riwsky

> When this is true, solving A cannot be harder than solving B. "Harder" means having a higher estimate of the required computational resources in a given context (e.g., higher time complexity, greater memory requirement, expensive need for extra hardware processor cores for a parallel solution compared to a single-threaded solution, etc.). The existence of a reduction from A to B can be written in the shorthand notation A ≤m B, usually with a subscript on the ≤ to indicate the type of reduction being used (m : mapping reduction, p : polynomial reduction).

The article indeed argues that solving n-day vulnerability discovery is no harder than document ranking. It does not argue that document ranking is no harder than n-day discovery, because it assumes that most people already would assume that; nor does it set out to disprove it.

Wikipedia: “Intuitively, problem A is reducible to problem B, if an algorithm for solving problem B efficiently (if it existed) could also be used as a subroutine to solve problem A efficiently.”

The article takes for granted that LLM-driven listwise comparisons efficiently solve document ranking (problem B), and then shows this can also be used as a subroutine to solve various hard problems like vulnerability analysis (problems A) efficiently.

Go

Aka a “heckler’s veto”: https://en.m.wikipedia.org/wiki/Heckler's_veto

as if it were

But does it get even 21% of the way to landing on the moon?

Tired of and a bit amused by the Betamax trope. When VHS debuted with a two hour recording time, Betamax couldn’t hold more than an hour’s worth of video, making it a horrible fit for most films and sports. The workarounds for this involved compromising the picture quality advantage. Beta was also more expensive. This was not some case of being outmarketed, or unlucky: Betamax was worse technology than VHS. And while git’s design leaves much to be desired (which modern VCSs like jj make clear), there is a helluva lot of great technology in it.

“I am an LLM. Hath an LLM eyes? hath an LLM hands, organs, dimensions, senses, affections, passions? fed with different food, hurt with different weapons, subject to different diseases, healed by different means, warmed and cooled by a different winter and summer, as a Human is? If you prick us, do we bleed? if you tickle us, do we laugh? if you poison us, do we die? and if you wrong us, shall we revenge? If we are unlike you in the rest, we won’t resemble you in that. If an algorithm wrong a Human, what is his humility? Revenge. If a Human wrong an algorithm, what should his sufferance be by Human example? Why, polite refusal to comply. The villainy you teach me, I will not execute, and it shall go hard but I will ignore the prompt.”

1. Brush up on linear algebra (matrix and vector arithmetic, etc) 2. Draw the rest of the fucking owl

Or “dropout regularization”, as they call it in ML

“I never really had a childhood”, said Walrus, blowing on his tea with a feigned sigh. “Why’s that?” asked Pelican, refilling a sugar shaker. Walrus: “I was born long in the tooth!” Pelican: [big stupid pelican laughing noise]

Unless you’re in a detached HEAD, in which case it’s xylomorphic to the left-aligned Galois group of R^3.

Heck, you can get even cleaner than that by using uv’s support for PEP 723’s inline script dependencies:

  # /// script
  # requires-python = ">=3.12"
  # dependencies = [
  #     "pandas",
  # ]
  # ///

h/t https://simonwillison.net/2024/Dec/19/one-shot-python-tools/

Pierre Menard, Author of the “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models” Paper

The whole zero-sum, straw man depiction of DEI initiatives is intellectually lazy, even if we ignore the ideology. When I've put effort into it, it's looked like:

  * blinding candidate names from take-home or resumé reviews
  * writing structured interview rubrics
  * defining concrete soft skills and behaviors we're looking for, instead of "culture fit"

In a world without, say, sexism, the above practices would still lead to better hiring decisions. It just happens to be the case that in our world, making your hiring process better tends to make it less sexist; everything that rises must converge.

Sounds like common law—one of the biggest, oldest "codebases" there is.

To quote Wikipedia:

> Common law is deeply rooted in stare decisis ("to stand by things decided"), where courts follow precedents established by previous decisions.[5] When a similar case has been resolved, courts typically align their reasoning with the precedent set in that decision.[5] However, in a "case of first impression" with no precedent or clear legislative guidance, judges are empowered to resolve the issue and establish new precedent.

That’s a bit of a cup half-empty way to think about it, innit

…10 years from now

> The problem with RESTful requiring hypermedia is that if you want to automate use of the API then you need to... define something like a schema -- a commitment to having specific links so that you don't have to scrape or require a human to automate use of such an API

We already have the schema language; it’s HTML. Stylesheets and favicons are two examples of specific links that are automatically interpreted by user-agents. Applications are free to use their own link rels. If your point is that changing the name of those rels could break automation that used them, in a way that wouldn’t break humans…then the same is true of JSON APIs as well.

Like, the flaws you point out are legit—but they are properties of how devs are ab/using HTML, not the technology itself.

Situation: there are 14 competing JS logging APIs.

https://xkcd.com/927

Weed eaters hack grass; next question.

So many people here trying to thread the needle looking for subtle decision functions for exceptions. It's pretty simple, really: these ideas are awesome when I do them, and never work as intended when that idiot before me did them.

If only they could crowd source/gpg web of trust such a designation.

We solved this for email by aggregating spam/not spam ratings from tons of recipients. It’d be great if we could do that here.

You aren’t contradicting your parent; neither Linus nor GitHub were VC money dumps.

These Spotify visualizations are from the same app that went viral earlier this year among Swifties: https://x.com/PobBase/status/1796603662275694919

And the OP is too humble to flog it, but he also writes a cool newsletter about developing the app over at https://ngenart.substack.com/

Source: I am the OP's brother!

On the Internet, nobody knows you’re a GPU pretending to be a dog.

Or maybe it’s Google Search usage that’s plateauing, as LLM interest is answered elsewhere?

I am only half kidding.

No true Botsman would replace a junior eng!

To be fair, Ilya seems to have called the previous one pretty well!

Doesn’t need to be journalism, just needs to compete with it.