in the cost of the modem + service in order to keep selling high margin supplies.

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.

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.

Like none of these TV's hooks to an Open Wifi Hotspot and abuse you.