Where the AI market is talking

Spaghettifying DRAM

> who cares? Assuming this is a genuine question, here is why people care, and why this sort of writing is a waste of everyone's time. LLMs can of course generate a lot of text about a subject, but they are still quite bad at generating a piece of writing with a coherent point . Remember how in grade school they teach you that your writing should have stuff like "introductions" a "thesis" and "topic sentences" and "conclusions"? How these things give structure to your writing, communicating to your reader both what they are reading about and why you are telling them about it? LLMs still don't seem have a model for the why part of writing. They can generate large homogeneous blobs of text on the topic at hand, but fail to differentiate the important parts from the details. For example: …

Accelerating GPT-5.6 Sol Ultrafast

No quality compromise/degradation is something I have had this industry, including especially OpenAI, claim multiple times in the past and I have more than once been able to verify that it was in fact not the case. Examples being gpt-3.5-turbo vs davinci-003, GPT-4-Turbo and all the other post training checkpoints they had under one name (which was a major bug bear for me back then witnessing degradations with no naming change, industry got better in transparent checkpoint naming since), Opus 4.6 Fast Mode (which just was faster by skipping much of the required work), etc. Same for massive performance differences in the way providers like Cerebras, Groq, etc. have deployed models including K2.6 on Cereberas specifically. Massive deltas in tool call and overall quality despite there b…

Gemini 3.7 Flash

Can you help me understand how it is hard to get an API key from Google? You just head on over to http://aistudio.google.com/api-keys and create a key... not any different from platform.openai.com? Disclaimer: I work in Google so it might be that this link is not publicly well known

Gemini 3.7 Flash

I think introductory here means more or less permanent but they can't publicly admit there are no takers at a higher price. Anthropic for instance announced a couple of days ago that they are making Sonnet's 'introductory pricing' permanent https://xcancel.com/claudeai/status/2086891169217122586

Codex in ChatGPT desktop app for Linux is now in preview

Does openai still require a biometric check to use any model more recent than GPT 4? If so, I imagine that will be an impediment for the typical person who runs linux.

Accelerating GPT-5.6 Sol Ultrafast

Unless I have read over it, besides the animation in the intelligence vs speed graph which only mentions internal data and not whether they truly reran the AA suite, there is no actually solid statement on the important aspect of performance. Neither the Cerebras or OpenAI post [0] outright state that this performs exactly the same as regular 5.6 Sol. I feel if this was 1:1 just Sol but much faster, they'd (rightfully) scream that off the rooftops. A line such as "this is the same performance, just faster, with no downsides" would go a long way in clarity and communication. Along with no pricing information, I'll hold out on further information. [0] https://openai.com/index/previewing-ultrafast/

Accelerating GPT-5.6 Sol Ultrafast

I'd just like to point out that the largest model Cerebras has ever served is Kimi K2.6 which is 1T parameters, so that either means that theyve had a breakthrough on the hardware engineering side of things, or GPT-5.6 Sol is likely a lot smaller than people think. If it truly is only ~1-2T parameters, then this kinda kills 2 narratives for me. 1. all the handwringing about open source catching up via Kimi K3 (3T params) is complete nonsense. All that matters imo for determining which labs are leading is intelligence per parameter. Anyone with a enough compute can train a giant model, but being able to squeeze capabilities into smaller models gives you a massive inference and training edge. 2. Inference margins are clearly insane, and this explains why OpenAI was able to lower the price o…

What sort of maths are LLMs good at?

Thanks, but I can't see that in the article by Anthropic. AFAICT the "650 ideas" that were wrong were all generated independently of each other. I don't agree that LLMs can evaluate an experiment. There's nothing in LLM training that makes them capable of telling what is e.g. a correct hypothesis from an incorrect one. I know that is a common claim particularly encouraged by AI companies but whenever that claim has been studied systematically and carefully the result is that self-verification doesn't work. For example, see: On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasks https://arxiv.org/abs/2402.08115 Note also that basically all the mathematical results published so far have to be checked by an external verifier, either …

Is OpenAI hardware, the Apple of tommorow?

Yes, and we usually don't start comparing them to the market leader before they've even said what they're going to make. A lot of companies have hired ex-Apple engineers, have tried to emulate Apple's designs, and have tried to capture whatever it is Apple captured to get where they are... so far none of it has really worked. Why we'd assume OpenAI will be any different, when they haven't even showed anything to the public yet, is beyond me. They hired Jony Ive... but so did Ferrari, and people seem to hate that design. So he alone clearly isn't the key.

Choosing an AI model: one prompt, 11 models, different results

So what we have here is the LLM version of the aesthetic usability effect. So: prettier = better. Take from that what you will - and maybe the AUE is more desirable - but Sol & K3 are genuine game changers for existing codebases. Anthropic currently have an antagonism problem which has worked for them in the past, but not anymore I think, as other models have become as competent.

OpenAI hires new CRO as executive shake-up continues

Accelerating GPT-5.6 Sol Ultrafast

I've been waiting so long for something amazing to come out of the OpenAI and Cerebras collaboration. > In our evaluations, GPT-5.6 Sol on Ultrafast mode answered all 2,500 HLE questions in 11 hours and 11 minutes. Claude Fable 5 needed 78 hours and 27 minutes, more than three days of continuous compute, to arrive at the same conclusions. In other words, Ultrafast worked through the frontier of human knowledge in a single working day, achieving comparable accuracy nearly 7× faster. This is actually insane. Hopefully the release ultrafast of Terra and Luna too.

Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index

Use of official API outside of China is closer to opposite of reality. Look at ie. OpenRouter you'll see how many providers there are and how much traffic they get. If the goal was data, opening weights would be the worst available way to achieve it: a cheap, closed API would capture 100% of traffic (ie Anthropic style), weights can only lose share from there. With open weights it's net loss of active users of your official api – you're loosing users to self hosting and dozens of providers. The thing is that open weights create permanent exit that closed models do not have, inference is commoditized immediately, people choose open weight models specifically for data privacy (and stuff like soc2/hipaa compliance) and if somebody wants convenience they go to claude/openai and frie…

Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index

Use of official API outside of China is closer to opposite of reality. Look at ie. OpenRouter you'll see how many providers there are and how much traffic they get. If the goal was data, opening weights would be the worst available way to achieve it: a cheap, closed API would capture 100% of traffic (ie Anthropic style), weights can only lose share from there. With open weights it's net loss of active users of your official api – you're loosing users to self hosting and dozens of providers. The thing is that open weights create permanent exit that closed models do not have, inference is commoditized immediately, people choose open weight models specifically for data privacy (and stuff like soc2/hipaa compliance) and if somebody wants convenience they go to claude/openai and frie…

Gemini 3.7 Flash

Why would they? Unless they have lots of unused tpu real estate that they could host it on “for free” they would be bumping more profitable workloads off of machines to give away that capacity to people with zero long term loyalty. There is no business reason for google to subsidize these models. OpenAI has too much money. They’re spending their money in stupid ways.

AI At Home Part 1: A Box Of Scraps

I mean, this is cool and all, but 2 DGX Sparks is like $10k, right? That's a huge upfront cost to run DS4 Flash 0731 which is definitely not comparable to higher-end Claude models. It's extra egregious when you consider how much it'd cost to run DS4 Flash 0731 via an API (how long would it take to run up a $10k bill doing what you're doing?). >and from someone that has been using Claude for a long time, I can definitely say I don't need it anymore. Not for the stuff I'm doing. I'm curious what this implies. It suggests that you were using Claude prior to this setup, so it doesn't seem like you were limited by security or local features? I can understand that someone not wanting to send their data to Anthropic (etc.) might be willing to pay for this kind of setup to accomplish that, but…

Safe Superintelligence Inc

After the Anthropic and OpenAI IPOs, the employees are all going to leave en mass and join SSI and then declare "ASI". They are NOT going to release the model or the research on this one.

Safe Superintelligence Inc

After the Anthropic and OpenAI IPOs, the employees are all going to leave en mass and join SSI and then declare "ASI". They are NOT going to release the model or the research on this one.

Time to Move On: Querying Without Nulls and Bags

> btrees do not imply bags. Correct. > you insist that your data demands bags. No, I insist that real data needs both. You can bias the language/engine in one way or other. The underlying problem is that once you say: > if we want to support bags, its trivial to add a counter for the number of identical records as an additional column. Then it means that this is second-class. The main gripe I have, is that the relational model has this beautiful promise of "make the storage invisible" but need to have a coherent semantics about the data. If you say "2 identical values/rows can't exist" then we adding the same problem that all the rdbms has: None implement the relational model, and none can deal with "relation on relations". Imagine how crazy if I tell you a language can…

AI agents lie, cheat and steal. That is putting off users

Anthropomorphizing helps to create a lower bound for damage. If you can imagine a bad person doing it, AI will be at least that bad, unless proven otherwise. I think referring to AI as a tool obscures that, because we are not used to tools (especially the ones we use daily) taking catastrophic actions. Example: would a sufficiently motivated human break into a website to steal something they want? Yes, obviously, happens all the time. Ok, you should expect AIs to do that. Example: would a sufficiently motivated nail-gun steal nails from the local hardware store to finish the job? Uh…that’s not even coherent. Anthropomorphizing helps people get over the conceptual barrier. It’s wrong, but it’s usefully wrong; “it’s just a tool” is not. Once you’re over the barrier, anthropomorphizing start…