chronogram

I don't see the author complaining about it, instead he thought it was an interesting story. So I guess he already did that.

I have my business email read first by an LLM, I imagine someone who works at a tech-oriented website does too so those emails are basically filtered into a pile anyway.

Put very simply, that is how they're supposed to work. They are supposed to result in the most likely result. Same reason they, unless made not to, output '17' when prompted to "pick a number between 1 and 30", because that is the most likely result from training; but that means that all LLMs output 17 except for forced temperature or reasoning in RL making them run a python command with random.randint(0,31), or that the prompt is recognised and something else is supposed to be chosen than what is statistically the most likely text response.

Therefore asking a group of people would result in more varied outcomes of something visual, but from the textual training data the direction is statistically the most likely so is what is output, that sky is the most likely, time is the most likely, cloud coverage, etc. You're prompting with as little as possible so it fills in the blanks with what as likely as possible. When you ask Alex from the North Pole they're going to fill in the blank with a different kind of weather than Bob from the Equator.

That does mean that an LLM itself is in essence not as creative as a human, but then it is a tool to be used by a human doing more than oneshotting "Generate an SVG of a pelican riding a bicycle" and expecting creativity from the tool rather than the user. If the user is lacking in creativity they can prompt an LLM to get wacky scenarios to change it up a little.

What about the quality of the transcription?

Looks really promising. Once hardware codecs and camera support arrives it might be useful for old laptops. https://github.com/LineageOS/android_device_mainline_generic...

Assembly doesn't add anything here. You'd still need to implement the IO, portability, safety, etc etc. You'd end up with the same thing so you might as well use Go from the get go. Assembly is useful for dedicated SIMD like in dav1d.

It's part of their job. If they can't do that right they certainly can't do the rest of their job right. Just like it's OK to get queazy if your doctor is functionally illiterate or your lifeguard can't swim. Also things that match average people in some areas.

Is that your first time using an LLM? I don't mean that in a snarky way, it just seems like things everything could do for years.

What exactly do you mean by desktop security? Security from websites, installable software, physical access?

Same here. Astra has been the best thing I've seen. Astra on Low has been my favourite thing so far. Higher levels just mean more cruft, not useful.

I use it in the car to talk to because it's built-in, mostly for navigation via voice. It seems to be an old or quantised model, because it seems a few years old, and it has strict rate limits unlike Gemini, where it really just shuts off if you talk to it a few times in one day, but it's still fun to show passengers who are still new to LLMs.

They can't increase capacity by raising prices. Right now they calculate that they're at an optimal revenue curve. Simply reducing demand by increasing prices doesn't mean they get more money. The limit is in the ability to buy hardware.

Absolutely nobody commenting that has done that, it's just roleplay. Flash-0731 and Pro-0813 replaced previous models too, and the V4-preview models replaced V3.2 before that. You have your official API from a tiny cutting edge research lab that can only realistically host one model at a time, which you know from every model they released before, but they fully openly provide every model so if you want that infinite stability you can easily host it yourself for eternity. So those commenters want to pretend they require RHEL-like stability for their prompts with enterprise budgets, but somehow can't host those models, yet at the same time offload their entire RHEL-stability requirements to the research lab. Even Google and OpenAI retire models that were still relevant a year ago, but open models actually last forever.

The original from 2021, so 60% health after 5 years? Coupled with how they don't have a bypass sounds to me like they don't have their battery in order. Your linked thread also mentions that it happens at 100% too, which would explain the extreme degradation: leaving it plugged in by default thrashes the battery without a bypass.

Not OP, but I know that's not mainstream for phones at all. I think only Pixel phones enable it as a normal thing, when you have the 80% charge limit enabled. Sony (since 2020) and Samsung (since 2023) have it, but only when you enable it within some game settings for selected apps. Maybe gaming phones have it because they're plugged in more than others and care about heat. Apple doesn't do it.

I started programming this way as a kid with books and notes that I'd then periodically would be able to enter into a computer to see if it runs.

We'll make sure there will be enough regulations to have neither the electricity to power it or the companies to build it.

I use it directly: https://platform.deepseek.com/usage

3rd party providers on OpenRouter can be cheaper but it's already so cheap.

A fine doesn't give immunity for the future. You pay the fine for the past behaviour, then you start paying fines of up to 5% of daily global revenue for every day you keep being non-compliant. That's the start. The end is not doing business.

I'm not so sure. Especially with 3.6 Flash being 24 to Sol's 22. The 3.x Flash models were thought to fit on a single TPU 8i and it's the odd one out in that list at a whopping 234.7t/s compared to Sol's 64.4t/s and Opus's 56.3t/s. Even 3.1 Pro is a much higher 113.9t/s than the others.

I think AA-Omniscience Accuracy follows your expectations better. An ultra size Fable at 61%, followed by large frontier models like Sol, 5.5 and Opus. With Flash being up there. I assume because Gemini is more focused on general knowledge to operational cost in particular, rather than getting the highest scores in coding benchmarks. If you go to Domain Score (Normalized) you'll see that the Gemini models are only less competitive in Software. And that's where Sol goes from 6 in Health to 71 in Software.

What part of Afterburner caused issues? I have it installed to undervolt an old GPU but not actively running.

An LLM is stateless so every time you send a message a command is run with the parameters, the prompts, and a copy of the chat history when using one of those chat interfaces, and then it returns text. If the output includes explaining it to an LLM that would then be included in the copy of the chat history placed in the new command.

It would be a bit like opening a text editor and typing in that you want to increase the font size. Someone external to the text would have to come along and click the font options.

This is amazing! I can see in a few years having Home Assistant use an LLM to coordinate my home while I'm away. Tell the LLM to make an inventory of my home and then send the vacuum and the robot around to clean everything up. A multimodal LLM can inspect pictures for items that need to go in the dishwasher, the laundry or needs polishing and reorganise my closets to best allocate space, make the beds, take care of the plants. It has all day to do it while the solar panels provide ample energy anyway.

And DeepSeek has been making huge progress on efficiency, and publishing about, so they came with a 1.6T model that is both fast and cheap to run.

I use cjpegli as encoder and it compresses best with its default progressive and full 4:4:4 approach, so it's not only a nice to have feature.

So it's like a more limited Streamdeck.

The US already has their Strategic National Stockpile for medicines.

Shorter expirations are managed by constantly selling the old to the domestic industry and purchasing new, like China does for grain and frozen meat while simultaneously being able to keep the market more stable by selling high and buying low. Switzerland has a lot of stockpiles, even including coffee[0], which the companies also go through FIFO.

0: "The 15 big Swiss coffee retailers, roasters and importers, such as Nestlé, are required by law to store heaps of raw coffee. Together, these mandated coffee reserves amount to about 15,000 tonnes—enough for three months’ consumption. The government finances the storage costs through a levy on imports of coffee. All 15 companies are in favour of maintaining the coffee reserve—as long as they are paid for it." Economist 2019

Or travel. Even in the developed world you can be without internet or slow internet. I have Gemma 4 E4B on my phone that can process audio, image, and text if I have need to.

Even if it reports 44.1 it resamples internally to 48 in all hardware I've seen.

Yes. Section 5 talks about real-world deployment: 5.1: "The DSpark draft models are co-deployed with the preview versions of DeepSeek-V4-Flash and DeepSeek-V4-Pro"; 5.4: "MTP-1 represents the former production setup, having been superseded by DSpark two weeks following the DeepSeek-V4-preview release."

mpv is that. In 2010 mplayer2 forked from mplayer, in 2012 mpv forked from mplayer2. It's all I've used since.