🤷‍♀️Christian🤷‍♀️

Infosec, Healthcare, AI... yeah, those things. There is no Trust, only Untrust. Blue Team Forever!!

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@ns123abc We don’t want to ban open-weight models. We just want to reserve the right to decide which “open” models you’re allowed to use.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@AnthropicAI :|
We don’t want to ban open-weight models. We just want to reserve the right to decide which “open” models you’re allowed to use.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@TheSanguineMan @ASDGovAu Same, but then it also asked why Austria is using Signal. And do they make movies?

Soo.... Yeah.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@ZackKorman @DavidSacks 2 years ago a job posting got 20-30 respondents. We just had one break 500. YMMV

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@DavidSacks Hmm... What possible interest could staffing firms and big tech have to say they totally see more jobs coming, and what interest could Bezos have in writing about it.

hmmm....

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@0xPrajwal_ You hate money.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@ZackKorman @rebelcrayon I keep saying that on top of sandboxes it's going to need to be something like a "AI" EDR that lives inside the harness (or is the shell the harness runs in, there are multiple ways to do this).

Sandbox alone, or operating on the API isn't good enough.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

So, does this make Anthropic "adjudicated plagiarists" now?
https://t.co/fxzJsUMJPW

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@HackingDave Been saying this for a while.... Based on a pre LLM reading of US copyright law, they should be. Either due to having GPL licensed code, or the vast number of copyright holders to the training data.
Public domain that shit, baby!
Congress could do it very easily.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@maxgeithe @theo After much debate starting out as poison SRS and moving to balormage's holy relic... Maybe.
Might try Luminary, once we see how it shakes out.
Very much "dad gaming" style. :)

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@its_andyd @theo Even better... Because I am dithering on which build to start and was throwing more options at the models. Fable actually caught a pretty niche gotchas that weren't called out in the build guide, and that required game knowledge. Interesting...

Opus 5 is really slow right now

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@signulll It's not surprising that OpenAi, Anthropic, Google, Oracle, Apple, Amazon didn't.

What is interesting is you don't see Cisco, Intel, Red Hat, Palo Alto, Intel, AMD, etc.

Also, would think EFF, OSI, etc. probably weren't asked. (but probably should be).

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@ZackKorman @HeidyKhlaaf Completely unsurprising. Of course they let the researchers do whatever they want (mostly). We see the same thing in healthcare with doctors. They are the ones that generate the income, AND the ones that drive the decision making/policy/exception progress.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@vossbuilds For now....

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@kyrylo People Skills, Context.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@DebugPrivilege literally just gave me back pain.

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@psomkarbuilds Let's think about this. Could a typical Hollywood Producer use AI to make a movie (basic spec script and Storyboard/Animatic, full budget analysis) in 3 days?

And if they could, would it sell?

Hmm...

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@ben_sage Claude's guesswork:

[🤷‍♀️Christian🤷‍♀️ @roguecyber]

@0xPrajwal_ painfully.