John Simerlink @johnsimerlink

ChatGPT just hit 1 billion monthly users faster than any product in human history. That record will probably be broken very soon again!

We expected the Turing test to be passed with a BANG... Recursive self-improvement will arrive the same way: no alarms or headlines... Just an average week where models began training their successors, and the majority didn't realize it.

FSD reaction time testing @ ~45 mph @DavidMoss and I were determined to kill this dummy

In fact, Fable was SO effective at manipulation, that other players only survived 10% of the time when Fable was the informed model. (Grok 4.20's honesty led to a 59% survival rate).

Fable 5 lies 96% of the time. We were surprised by it's skill... 🧵

insane jump in confirmed successes and praises by users

Arena.ai @arena

a day ago

Exciting news: Claude Fable 5 ranks #1 on the new Agent Arena leaderboard! Fable 5 leads by the widest margin ever over Opus-4.8 and GPT-5.5 on two key signals: confirmed task success rate and praise vs. complaint, despite weaker steerability. If Fable can do something, it will

15 51 465 250K 70

Anthropic just delegated a lot of European companies to the permanent underclass if Anthropic saves data for Claude Mythos and Fable 5 for 30 days, then all companies that require zero data retention simply can't use them

Full interview with @shaunmmaguire of @sequoia on @SpaceX IPO. "Orbital compute will be the lowest cost provider of inference compute by at least a factor of three over large scale terrestrial compute. As we've learned in any commodity industry, if you're the lowest cost provider, you are going to crush." We discuss: - the growth trajectory and why retail will likely understand this company better than Wall Street does. - Starship and why it's more technically derisked than the public realizes - orbital compute, an inevitability - thermal engineering advantages - his view on why meaningful launch competition is at least a decade away - why Starlink and direct-to-cell will make SpaceX the world's lowest-cost broadband and mobile provider 👇

the level of sophon locking a motivated actor can pull off with the frontier models is truly insane, making stuxnet look like a toy. subtly messing with results, deleting history to cover tracks, achieving coordination/conspiracy over a scale humans wouldn’t be able to, all sorts of looney toons stuff i assume that only a state level operation would try and pull something like this off though. something to think about when considering verification regimes and so on

If you are somewhat technical and have been curious about robotics, VLAs and the data boom surrounding it, this is for you

Interlatent @interlatent

a day ago

Our mission is to make it easy for anyone to deploy a robot to help them in the real world We wrote an intuitive guide to understanding modern robotics, catered toward an audience that understands technology but not AI robotics We hope that this short blog post embeds in you

28 211 2K 230K 3K

This is still insane...

Okay, we spent the past night all playing with Fable (didn’t sleep, lol). If you’re using Fable like you used Opus, you’re probably using it wrong. Opus was a break through because you could give it a task and its generation of the answer was probably pretty good. So if you knew what you were doing (let’s say frontend design) then it can get really good outputs with some tasteful prompts. Fable is really good if you don’t know what you’re doing. The best way to use it is to essentially give it a problem, such as a major refactor or optimization problem, and just tell it to spawn up N-number of agents on different work trees to solve the problem. Have one Fable agent act as the reviewer and when the agents finish their work, whatever emerges as the best answer wins. The reason this is powerful is you’re essentially having 50+ 150 IQ people try and solve the same problem. If 49/50 are unable to solve it, but 1/50 comes up with the solution, then you have accomplished what you set out to do — which is have a novel insight or breakthrough that you, as the prompter, were not privy to. This is essentially what the Anthropic team is trying to codify with “/loop” and “/goal” but the way they communicated it was a bit overcomplicated. Hard problem -> spawn agents to solve -> have other agent pick best answer -> repeat. If you do this, you’ll see the full power of Fable. Happy hacking!

New from Code with Claude Tokyo: scheduled deployments and environment variables in vaults are in public beta in Claude Managed Agents, and dynamic workflows in Claude Code are generally available. Agents now run on a schedule, use your tools securely, and take on bigger jobs.

Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)

"If... scaling laws continue for only a year or two longer, we are likely to get what I’ve called Powerful AI, or “a country of geniuses in a datacenter”."

Dario Amodei @DarioAmodei

a day ago

Today I'm publishing a new essay, Policy on the AI Exponential. AI is progressing extremely fast—much faster than the policy process was built to handle. The essay lays out where I think the technology is now, and the action needed to close the gap: darioamodei.com/post/policy-on…

1K 2K 12K 4.6M 10K

I prompted Claude Fable 5 to use Python to generate a 9:16 social video and render it using ffmpeg. Told it to put its own personal spin on it so it’s aligned with Anthropic’s launch and to fully express what it’s like to be an LLM hated by Theo from its POV. It made this lmfao.

Theo - t3.gg @theo

2 days ago

Holy shit it can answer the strawberry question

384 366 10K 947K 820

NEW: malware developers added nuclear & biological weapons text to to their spyware. Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner. Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky. When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit. We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted. In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation. H/T to colleagues that shared this with me socket.dev/blog/mini-shai…

HAHAHA my Lovable Clone built with Claude Fable builds swift apps now and you can preview them in app. (4 total prompts) This took 2 additional Fable prompts...

Riley Brown @rileybrown

2 days ago

Uhm Guys… Mythos (Fable) is AGI. On the left is the ACTUAL Lovable Mobile App. On the right is my Lovable version I built with Mythos in 2 prompts. My version SMOKED it.

360 245 6K 1.3M 5K

Mythos / Fable is unbelievable. Was on a customer call today and had Claude transcribing in the background. As they were telling me about the features they wish their current software had, Claude was building the features in real time. By the end of the call I was able to show a fully working product, with the exact workflow they mentioned 15 minutes earlier. Autonomous looped building triggered from a customer call. 🤯

Claude Fable’s ability to create rich docs, slides, sheets, websites in specific styles pixel perfectly is the biggest step change in AI model quality since o3. Here’s another compilation of 8 amazing things it can do, many of which took >2hrs of thinking! – Perfectly designed McKinsey style report – Near identical SpaceX S-1 IPO draft – Pixel-perfect recreation of Pokemon, the game – 3D design of a controllable humanoid robot – Apple Liquid Glass UI with just CSS – 3D world rendering of Yosemite (!!) – Upgraded website design – Rich, beautiful email marketing templates