urams

Nikki Smetters got screwed. Clearly the best, imo.

The primary use case is the same use case as having 388 hp in a Porsche Carrera. It's cool and it's fun to go fast.

Half or more of Hacker News is constantly pushing the idea that frontier LLMs are stochastic parrots and basically useless, even in the face of the Hugging Face incident which most people also would have described as movie plot fiction until it happened. I expect biologists are even less well versed in the abilities of frontier models.

What's more, you can just try a jailbreak on a model yourself to see just how much detailed, step-by-step direction you can get to build bio-terror materials.

Interestingly multicast is used extensively by exchanges to distribute trading data to participants. I think it's the primary commercial usage outside of some IoT things.

So Opus 5 is basically "distilled" Fable? The benchmarks look often better than Fable.

> What disturbs me is that there likely won’t be a big enough reaction to this policy wise.

Anthropic was blocked from releasing Fable without any such level of incident. OAI was also briefly blocked from releasing 5.6. Why do you think there is no policy appetite?

> In Singapore, policing does not have the dystopian surveillance thing attached to it, because a large number of police are civilian conscripts.

I would not have guessed in a million years that someone would argue that country's surveillance doesn't have a dystopian vibe because the country conscripts people into service.

> However, once the cameras go up, you will never ever create one there.

This is directly refuted by Singapore, Japan, arguably the UK, and others. Singapore is very interesting to visit. Externally there's somewhat of a narrative of a harsh police state but when I visited I almost never saw a single police officer and they don't have particularly many police officers on a per capita basis (170/100k vs 440/100k in NYC). Most of the policing naturally happens by aid of the surveillance system and a very strong effort to take action to bring justice for any crime. As a result, everyone believes crime will be punished and so you end up with a very safe society. This expectation of crime being taken seriously actually _creates_ trust because you can have high faith that any criminal action someone may take will have consequences and therefore be deterred.

> We need open weights, open code and open data.

Even with this, the cost of verification would be enormous. You would need a massive cluster to repeat the training E2E.

It's a Princeton math PhD who posted. The verification is quite straightforward and was posted by the tweet author. Wolfram would have to also be producing incorrect outputs for the counterexample to be false. The counterexample works as claimed and conjecture has been proven wrong.

If you can afford to test it seriously, running both in parallel, it's worth a test to see which you prefer. If you can't, don't bother. You're not likely missing anything since they are close to personal preference with most people I know who have meaningfully tried both preferring Claude

It boggles the mind just how egregious this was.

I would not classify myself as anti-Israeli, fwiw. I just think wearing the military uniform of a foreign nation to your job governing our nation is despicable and borderline treasonous.

I will remember from now on that you can compare war to Lebron playing in a championship game but not to him playing in a pickup game.

But to be clear, the comment I replied to was one in which you made an abstract point that it doesn't make a difference to someone if you can do something but in practice don't/aren't willing to and I think that this is obviously wrong (just because Lebron didn't dunk doesn't mean he can't or is a bad ball player). You don't like the Lebron analogy, that's fine. Let's use a war analogy: in 2025 Pakistan and India, two nuclear armed countries, exchanged significant fire. Neither was willing to use their nuclear arsenals. Should we now conclude from this that they can't use their nuclear arsenals and are therefore equivalent to being non-nuclear countries? I mean who cares if they have nuclear weapons which can (can't?) kill millions if this one time their political will wasn't there for them to use them in their defense?

> Rather, this is an object lesson that war is politics by other means, and here we tried to do war without any politics and it has not gone well for us.

Be careful not to trip over your rhetoric in an attempt do display profundity. If war is politics by other means, then doing war is always done with politics. This whole statement is word salad nonsense.

Is it the case that if someone doesn't do something some time then they can't do that thing? Like, if you were playing basketball and Lebron James walked by and you threw the ball to him and said "dunk this!' and Lebron said "no, I'm not willing to" does this mean Lebron can't dunk?

Because personally, I'd still take Lebron on a basketball team even if he wasn't willing to dunk the ball that one time.

This is really a non-answer. If your point is "Dropbox is a struggling company and therefore all criticism of it ever is fully validated no matter the timeline" then any criticism of any company ever will be validated eventually which is absurd.

Point 3 was not "'viral' or income generating" and DBX pioneered one of the most viral campaigns (give-get) and generates almost $1B a year in free cash flows? How is that vindication?

> Plus Uber's only increased their revenue 11->14B in the last 5yrs.

This is just incredibly off. A brief look at Yahoo Finance shows revenue has grown from $31.9B in 2022 -> $53.7B in trailing 12mos.

Very possible Elon is doing this to make give Anthropic better chances against OAI while he attempts to reshape xAI.

Also possible he sees infra as the future of xAI if he really believes in the value of space compute.

Hard to see this any of this as anything other than a bearish sign for Grok though.

I think you would be right if their post was substantive in relation to the topic, but it's not. It's a list of grievances almost all of which are unrelated. Despite this, it was at the top of the replies to the topic.

It should be noted that this user is basically an OAI shill account. You can look through their history to see this quite clearly.

Anecdata, but I have a friend at OAI who claims that on both twitter and HN there is mild coordination of OAI employees to signal boost pro-OAI and anti-competitor messaging.

Yes, I have a similar only usage but using JSONB_EACH. You can actually replace "CROSS JOIN LATERAL" with a comma which I think is clearer.

However, it working as a lateral join is critical as you need the function to fire for every row.

Probably the biggest issue is that it's primarily useful in the context of exploratory analysis and makes iteration on a plot much slower, requiring you to re-run the query to get a new viz. Iterating on a viz is best done with the data cached locally or elsewhere.

In the context of a query used for a dashboard in prod, you're likely using a different viz environment so it's not useful at all there.

What feels like a lifetime ago, I made almost all of the R tooling at Uber and actually implemented what was effectively exactly this on top of the R DB tooling. Everyone, I think correctly in retrospect, I showed it to thought it was more or less useless. I had hoped it would be a nice way to pull and instantly visualize data but that was rarely all that valuable like this.