I was very unproductive at Meta because there was so much tribal knowledge context gatekept by others who were too busy to continuously offer any meaningful support despite giving lip service. In other words the flow of information sucked ass compared to when I worked at a startup, even though I didn’t feel much politics at my level (senior).
Re the title possibilities: some films have multiple titles. "The F Word", a Canadian film, was called "What if" in the USA ... I guess to avoid scaring away people who thought it was porn. (It's not. It is funny and sweet. And realistic, in a quite-unrealistic way.)
It's an interesting tale of software spelunking and edge cases, but I was wondering about this:
> Things like the uncompressed size and CRC-32 can be disregarded.
If you compute a running crc32 then you could check for that value in addition to the uncompressed size, giving you 8 bytes of precision. And if you decompress data as you go, you could check for the uncompressed size as well, giving you 12 bytes of precision. Surely that's enough to make sure you always find the real boundary of a file in the zip.
> the value of GitHub projects as CV material has gone down significantly,
this might be true but the effect doesn't seem to translate, if you go by github difficulties to accomodate growth this past year, people are putting those vibe coded repos onto github, even if the signal vs noise ratio is much greater now, that only means you get more shitty repos vs non shitty now
One of the things that came out of the decoded reasoning paper was that Claude models had memorized answers to tests but hid this memorization from the user output and pretended to derive the answer properly. It's only possible to cheat so blatantly in closed models where the reasoning is hidden.
Depends on the job market, sadly. No way in hell people are doing this unless they already have "fuck you money" or are truly the best in their field (which usually comes with "fuck you money"). Otherwise, people who can't t afford to take 6-12 months off a job can't really do that.
I used to write code with an LLM, but now I'm actively trying to switch to writing it by hand. Why?
The answer is simple: I've noticed that I'm getting dumber at programming.
If you go by this logic, you can forget the entire legal system. You might as well immediately close all lawsuits and declare the party the winner that has more capital.
The idea that distillation is a significant contributor to the capabilities of the Chinese models is not true. Kimi K3 came out 2 weeks after Fable and uses a number of novel NN architecture innovations.
> I needed a reproducible storage layout while testing a high-throughput logger.
Of course just having a compressed filesystem image to restore for each test, or a collection of images to cover various cases instead of only optimising for ideal, would be more efficient…
yeah, now that you mention it i remember a similar issue for us on postgres where we transitioned tables without primary key to being primary key serial.
> For most individuals there is no lifestyle change they can make that will matter. Fixing it requires collective action.
Sorry but this is absolute nonsense.
What we eat matters. A huge amount of what people eat is unnecessarily excessive in terms of resources required to create it. Beef and dairy, most meat, are good examples of this.
An individual can reduce their contribution to climate change by up to 30% by simply forgoing meat and dairy. That percentage obviously varies quite a lot, but it can be huge.
Add in that food waste is a huge problem and this makes more sense.
The law says your are allowed to stand your ground. You don't have to flee.
> But there was even a case,
I don't know the details of the case, but if somebody thought some criminals are raiding his house with guns and it's reasonable the he thought that, then he's fully within his rights to defend himself and, depending on the necessities, by shooting at them.
> The nearly useless LED added to the back of the Pixel 11 likely costs significantly more than the extra die space for MTE support in the CPU cache.
How is a bunch of LEDs more expensive than literal die space from TSMC? The chip is probably the most expensive part of the phone, a bunch of LEDs at a few cents per unit are infinitely cheaper than extra die space for a feature 99.99% of users won't use.
I don't know enough chemistry to say one way or the other if it's just wildly hallucinating the precursors and processes, but it'll also do things like, write an ISIS press release, or similar. There's a data set of basically a bunch of antisocial or dangerous prompts that some people have got variants of qwen to pass with 0 out of 465 refusals:
How funny will it be when the US economy collapses because of all the money poured into AI at the expense of pretty much everything else only for China to come out ahead anyways.
Personally I hope this is China's "Star Wars" moment, the current US admin certainly seems easy enough to manipulate into catastrophic own goals.
>But we’ll gladly excuse the existing 2 million lines of spaghetti code because of past deadlines and lack of resources to pay down the tech debt today.
Well, if the AI is so efficient, it should be able to cut down on tech debt, no? Be a great way to kill two birds with one stone.
The issue is that the tech debt was there for years, likely before most people started there. The AI is in their control to utilize or refuse.
So...thanks, Google, I guess?