underdeserver

7 ms·

61.
by underdeserver 6mo ago
I would suggest watching Nicholas Carlini's talk and Heather Adkins and Four Flynn's talks from unprompted: https://youtu.be/1sd26pWhfmg?si=onOai_ocxkZeNWP0 https://youtu.be/B_7RpP90rUk?si=HkRBhw95
62.
by underdeserver 6mo ago
Interesting also is what they didn't find, e.g. a Linux network stack remote code execution vulnerability. I wonder if Mythos is good enough that there really isn't one.
63.
by underdeserver 6mo ago
Key point: available to participants .
64.
by underdeserver 6mo ago
It's not just about cost - a $2m drone and a $200m drone can both be sacrificed if cost/benefit analysis merits it. You don't sacrifice pilots, ever.
65.
by underdeserver 6mo ago
Just wanted to point out that tptacek is the blog post's author (and a veteran security researcher).
66.
by underdeserver 6mo ago
Specifically in software vulnerability research, you mean. Fixing vulnerable code is usually trivial. In the physical world breaking things is usually easier.
67.
by underdeserver 7mo ago
Where in the FAQ?
68.
by underdeserver 7mo ago
I'm basing that opinion on the FAQ that states that most politics stories are irrelevant. But sure, I'm one vote among tens of thousands, and it's up to the mods to decide. It's most of my comment history recently beca
69.
by underdeserver 7mo ago
> Qatar, Kuwait, UAE, Bahrain is asking questions, if US can't defend us and is moving all defensive missiles to protect Israel, why should we even be ally with them Where are you getting this information? The UAE, for instance, is
70.
by underdeserver 7mo ago
Irreparable? That's a big claim. Long term I think if the regime in Iran changes and better economic cooperation becomes possible everyone will benefit.
71.
by underdeserver 7mo ago
I am saddened by all of these comments. Will not one of you try to steelman this decision? Or do you truly, fully believe the entire US government and intelligence complex, supported by roughly 50% of your compatriots, are warmongering babo
72.
by underdeserver 7mo ago
Truth social post: https://truthsocial.com/@realDonaldTrump/posts/1161445529692... Six-month phase out period, that's something. I guess the DoW's been using Claude quite a bit.
73.
by underdeserver 8mo ago
Can you point to any pro-Israeli posts on HN since October 7, flagged or not?
74.
by underdeserver 8mo ago
Mark Russinovich and Scott Hanselman are both prolific engineers and content creators. Russinovich wrote SysInternals, which I remember using like 25 years ago to figure out what ran on startup. Hanselman has a pretty popular podcast, Hanse
75.
by underdeserver 8mo ago
What a glorious utopia we could have
76.
by underdeserver 8mo ago
> when agents started to compile the Linux kernel, they got stuck. [...] Every agent would hit the same bug, fix that bug, and then overwrite each other's changes. > [...] The fix was to use GCC as an online known-good compiler o
77.
by underdeserver 8mo ago
jinx
78.
by underdeserver 8mo ago
Anthropic's version is in Rust though, so at least a little different.
79.
by underdeserver 8mo ago
> At a bare minimum, the agent must have the ability to: read files, execute programs, and make HTTP requests. That's one very short step removed from Simon Willison's lethal trifecta.
80.
by underdeserver 8mo ago
I still remember old bioweapons threat levels. Level 1 was no threat. At level 3 everyone was required to shave.
81.
by underdeserver 8mo ago
I don't think you can really learn from this experiment unless you specify which models you used, if you tried it against at least 3 frontier models, if you ran each eval multiple times, and what prompts you tried. These things are non
82.
by underdeserver 8mo ago
Should be (2024).
83.
by underdeserver 8mo ago
Probably should be marked (2025).
84.
by underdeserver 9mo ago
You can have memory corruption in pure Go code, too.
85.
by underdeserver 9mo ago
I'm sorry what
86.
by underdeserver 9mo ago
For those who missed it, tptacek wrote TFA.
87.
by underdeserver 9mo ago
Looks like it doesn't compile for at least one other guy (I myself haven't tried): https://github.com/wilsonzlin/fastrender/issues/98 Yeah, answers need to be given.
88.
by underdeserver 9mo ago
On Twitter people are saying GPT-5.2 is better. That's also what Cursor used in their testing. Maybe try it?
89.
by underdeserver 9mo ago
Technically correct, but a better benchmark would be a known compressor with an unknown set of inputs (that come from a real-world population, e.g. coherent English text).
90.
by underdeserver 9mo ago
"By jamming GPS and satellite signals, authorities have reportedly degraded Starlink performance, with data suggesting packet loss has reached 30%."

More ›