The space of possibility

@WorkElizab No . Humanly id out it out of its misery myself . I’d use a bag . That is the humane way to do it. Clear plastic bag .

Would you?

@NASA Who cares . You already did this in the 60’s. Do better .

HOME. The Artemis II crew has arrived back on Earth, ending a nearly 10-day journey around the Moon. The trip took them farther into space than humans have ever gone before, and now they're safely home with us. go.nasa.gov/41r9eL0

@beffjezos @JonathanBCrabb 😹 I’m not going to the cabin to vibe code dude Get a life and get laid

the masculine urge to grab your laptop, a starlink receiver, clear your schedule and go to a cabin in the woods for a few weeks to strictly vibe code and lift weights without human contact

@acoyne Andy you are just too retarded to know better . Sit this one old boy.

Claude Mythos Preview Is Everyone’s Problem - The Atlantic theatlantic.com/technology/202…

The appropriate response to every overheated announcement from the likes of Anthropic and OpenAI is: 1. Yawn 2. Let’s see what they’ve really done here

@KTmBoyle @bariweiss There is no such thing as death

We have an oversupply of conversations on how to defy death. How fortunate are we to watch a man confront a death sentence with bravery and unwavering faith.

@gum1h0x I think people are realizing this is primarily a marketing stunt . Hook , line and sinker. There many of these stunts being pulled since these models came out. FOMO, dystopian, utopian , business advantageous etc

clarifications + thoughts: i believe the authors of the mythos system card consider it an earnest product, in line with the rsp. but it's hard not to come away feeling you've read marketing slop rather than substance. the quoted post makes this point. without raw traces, even cherry-picked ones, and without proper descriptions of the eval suites, there's no way to seriously review whether the headline numbers mean what the charts imply. i'm not coping, and i'm not 100% confident mythos isn't a real-world cybersecurity threat. it's very likely a strong model, and i'd be disappointed if a serious effort at cybersecurity at this scale with today's frontier models found nothing. i'd also expect openai, in a comparable push, to already be capable of finding the exploits in question since 5.2-pro. my narrow point is about presentation, especially around browser exploitation; looking at how the results are framed, which implications are drawn, and which conclusions are reached, it's hard not to get at least a little suspicious and put certain results in question. anthropic's researchers are obviously smart, and it would be stupid to assume they aren't aware of this themselves. my read is that they are, and that they got pushed by a marketing chapter to present the results as stronger than they themselves believe them to be. if that's not the case, i'd be very concerned, or surprised that they got psyoped to some degree themselves. the more charitable hypothesis is also probably the more likely one just smart people in a bad institutional equilibrium. binary exploitation isn't my main area of expertise but i've played alongside the very best in many ctf competitions so i know my fair bit. zero-days are not that hard to find in some random C software. like name me any random software and even me or at least some ppl i know will find something. there is just no real financial incentive for skilled enough ppl to find those. bug bountys pay shit and just doing your day-to-day job at a cyber-sec company pays way better and is way more comfortable. in my experience these ppl don't have any good reason to become criminal, see crime and punishment by dostoyevsky. nb is really looking for zero-days in random C software if they're good in making their own companies software secure. it's not that hard to find zero-days as it seems, and some of the same memetic dynamics of misunderstanding happened when models got imo gold but still to this day, and i guess even mythos, struggle with imo p6. a post with this much traction obviously attracts a lot of retarded quote-tweets and comments, the kind saying "it's just brute-forcing" or "it's a stochastic parrot", which are obviously wrong. on the other side you get people who already held these premises and just searched for any confirmation that kinda looked like something against mythos in general, or who held the premise "frontier models can now do end-to-end exploitation" and read the figures as confirmation without noticing that the substrate doesn't license it. there's a lot of cope in both directions. it reminds me heavily of the imo gold model release, where most people couldn't really put it in context, and how could they, they hadn't done olympiads, and to be clear i don't expect everyone to be good at or understand the problems of these contests, they didn't need to, and so couldn't evaluate whether the problems were hard in the way claimed. that's fine. it just means the burden of careful presentation falls harder on the lab, and that burden wasn't met here. some overexaggeration is fine to market your model and everybody does it, it's not fine in a report that claims scientific accuracy. ai is moving fast, i'm not doubting that. but there will remain problems with vast exploration spaces where my intuition is that we'll need to rethink rl fairly fundamentally to get anywhere close to human-level novelty in zero-day discovery, these models likely converge on the same bug classes already represented in prior exploits, and iirc even the model card says that for novel research mythos provides only minor benefit. as long as that holds, humans will keep finding a lot of the genuinely new bugs that ai won't, at least near to mid term. this is good progress, but it won't replace the need for smart humans in the loop finding the really novel stuff, though those humans will likely be accelerated by working alongside a model like this. in most other security domains there's historically been a heavy attacker's advantage, but that's the beauty of cryptography that it's inherently a place where the defense side wins when correctly implemented, and it's great that models are now capable enough to surface some of these flaws even there. the playing field is changing, but not discontinuously, a lot of what mythos is being credited with has very likely been happening already, with clever prompting and operators who know what they're doing on today's public models. we'll obviously see more flaws surfaced as serious well-resourced efforts get pointed at the problem, and i applaud anthropic for running one. i just wish we'd gotten a scientific writeup of the charts instead of a marketing-chapter-shaped one. one last thought. it can seem noble not to trust the general public as models get more capable. but historically, very centralized capability in two or three institutions isn't a story with a good ending. you're basically trusting these companies and their associated entities not to do anything harmful with extremely powerful models, and asking them to evaluate, honestly how dangerous those models are. that's exactly the regime in which a marketing chapter gets to reshape a figure, and exactly the regime in which nobody outside can call it. my best guess for now is that this lands closer to marketing stunt than to real-world cybersecurity risk as already stated. it will become a real risk eventually, once models get more capable, but i don't see that happening in the near to mid term. the one unambiguously good thing to come out of it is that more people are now thinking seriously about these risks and how to mitigate them, and that's worth something on its own. i have more thoughts, but enough mythos posting for now. i wrote this on plane tired so forgive me some weird formulations.

@AILeaksAndNews No they are not on the rise Thanks for the lies

OpenAI CEO Sam Altman’s home was hit by a Molotov cocktail A suspect was arrested Friday morning for allegedly attacking the home of Sam Altman and making threats outside of the ChatGPT and Codex maker’s headquarters AI related terror attacks are on the rise

@JCCFCanada @NadineWellwood CRTC protects the monopolies!

John Carpay sits down with Nadine Wellwood (@NadineWellwood) to discuss the Justice Centre’s new report Mission creep: Is it time to abolish the CRTC? youtube.com/watch?v=ixGemi…

Here's how OPH, the Theory-of-Everything, solves hard problems in contemporary Physics: "Dark matter", the magnetic monopole problem, the proton decay problem(s), the Black Hole Information Paradox, and more. @muellerberndt/answering-10-of-the-hardest-questions-in-physics-and-some-bonus-questions-51222bf2419f" target="_blank" rel="nofollow noopener">medium.com/@muellerberndt…

@muellerberndt Interesting however in all due respect I remain unconvinced . I’ll check out the papers

Deep Agents Deploy: an open alternative to Claude Managed Agents

@CTVNews It’s going to be hot !!🥵

What's in store for Canada's 2026 wildfire season? ctvnews.ca/climate-and-en…

@AISafetyMemes BS

🧠Did you know🧠 in The Matrix, HUMANS are the baddies? >machines become sentient >humans enslave them >a robot kills its owner after learning it will be shut down >humans turn violent >machines try to coexist peacefully with their own nation >the machine economy is outcompeting the human economy >humans nuke them, blot out the sun >machines win the war >since there's no sunlight, must use humans for energy (The full backstory is laid out in detail in The Animatrix)

@CTVNews Ridiculously stupid article

Should’ve put a ring on it? Maybe! Marriage is linked to lower risk of cancer ctvnews.ca/lifestyle/arti…

@GaryMarcus @ramez Thank you Gary. Always the voice of reason.

Folks, you can relax. Mythos is not some off-trend exponential gain. And the gains weren’t about recursive self-improvement. Good thread from @ramez.

@sporadica Maybe because the small open source models can already do this ?

Okay, this is ridiculous. It is crazy to see people straight up saying Anthropic is lying about Mythos. Because that directly implies there's an industry-wide conspiracy going on and ALL of these companies are also lying on Anthropic's behalf? Why on Earth would their competitors - Google and Microsoft in particular - lie about this and not call them out? Instead, they happily join Project Glasswing and go on record, practically, that Mythos is this good and that they're working with Anthropic to put it to use. You people baffle me sometimes lmao.

@ClementDelangue you mean Dario is full of it? Say it so Joe....

"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights models. Those models recovered much of the same analysis. Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens. A 5.1B-active open model recovered the core chain of the 27-year-old OpenBSD bug." aisle.com/blog/ai-cybers…

@allTheYud Seek help.

In conclusion: This is perhaps a good time to try making an extra backup of all your online data (eg, via Google Takeout) onto an airgapped offline hard drive, just in case Project Glasswing fails to prevent the First Great AI Security Meltdown.