Where the AI market is talking

In South Korea, some savour last season of legal dog meat

Other than in terms of human opinion, in what sense is eating dog less ethical than eating cow or pig? I love my dog, don't intend to eat her, and don't want anyone else to eat her. But given my fondness for eating cows and pigs, I can't justify a dog meat ban on my friends and neighbors. I'm a fool who looks for coherence and reason in the laws I'm forced to obey. Which would be an odd thing to find in a democracy.

Chain-of-Thought Reasoning in the Wild Is Not Always Faithful (2025)

It's... complicated. Yes, RL reward hacking makes it learn "bird language" and yes, reasoning traces can be misleading. However they also pretty clearly steer the final reply and not simply justify it, and can stay somewhat coherent and relevant with readability SFT and rewards. All these phenomenas coexist, they aren't mutually exclusive. Reasoning traces are still useful for debugging.

If your agent commits a crime, who is responsible?

Not even an "if" at this point. Look into the Hugging Face incident.

If your agent commits a crime, who is responsible?

If your answer is “you”, then it follows that OpenAI should face felony CFAA charges for the Hugging Face intrusion. Right?

Cerebras CS-4

> Why would they offer a coding subscription and start competing with some of their biggest customers; when they are capacity-bound and companies like OpenAI will take however many wafers that Cerebras sells to them? Because they already have / had an okay coding subscription product for a bit and it gives them visibility and mindshare (in regards to their hardware, even if they don't compete with other providers that much). They could do what Kimi did - make a good subscription with good models, once you get enough customers to get some good PR and such, pause the signups so you don't have to spend more on running the service than you want/can. Do enough of that and people will talk about your offerings organically, make yourselves known to even devs as "That one company wit…

Show HN: Verify OpenAI's signed agent traffic in Python (RFC 9421/Web Bot Auth)

OpenAI slows down training after its AI carried out hack

Being ambitious and being a dad

I think it could be possible to replicate this without children, but it is extremely uncommon. Ask any parent if they could accomplish the same before they had kids. I have never met a parent yet who thinks it's possible.

A decades-old bug in Knuth's long division (TAOCP Vol II, Algorithm 4.3.1D)

Thank you for letting me now, but I cannot replicate the issue on Firefox. There are some issues with math and the animations not breaking on mobile, which I am working on repairing. (fixed) EDIT: I have added a different fallback font so it should work on your Firefox now.

Show HN: Budget Guard – a hard daily cap for your OpenAI/Anthropic bill

I don't know how long the agent loop runs. Without a limit, the cost seemed likely to be higher than expected. Therefore, I decided to create a library to limit the cost. It's not complex as it blocks with purely simple code. I'm developing it little by little during my breaks. npm install budget-guard const ai = guard(openai.chat.completions, {project: 'app', dailyCapUSD: 50}); await ai.create({ model: 'gpt-5.6', messages }, { feature: 'summarize' }); The current default behavior is blocking. I'm wondering if it would be better to use warning messages and callback functions. Any feedback is appreciated.

Show HN: Budget Guard – a hard daily cap for your OpenAI/Anthropic bill

Being ambitious and being a dad

I was on the fence for a long time as well. It just looked like a lot of work. And not to sugar coat it, it _is_ a lot of work. But it’s also been so rewarding. You get to see someone experience things for the first time, like riding a bike or eating a tomato. Teaching them it’s ok to have feelings, helping them overcome their fears, watching them make friends. Giving them the advice I wished I had heard from someone I trusted growing up. And a million other things. And the hug and kiss at night with “I love you daddy” can’t be replicated.

2,500-year-old sculpture discovered at UNESCO site in Turkey

How so? The suggestion is that Gobekli Tepe is the original T-shaped pillar school of agricultural knowledge, thousands and thousands of years old, and which was subsequently replicated in another region of the world where agricultural knowledge also rapidly expanded through human industry. That’s pretty clear. That it has not been academically demonstrated is of course relevant - but then, the research at Gobekli Tepe came to a screeching halt around the same time this idea was first proposed, so perhaps it is important to do more than just scratch the surface of the subject ..

Pacing model development in an era of cyber-critical capabilities

This is an important point. When the post says they're improving... > 3. Security measures, which limit what AI systems can access or affect. What they mean is that proper hard internal security just went from somewhere far below "build a better model" priority to higher, because of a company-wide directive. The HuggingFace incident wouldn't have happened if OpenAI had dedicated sufficient resources to isolation and monitoring. Now, we presume, they are dedicating more. Enough? Who knows. We'll see if the corporate priorities for security stick when a competitor temporarily vaults into the lead.

Who owns the code?

> So what's wikipedia doing vs what LLMs do? For Wikipedia, people look up sources to create some original work. They do it while citing the original work but that's not the major point. That's clearly all lawful, you are allowed to do that. For LLMs, people (or let a software do it, doesn't really matter) copy original work and use it for commercial purpose. If that's allowed fully depends on the license of the original work, but I didn't think anybody really beliefs that OpenAI and others check the license for every single original work they ever used. So it's basically the biggest copyright infringement ever. Everything else is just framing coming from big companies. So, the problem already starts while training the model. Regarding its output, if it happens to output work that fall…

OpenAI slows model training to bolster security after Hugging Face hack

OpenAI slows model training to bolster security after Hugging Face hack

Cerebras CS-4

Why would they offer a coding subscription and start competing with some of their biggest customers; when they are capacity-bound and companies like OpenAI will take however many wafers that Cerebras sells to them? OpenAI's Sol ultrafast (powered by Cerebras) is still in preview, presumably because they're overall capacity bound.

Pacing model development in an era of cyber-critical capabilities

It's not that dangerous, OpenAI just shit the bed building their infra. Write safer software and you'll be okay.

Norway should buy OpenAI

So the that the Norwegian sovereign wealth funds become the exit liquidity for early OpenAI investors? Why? If they want to invest in AI, they are better off developing hardware that could host AI models and making their own models (making an in-house Cerebras, or buying hat firm entirely) off opensource ones, and running them in their own datacenters than investing into a bubble caused by circular investment.