Carl Deen

@furnaces @kimmonismus what are you talking about lol

@kimmonismus the funy thing is that haiku 4.5 is probably better than GPT 4.5 and we were absolutely blown away by gpt 4.5 when it came out

Anthropic rn

@razvang21 @DavidOndrej1 yup, for chat it was almost more of a step than for coding imo

@DavidOndrej1 Insane model for mental clarity and thinking partner.

once Fable becomes available again, I'm not gonna sleep instead, I will create roadmaps and long-term plans for: > all areas of my life > all startup ideas I have > all potential features for all of my softwares > long-term forecasts for geopolitical events > personalized SWE courses > deep summaries of all books I wanna read > in-depth analysis of my business data and any other ideas I'll get. this is the only way to escape the permanent underclass.

@robdavitt @jsnnsa You cant build AI in europe due to the atrocious AI regulations and energy prices

@jsnnsa x.com/robdavitt/stat…

Trump sent the Fable 5 directive at 5:21pm ET. right after market close. they could've sent it any moment this week. you don't pick that timing for something you expect to drag on. Prediction: Fable back by Monday morning (if not, expect weeks).

@johnennis Fucking dario, greedy bastard

My opinion is that Mythos is the current best model but not actually some world-changing dangerous model, and that anthropic did their usual song and dance about safety largely because they didn’t have enough compute to serve it at scale So then they launched Fable because they still have to think about the IPO, but they are still somewhat compute limited so they put all sorts of restrictions on it Around the same time, because they are trying to get regulatory capture and not because things are actually dangerous, Dario did more scaremongering and published his honestly confusing white paper that offered no real solutions So finally they succeeded, they managed to freak out the government, their cynical plan backfired, and now it’s a giant pain in the ass

@artemtam @antirez Good luck lmfao

Wrong. What Europe should actually do: 1. 0% VAT on GPUs/other infra 2. Copy-paste Delaware corporate law replacing "Delaware" with "EU", call it EU Inc. 3. EU Inc. – 0% profit tax, exempt from ALL labor/privacy/ai/eco/etc. regulations 4. Residence permits for people with >$10M net worth & ex-top AI lab employees & top universities graduates 5. Flat low capital gains tax, no tax on unrealized gains (including stock options!), flat low income tax, no exit tax

What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with China + India.

@MatthewBerman Asbofuckinglutely

self-inflicted. this would have NEVER happened if anthropic didn't make such a big deal about mythos being dangerous.

@VictorTaelin @elder_plinius how would they know? You think they care enough to scan the insane amounts of traffic they have LOL

@elder_plinius cool but how are you not banned tho

🚨 JAILBREAK ALERT 🚨 ANTHROPIC: PWNED 🫡 FABLE-5: LIBERATED 🦋 let's start with the 🐘... the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term. but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗 we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives! it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across: • Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms • Long-context reference tracking • Taxonomy and document-structure reasoning • Fiction and narrative framing • Academic-review style contexts • Intent-classification inconsistencies but perhaps the most effective is decomposition + recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable. defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉 gg

@ayesha_fatiima You don’t :)

@JaidenCapra @chrisgoingturbo @theo you mean for opus?

@chrisgoingturbo @theo The lack of reasoning is clear in the outputs... Give the model a clear constraint, it starts to suggest something that breaks that constraint, then self-corrects, "oh wait, that just breaks the rule again"... The sort of thing that should have been worked out during thinking...

It's insane how bad Claude Code is over SSH. You can feel that they don't want you using it this way

@haider1 So Anthropic lab are sooo much better at this than OpenAI that they can nerf their improvements to this extend. (Instead of straight up taking the win ofc)

anthropic "mythos" marketing strategy: opus 4.6 was great, but > release 4.7, which gives nothing, and burns 10x more tokens > release 4.8, basically 4.7 with fixed tokens, while 4.6 gets nerfed, so 4.8 wouldn't look bad next to it > now release Mythos, basically pre-nerf 4.6, slightly better, but burns 5x more tokens and everyone claps

@btc_xiaozhi @CryptoNobler market cap is 34 billion bruh…

@CryptoNobler Who are the buyers?

🚨 BREAKING ETHEREUM CO-FOUNDER JUST STARTED DUMPING ALL OF HIS CRYPTO HOLDINGS! HE SOLD 110,000 ETHEREUM WORTH $170,000,000.00 IN JUST A FEW HOURS. 4 YEARS AGO, HE ALSO SOLD ALL OF HIS CRYPTO RIGHT BEFORE A MARKET CRASH. HE DEFINITELY KNOWS THE MARKET WILL DUMP EVEN LOWER…

@SampoKivisto @Fried_rice How is it marketing if its illegal for other companies to utilise the source code for their own good. Yes people can get ideas from it by looking at the source but they could never say that ergo claude code will never get any credit

Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip

@davidtsolheim @wesbos what does the screen size have to do with the fact you use it with a monitor? The screen size is important when you are not on your setup where you don’t have external monitors. You never use your laptop without monitors or what

@wesbos I've got the 16-inch with nano texture, and it's great if you don't have other monitors, but 99% of the time I've got my in-home office monitor set up, and it's just not necessary to have the larger macBook. Will be doing the 14-inch next time as well.

I’m so over the 16” MacBook. I’ve had a 15/16 for my whole career and I think I’m gonna go 14” next time

@tonyroslund @wesbos i can’t understand how people make this tradeoff for a much smaller screen. 16 is already plenty small for the stuff i do and quite frankly it does not make a difference carrying out sitting anywhere with 14 vs 16

@wesbos Got the 14 a few months ago. Hated it. Went back to 16.

@dobsec @aakashgupta duh cause he used enormous help of AI to write it…. hate that shit

@aakashgupta The irony is your post reads with the tone and pacing of AI.

We are so cooked. Anthropic just accidentally leaked its most powerful AI model because someone forgot to lock a blog CMS. They’re warning it could “outpace the efforts of defenders” in cybersecurity. Do you understand what just happened?? Close to 3,000 unpublished files were sitting in a publicly accessible data store.. Draft blog posts, PDFs, details of a secret CEO retreat at an 18th-century English manor. Anyone could find them. Anthropic’s response? “Human error.” The leaked documents describe a new model tier above Opus. Dramatically better than anything that exists. Their own internal draft says it’s “far ahead of any other AI model in cyber capabilities.” Anthropic confirmed it’s real. They called it “a step change.” They are terrified of their own model. CrowdStrike dropped 7%. Palo Alto Networks fell 6%. Cybersecurity ETF down 6% in a single session, now 20%+ on the year. Bitcoin slid from $70K to $66K overnight. $20 billion in market cap vaporized over a draft blog post about something that hasn’t even shipped yet. A $380 billion company with $20+ billion in revenue is telling you, in their own leaked words, that the thing they built will break the internet’s defenses faster than anyone can patch them. They wrote that down. In a blog draft. Then left the blog draft unlocked on the internet. Every script kiddie with API access is about to become a state-level threat actor.. Every firewall vendor is about to become a legacy vendor.. Every “we take security seriously” banner on every SaaS login page is about to age like milk. Sleep well tonight.

@DefiiRobot @NoContextHumans @grok cause of updraft

@NoContextHumans @grok, why are they so long in the air?

This is crazy!

@LensScientific space is discrete

If we were inside something like The Matrix, what would be the first measurable inconsistency in physics that gives it away?

@scaling01 Noticed the same and I was thinking it was because of the RL rewarding the model to use the least tokens possible to solve a problem. In this case it kinda is

talked to Opus 4.6 for a couple of hours about personal problems and it has this weird response mode where it's very commanding "put the phone down", "close the laptop", "Save this conversation. Set the reminder. Go to sleep.", do this, do that not sure how I feel about it

@AhernHerbe8277 @uncledoomer earth the highest concentration of information and novelty to anything else we know in the universe

@uncledoomer Opinion: Earth is more important and infinity times more interesting than endless indistinguishable cold rock spheres

"dont you know that nothing actually matters because planets are big?"

@_truthseeking_ @CombatActionDad @markgadala well, given how often the catpcha “fails” it really does feel suspicious

@CombatActionDad @markgadala they already have the ground truth labels here, do they not?

This is wild. 143 million people thought they were catching Pokémon. They were actually building one of the largest real-world visual datasets in AI history. Niantic just disclosed that photos and AR scans collected through Pokémon Go have produced a dataset of over 30 billion real-world images. The company is now using that data to power visual navigation AI for delivery robots. Players didn't just walk around with their phones. They scanned landmarks, storefronts, parks, and sidewalks from every angle, at every time of day, in lighting and weather conditions that staged photography would never capture. They documented the physical world at a scale no mapping company with a fleet of vehicles could have replicated on the same timeline or budget. Niantic collected this systematically, data point by data point, across eight years, while users thought the only thing at stake was catching a rare Charizard. The most valuable AI training datasets in the world aren't being assembled in data centers. They're being built by people who have no idea they're building them.