Mike McCaffrey7h ago
Lesley Carhart 
The thing about AI replacing analysts that I think is implicit to a lot of people who understand AI, but not at all clear to the people doing layoffs is this: adversaries use AI for automation and as a force multiplier too. Humans are the deciding difference in who wins, beyond equal automation.
5:00amWeb
Show threadMonsieur_Lepetit7h ago
@hacks4pancakes
"FAFO Tsunami incoming!"
Show threadv7h ago
@hacks4pancakes I even dug through my toots and couldnt find it but I definitely remember wondering how.. different this is for cybersecurity? Someone made a new automated scanner? cool! Why is this revolutionary? It doesnt change how you use tools: they give time, not skill. Yes even without skill you can spam a tool at something and have things happen but youre flying blind and the reason why it wasnt enough to be a script kiddie 10 years ago still apply even when your VERYYY expensive rube goldberg linear algebra doer works as a good scanner
Show threadv7h ago
@hacks4pancakes also the really cool linear algebra doer scanner needs do to so much linear algebra to do a scan it ends up costing 20 grand so idk about the economics of that. Most security holes dont even come from bugs they come from doing something that might be ok in one context but not ok in another, right? Is ai better at *not* doing that?
Show threadjohnny garbanzo7h ago
@hacks4pancakes perhaps one of the best takes I’ve seen in a while
Show threadFabio Tavano6h ago
@hacks4pancakes as if this is just another tool and not the accelerator for the environmental catastrophe..
Show thread🄻🅄🄽🄰4h ago

@hacks4pancakes 'multiplier' fits quite well in the generative output context

generative output can be a useful *multiplier*, if the result can be arbitrary and neednt be precise or deterministic

but brute forcing the most likely output if you need precise / deterministic output

is at best a waste of resources

at worst non deterministic behavior of critical infrastructure that no human operator can understand or fix

Show threadscotty86 🇺🇦🕊️1h ago
@hacks4pancakes Capitalism has a solution for this. 💸
Show threadGary McGraw21m ago
@hacks4pancakes also see https://berryvilleiml.com/2026/04/09/too-dangerous-to-release-again-software-security-and-ai/
Too Dangerous to Release (Again): Software Security and AI | BIML

Have you heard? The mythos model from Anthropic is so dangerously good at finding software vulnerabilities that its rele

Berryville Institute of Machine Learning