mroczekdev

@Alexvx_nft This is useful advice, not generic one you get from llm

if you're <5-10K and feel stuck on x check this 🧵

@AnthropicAI Please don’t screw it up!

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

@scaling01 It speeds up kernel task by 400, the llm training is 51 times

Mythos speeds up AI research by up to 400 times A 300X speedup over the baseline requires 40 hours of work by a human expert It also clears the >8h threshold of human equivalent work time on ALL tasks!

@sandislonjsak okay, give me your money xD

Okay so Anthropic's Claude Mythos is so good they are sharing it with big companies so they can patch vulnerabilities before releasing it to general public? Who will give me my 15 days to safely fix vulnerabilities in my apps which Mythos will expose to everyone? Chat, are we cooked?

@AlexFinn For people having problems with benchmarks. TLDR:

Good news: Anthropic just revealed Mythos- the most powerful AI model ever made Bad news: you'll never be able to use it I get it. It's so powerful that it could exploit cybersecurity But I hate it. I don't love that a company gets to hand select who gets to use the best intelligence. The companies who get access to Mythos will have a distinct economic advantage against those that don't That feels unfair I'm more of a fan of democratization of intelligence. This feels like an opportunity for OpenAI to release something as powerful but put it in the hands of consumers. Trust the consumer by default. Sort of like with the OpenClaw situation Another reason to root for open source

Claude mythos is a sick model! Unfortunately, it will be now only in the hands of big tech as a security measure, not form of class opression. Therefore, I present my new idea for security of "too good models" Solution: Special ethical developer certificate obtained from the hands of the pope. This is the only option to release safely to the public

@sandislonjsak this sounds like startup idea @garrytan can we start a discussion?

@chatgpt21 now in the hands of big tech, to play to fuck around and find out, how good it really is

Machines of loving grace coming a little sooner than expected 😅

@scaling01 Have and have nots, anthropic I thought you had our back

The permanent underclass began today Claude Mythos won't be available to the public, but only billion dollar companies, governments, researchers, ...

@bcherny I just wonder is it fair to rest of developers community. It implies that there won't be bad actors in big tech

Mythos is very powerful, and should feel terrifying. I am proud of our approach to responsibly preview it with cyber defenders, rather than generally releasing it into the wild. Model card here: www-cdn.anthropic.com/53566bf5440a10…

@suni_code dystopia wasn't on my bingo card

Elon Musk warns "Instagram is for low IQ peoples"

@pmarca I am a little worried that the keys to castle will be in the wrong hands. With Claude Mythos only the big tech gets the keys and we have to trust them with access

Every security flaw discovered by AI was there before AI, waiting to be discovered either by people or by AI. The world has never been good at securing computer systems; finally with AI we are going to get good.

@pmarca Thank your for bringing always important topics to the surface I would like to remind everyone: Grand Theft Auto 6 (GTA 6) is officially scheduled to be released on November 19, 2026.

Overheard in Silicon Valley: “Remember ‘hate speech and misinformation’? That was weird.”

@deedydas cnbc.com/2026/04/07/ant… too bad that only big tech will be able to use it for now. Sounds like the real division will be not claude vs. claude max, but have and have nots

Claude Mythos just obliterated every single benchmark in AI. I can't believe what I'm reading.

@arvidkahl this seems too good to be true, but I hope so

yeah that's pretty impressive

@garrytan I really enjoyed watching this interview. It always felt that YCombinator is just VCs, but it has so many stories about being human

Tech gave me everything I have Its capacity to lift people into abundance is incredible and there is nothing like it We must make that into prosperity for everyone

@GergelyOrosz they've been leading coding models for some time already and their have particular "flavour" with their llms that makes them especially nice to use

Of course, things change so fast with AI models that it’s hard to project what the situation will be, even a few months out. Competition breeds more competition - Anthropic (a company founded only 5 years ago!!) is a perfect example

Two years ago, if you asked me which lab will be the first to say: “this AI model is too powerful to release, so we’ll wait with it” - my guess would have obviously been OpenAI. Who else? That Anthropic got here first shows how quickly they’ve become the front runner AI lab.

@GergelyOrosz the pentagon controversy was the best marketing they could ever get

@jasonfried It might be the thing that Conway's Law is based on. Sometimes those higher structural flaws are impacting everything

High vs. Low Problems Problems are often described by their size. "Hey, that's a big problem". Or "Eh, that's just a little issue, no big deal." And if you do hear someone say “serious” you immediately think huge. It’s still about size. I sometimes still use those descriptors, but I'm trying to use them less. Instead, I’m reframing problems as High or Low. High problems seep. They stain. They coat what's below, like melted frosting dripping down a cake. Gravity always wins, and high problems eventually become low problems too. You see this in relationships. Little problems don't add up to big ones - it's the big ones that create the small ones. Something small really annoying you about someone? Good chance it's because something bigger is annoying you more. High problems leaking down, staining what’s underneath. You see this in organizations. People bickering about flaws in someone's work or style? It's not because all is right above - it's because something's spilling from above. Bad hiring, lax oversight, proper examples not being set. High problems are like water hitting dry sponges — they don't just soak in, they make everything swell and distort, turning small issues into bigger ones than they should be. You see this in product design. Big picture complexity trickles down into every crease and crevasse. It’s hard to be simple and straightforward locally when you aren’t globally. The right design right here can’t be applied because predetermined guidelines from above promote consistency rather than context. You can’t just link something up because all nav has to be a button according to a larger system someone else already defined. While it's possible for small things to pile up and color what’s above, that direction is rare. So when you encounter low problems - especially ones that wouldn't seem noteworthy to an uninvolved observer — it's safer to assume they're symptoms of something bleeding from above, rather than growing from below.