JSR_FDED

I used to think you were only cool if you used AWS. But then I was cult-deprogrammed and now I feel smart every time I spin up a performant and reliable VPS for peanuts.

See this is why I never understood LISP people proudly proclaiming that “code is data and data is code”… we were doing this in BASIC for decades already!

I’ve also noticed how powerful actual keywords are when trying to retrieve data from a sea of candidates that have high cosine similarity to the query. My quick and dirty ranking mechanism is simply keyword match first, and only then similarity score.

Super cool for how compact yet still readable it is. Shouldn’t there be a tool description passed to the LLM though?

Srinivasan had demanded a meeting with Malaysian Prime Minister Anwar Ibrahim, threatening a major withdrawal of tech investment if his demands weren’t met.

That is so NOT how things work in Malaysia.

Good riddance to these deluded Network State believers.

It’s wild that a single DOM read can tell which OS you’re running based on how the browser hyphenates words.

Details like this give you an appreciation for how hard it is to mimic a real user if you’re trying to automate web tasks like scraping when the site doesn’t want to be scraped.

I’m so tired of this. These execs saw the threat a long time ago, but then it was in their interest to not have pesky regulations get in the way of world domination.

Now they’re being outcompeted and so we should suddenly have regulations.

Time to put your big boy pants on and compete.

hyping an open model that isn’t open

Moonshot has committed to release it end of the month.

It would only be inconsistent if the router chose different models for the same task.

Considering that of 89 terminal tasks there were 11 that only Kimi got right, and 7 only Fable - having a router gives significantly more consistent results if you define consistent to mean correct.

They share their methodology and results. I learned things about the relative strengths and weaknesses of Kimi and Fable I hadn’t seen anywhere else. Should being in the model hosting business disqualify them from sharing?

Very interesting. They test Kimi K3 and Fable on a set of approx 1000 tasks grouped into 5 areas (SWE, Legal, etc).

They put a router model in front that predicts whether Kimi or Fable is going to give a better cost for a correct result. (They believe that ultimately such a router model should be continuously trained on your own workloads so it makes the best decisions for you).

Their router chose Kimi the majority of the time (72% in one category, all the way to 96% in another category), leading to cost savings in every category (from 1.5x to 50x depending).

It’s Cyber Ninjas all over again

If ever there was an article that needed a video or two, this is it

Couldn’t happen to a nicer company

Fortunately to keep you going while you wrestle your way through this turgid writing they intersperse the same photo 18 times

Zune would like a word

The judge called the outcome disturbing, as it leaves victimized children as "collateral damage" of privacy protections.

As sad as this is, end to end encryption means no CSAM scanning.

As an alternative Apple previously tried to do scanning on the phones locally but caught hell for that too.

This is one of those unfortunate tradeoffs but I see no alternative to privacy taking priority.

This website is an embarrassment

Maybe it’s the music of the first video, and maybe because it’s late, but this made me quite emotional. On the one hand it’s like a child learning to interact with its environment. On the other hand it’s a requiem for yet another thing that used to be human.

To me all these coding tools look indistinguishable from each other. If you changed the logo on one I wouldn’t notice.

Because you’re not that important. I don’t mean that in a mean way, just that if you’re a serious player in national security or something like that, you’re already not using Gmail, let alone ChatGPT.

Oh give me a break.

Using a Chinese LLM will not put a Marxist under your bed.

Did you know Gemini is shockingly bad at French poetry? Hasn’t stopped me for using it for all other tasks though.

Remember the Panama papers? That was a Wordpress hack.

Unlike iOS/Android choice which has broad personal ecosystem implications, changing IP address to another LLM provider is effortless.

Is there a story behind this?

I find the fact that 39% of papers use a style that signals lack of effort somewhat worrisome. Even if that’s 100% wrong, and they’re all high effort papers, the fact that they give off the same aura as low effort papers is a problem for the authors.

Spending a lot on making your top developers more effective is easy to justify.

But once you have an AI-powered application in production, why wouldn't you go with far cheaper and capable enough models? That's just optimising a business process like any other. You'd use cheaper providers whether Chinese, or other on-prem models.

I think we'll pretty quickly see a best practice emerging that any generated code will be subject to an additional pass scanning for vulnerabilities. The scan will be done by a different model than the one that created the code. That will help catch vulnerabilities created by models, whether intentional or not.

This should be done regardless of which model was used - American or otherwise.

Either that, or they don't want to be hostage to a handful of companies intent on owning the future. Personally I'm right there with them.

Kind of a problem though, if that’s who you’re getting mixed up with.