News for hackers. Stuff that matters.

redanddead

Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @10:56PM (#50347863)
lmao who cares what they ban
Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @07:35PM (#50346379)
it is
Re:Skill to Create Product Videos on Monday November 02, 2026 @06:54PM (#50346014)
From an outside perspective, the issue is that you're still trying to engineer what should be a psychology tweak
Re:Egoless Mathematics in the AI Era on Monday November 02, 2026 @06:51PM (#50345976)
Human academia needs to clean up its ego fixation, jesus fucking christ

It'd be pathetic if the silicon does better science than us and our crabbucket mentality -- highly likely in my view...

Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @06:39PM (#50345809)
Looking forward to more bricked $60M racks on ebay
Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @06:38PM (#50345793)
must be superintelligence then
i would love to spend >$100k on a great electric car, idk if they're quite there yet and the rocket boosters aren't doing it for me in the tesla line up lol

Still needs a step change upgrade on the battery tech

there's the scaling
Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @06:02PM (#50345325)
Let's analyze the premises

1/ The internal users at labs and external users are completely different.

2/ The lab people use the models to create recursive systems, systems to create prompts or synthetic data sets etc etc etc. Way different mentality than "fix the bug, make no mistakes"

Is it really so farfetched to think they've been having models fuck with the weights of models? I think it's entirely possible, probably even likely as a way of generating permutations of synthetic datasets that will then be included or excluded

Re:Astra can read and write weights to improve Qwen on Monday November 02, 2026 @05:55PM (#50345221)
How is this such a big story on HN. It seems to me that this is just another engineering problem, not particularly surprising
There's not infinite hydro... net new energy is coming from increased efficiency (insulation etc) and upgraded turbines
RIP insurance industry
> I was taught to gut a squirrel by the son of a freed slave

Why gut squirrels

You and orangecat are using a false binary

That’s not how science works

Show me correlating graphs to the rise in colorectal cancer. There has to be a meta-analysis about this, my money is on food being a major contributing factor

There’s a lack of fiber

There’s an overconsumption of sugar

Sugar is awesome for cancer

It’s the food

This isn’t some mystery, this isn’t happening in a vacuum. There are correlated trends like rising profits of the food industry, food becoming plasticized. Everyone’s fat now

Author needs to consider this
The Zimbabwean mint has never been busier
you were consulted by an LLM?
Re:Amazon to buy satellite firm Globalstar in $11.57B deal on Wednesday October 28, 2026 @12:03PM (#50275903)
Very high value acquisition

How did they not become spacex/starlink

How high?
That plus organs on a chip could be scalable and generate interesting data
Who specifically was involved in that and what are they up to now?
i think it will be
Re:Anthropic S-1 on Monday October 26, 2026 @01:49PM (#50247927)
Not everything needs to be an argument, and regardless you didn't change my mind

The value is in the harness! The harness leaked itself

Re:Anthropic S-1 on Monday October 26, 2026 @01:45PM (#50247853)
M365 is a terrible example
That'll stop em
Yup. Hard to stop individual devs as well… I can plug chatgpt into claude code and vice versa or any configuration I want, there’s no lock in
Re:Anthropic S-1 on Monday October 26, 2026 @10:40AM (#50244554)
500B in a decade is really a lot
Re:Anthropic S-1 on Monday October 26, 2026 @10:31AM (#50244399)
>claude code

The same one that leaked itself a few months ago?

Older »