NewEntryHN

> So the Chinese labs have thrown a lifeline to the Western loss-making labs, and I just have no clue as to why.

Because contrarily to the author's assumption, all labs, Western or not, have sufficient skills to discover the optimizations anyway, and publishing or not is not actually that important?

Good. We need AI-bullish folks to test it boldly and then report whether it works or not.

This is hard to argue without a fine understanding of how much insight OpenAI had about the stab at the problem from the "public rumor" alone.

If there was any sort of coarse insight that "they're trying to solve it this way", then both those things can be true:

- The massive amount of compute from OpenAI re-discovering Buckmaster's work solely from the coarse insight (and solving the rest as well).

- OpenAI still acknowledging they basically scooped the coarse insight using compute, and they're willing to credit Buckmaster.

Buckmaster says "Concretely, what Levent and I did was to take the Cordoba and Martinez-Zoroa program, which achieved blowup results with rough forcing, and, with a great deal of help from LLMs, push it to smooth forcing and to the incompressible Euler equations".

Could that simply be the prompt they used at OpenAI? How "stolen" would the proof be in that case?

You don't need any reproduction. Assuming researchers are using LLMs, you should just see the number of jumps increase as models get better.

> without access to the original source code

All models in the leaderboard probably have had access to the original source code in their training data.

I'm not sure I understand how important "role perception" is when following instructions from a tool call rather than the user is currently a legitimate use-case (applying steps from documentation, or shell command instructions on stdout, or really anything that can be deduced from the content of a tool call).

Very easy for French speakers ahah

Maybe shadow hierarchy are still more productive than official ones? Looks like something that wouldn't have that many meetings.

Hmm I'm not sure why in an emergency situation accessing a webpage would be easier than making a phone call.

If the thing "outside of reality" ever reveals useful to explain anything about reality, then it becomes part of reality.

I think OP's question is more how could Newton's law "pass" a test any more than General Relativity would, considering that it's merely an edge case of GR?

Why is the article titled "Newton's law of gravity passes its biggest test" if it doesn't explain the movement more than MOND?

You are comparing the fight between a p2p program and the entire music industry with the fight between the entire LLM industry and a newspaper. Notice how the order seems inconsistent.

Ads are for the free tier.

It's a spectrum. Installing Postgres locally is not 100% future-proofing since you'll still need to migrate your local Postgres to a central Postres. Using Sqlite is not 0% future-proofing since it's still using the SQL standard.

If the only argument for a piece of tech in comparison to another one is "future-proofing", that's pretty much acknowledging the other one is simpler to setup and maintain.

He says it mimicks what is described here: https://cloud.google.com/blog/topics/threat-intelligence/unc...

Which is basically phishing:

> The meeting link itself directed to a spoofed Zoom meeting that was hosted on the threat actor's infrastructure, zoom[.]uswe05[.]us.

> Once in the "meeting," the fake video call facilitated a ruse that gave the impression to the end user that they were experiencing audio issues.

> The recovered web page provided two sets of commands to be run for "troubleshooting": one for macOS systems, and one for Windows systems. Embedded within the string of commands was a single command that initiated the infection chain.

Nothing worse than what would have happened to the Astral team if they had ran out of funding rounds without an exit...

This assumes you always learn something new with every new program you write.

I would not have helped. People are losing their mind over agents "security" when it's always the same story: You have a black box whose behavior you cannot predict (prompt injection _or not_). You need to assume worst-case behavior and guardrail around it.

Netscape had 20 millions active users at its peak, out of 6 billions humans.

ChatGPT has 800 millions monthly active users currently, out of 8 billions humans.

This implication completely depends on the elasticity (or lack thereof) of demand for software. When marginal profit from additional output exceeds labor cost savings, firms expand rather than shrink.

One reason you see a pareto distribution in "normal sized" teams is not solely because of competency, but because the 80% can rest on the 20% and therefore don't feel too pressed to work that much. Therefore the pareto model breaks down in 1-man teams.

To be fair, the diction in modern movies is different than the diction in all other examples you mentioned. YouTube and live TV is very articulate, and old movies are theater-like in style.

"Software engineer complains bearing the burden of everything and concludes everything would be fixed by firing everybody except themselves."

Of course. IQ tests measures nothing more than the ability to pass an IQ test, which is proxied by a lot of things such as western culture, education, propensity to cram tests, etc.

Why doing it themselves instead of distributing the work to data owners?

This is not just branding, MCP is an implementation detail; the product is chatting with apps.

What's up with the "prompt refinement" business? Are folks trying to get it right with one shot?

My experience is that treating the generated code as a Merge Request on which you submit comment for correction (and then again for the next round) works fairly well.

Because the AI is bad you get more rounds than in a real code review, but because the AI is fast and in your command each round is way faster than with a code review with a human (< 10 minutes feedback loop).

Isn't that mostly the fine-tuning phase? RLHF being cherry on top?

Would't any additional item increase safety?