New Mythos checkpoint shows continued improvement: “On a 32-step corporate network attack we estimate takes a human expert ~20 hours, this checkpoint completes the full attack in 6 /10 attempts.”
can't they solve cancer already?
>>108816621we can have public data leaks instead
>>108816613The only thing I'm impressed with is the fact that LLMs can do ANYTHING in 32-steps. Usually they fail EVERYTHING that requires multiple steps and planning, corrections along the way etc.But let me gues. They are using some kind of harness, IDE, some sort of contraption to hardcode the steps?
>>108816613>our model is so good>show us>no, just trust usfuck off
>>108816683Mythos literally fixed like 1000% more bugs for Firefox last month
>>108816699>just trust us
>>108816683what a clown show indeedcompany 1:>we have this super secret agi model that no one can use, just trust uscompany 2:>nuh uh we have a better super secret model that is a master cyber hackerliteral retards on a playground fighting for hypothetical market cap
>>108816621This, where are OH MY SCIENCE breakthroughs, its all some gay codemonkey shit
>>108816621>>108816728How the fuck could AI speedrun scientific research? Unlike running code, medical research costs money and the government has gutted scientific research funding, including cancer research. We gotta fund a 1.5 trillion defense budget instead.
>>108816740Very antisemitic post, israel won't fund itself
>daily anthropic spam threada reminder that even within the same space, other models have done better without needing to spam themselves everywhere and shoving themselves down everyone's browsing feed.https://daniel.haxx.se/blog/2026/05/11/mythos-finds-a-curl-vulnerability/
May i see it?
>>108816775>>108816822Would he ever lie to you?
>>108816613Turns out despite what retards keep screeching you can actually just scale this shit up. In other words if you parallelize enough monkeys with typewriters it turns into a useful cyberweapon kek
>>108816828Why are all these companies ran by jews again?
>>108816613I'm going to laugh so hard when the inevitable happens and someone is going to use Mythos et al. to attack the AI companies, breach their infrastructure, and leak all the sordid details on their fleecing scams.
>>108816868because they depend on massive amounts of financing to exist, which is the jew's forte
>>108816672>t.
>>108816613Based. Luddites BTFO and seething rn
>>108816621>>108816740The difference is open training data. Open source software is a cultural staple and github can be easily trained on.We don't have a history of open source medical data. Journals/people that publish studies don't need to open source their data. If anything we have the opposite, a culture of privacy and protectionism for medical data and research.
>>108816613Defeated by paying a $100 hooker to give the IT backup operator the “jewish girlfriend experience”https://www.youtube.com/watch?v=zgPLFGaD9fY
most IT infrastructures are held together by spitshine and a fucking dream i look forward to the imminent collapse
>>108818139Open medical data wouldn't actually fix much
>>108816613thatd be 100$ per M output plus tip
i hate when doomers fearmonger about ai when it's not even doing much yet
>>108816672We already know about Claude's 512k LOC script, would be weird for them to suddenly go without one.
>>108817817Go to bed libtard and stfu
