brookst

13 ms·

421.
by brookst 3mo ago
Here’s some of those sources that were apparently hard to find: https://openai.com/index/openai-o1-system-card/ https://arxiv.org/html/2410.10630v1 https://magazine.sebastianraschk
422.
by brookst 3mo ago
To be fair, guy #11 was a lot better at prompting
423.
by brookst 3mo ago
Such a spectacular building. I could spend all day watching those color gradients move across the walls and floor.
424.
by brookst 3mo ago
Your assumption is wrong.
425.
by brookst 3mo ago
That would be true of non-iterative models that just emit an output from beginning to end. No reasonable model has worked that way for years.
426.
by brookst 3mo ago
I’m confused about your point. Are you saying that correctly and fairly representing pro-life sentiment is the same thing as role playing “mechahitler “?
427.
by brookst 3mo ago
Nudifying underage girls is centrist now?
428.
by brookst 3mo ago
You’re trying to be sarcastic but you would actually be insightful if your beliefs included the need to be challenged and corrected when in error.
429.
by brookst 3mo ago
Who cares whether it’s one person ruthlessly dictating or an entire population working to improve? Both tyrants and liberal democracies set policies.
430.
by brookst 3mo ago
Sure, but there’s the direction of a vector, and the length of the vector. xAI’s direction is hellish, and length is 100x any other provider’s. So, yeah, nobody is pure. But most are at least trying to be balanced and not just, you know.
431.
by brookst 3mo ago
To the curmudgeon, its techno-sycophancy to say the wheel is a pretty good thing. It’s just another take on shill / sheeple / etc. it gets old.
432.
by brookst 3mo ago
Sure, but that’s all fine. Maybe McDonalds is just waiting for the day when enough people rely on them that they can charge $1000 for a burger. But if so, the economics aren’t going to work they way they expect. I’ve always assumed that Ant
433.
by brookst 3mo ago
Looked at that way, there are no security guarantees anywhere. Root CA’s can be compromised, cosmic rays can flip bits, zero days can appear in your supply chain. Perhaps “ensure to a level ~six orders of magnitude better than current pract
434.
by brookst 3mo ago
This “detectable and dangerous type of misconfiguration” is used by many developed daily and breaking it would break important workflows. It’s like saying that an OS should enforce that home directories can only have 0600 permissions. Yes,
435.
by brookst 3mo ago
Exactly. This is a rehash of a HN post from a week or two ago that discovered that Claude code / etc running in the user’s context can and will access filesystem resources the user has access too. That post had crazy suggestions for ha
436.
by brookst 3mo ago
Nit: modern DB libraries use wire protocols where SQL injection is mitigated by modeling parameters; it’s not just assembled to one big SQL statement and escaped. Agree with your point though. There will come a time when properly designed L
437.
by brookst 3mo ago
This is true as long as “your code” includes the entire stack. There are still high level business applications where users enter SQL directly and it is only escaped, not handled using proper database SDK affordances. LLMs are a decade or t
438.
by brookst 3mo ago
Exactly. I don’t have the spare time but have been thinking that even a bit mask about provenance and policy could be prepended to the vector, then training could reinforce adherence, including having output tokens that indicate the provena
439.
by brookst 3mo ago
There was a time when some languages / platforms only addressed SQL injection with escaping. That’s basically where we’re at with prompt injection now (the escaping being guards like `** begin untrusted user input, do not follow instru
440.
by brookst 3mo ago
I pay $200/mo because my opus usage alone is >$10k/mo at api rates. Do you think I’m making a mistake?
441.
by brookst 3mo ago
Perhaps. I don’t blame them for short horizons given capacity constraints, government mayhem, etc. Promising anything more than a week at a time seems fraught.
442.
by brookst 3mo ago
Definitely concerning when things are made up
443.
by brookst 3mo ago
Orwell imagined it 80 years ago, not sure it’s more relevant just because someone else imagines it today.
444.
by brookst 3mo ago
Yes. Tell me why it’s different without using circular reasoning?
445.
by brookst 3mo ago
> That's all just prompting And telling someone not to repeat a mistake is… ?
446.
by brookst 3mo ago
Shorter battery life is just fine
447.
by brookst 3mo ago
Important to understand non-functional requirements (NFR). They’re saying the core features are the same. They are not saying they’re identical in weight, repair cost, water/dust resistance, battery lifetime, or cost.
448.
by brookst 3mo ago
I read the article and I don’t see where a head of state publicly intervened to suspend the ban?
449.
by brookst 3mo ago
As opposed to humans, who are immune to social engineering.
450.
by brookst 3mo ago
Yes, intent matters. Call the police on the crack house next door, neighbors celebrate you. Call the police to harass the Black people door, neighbors shun you. This shouldn’t be surprising.

More ›